From 383bdca0b4d20ad07527ee6c69f2aad1739778de Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Wed, 25 Mar 2026 16:44:32 -0400
Subject: [PATCH 01/22] DataFusion 53 Release Blog

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 337 +++++++++++++++++++
 1 file changed, 337 insertions(+)
 create mode 100644 content/blog/2026-03-25-datafusion-53.0.0.md

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
new file mode 100644
index 00000000..66a1f72c
--- /dev/null
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -0,0 +1,337 @@
+---
+layout: post
+title: Apache DataFusion 53.0.0 Released
+date: 2026-03-25
+author: pmc
+categories: [release]
+---
+
+<!--
+{% comment %}
+Licensed to the Apache Software Foundation (ASF) under one or more
+contributor license agreements.  See the NOTICE file distributed with
+this work for additional information regarding copyright ownership.
+The ASF licenses this file to you under the Apache License, Version 2.0
+(the "License"); you may not use this file except in compliance with
+the License.  You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+{% endcomment %}
+-->
+
+[TOC]
+
+We are proud to announce the release of [DataFusion 53.0.0]. This post highlights
+some of the major improvements since [DataFusion 52.0.0]. The complete list of
+changes is available in the [changelog]. Thanks to the [114 contributors] for
+making this release possible.
+
+[DataFusion 53.0.0]: https://crates.io/crates/datafusion/53.0.0
+[DataFusion 52.0.0]: https://datafusion.apache.org/blog/2026/01/12/datafusion-52.0.0/
+[changelog]: https://github.com/apache/datafusion/blob/branch-53/dev/changelog/53.0.0.md
+[114 contributors]: https://github.com/apache/datafusion/blob/branch-53/dev/changelog/53.0.0.md#credits
+
+## Performance Improvements 🚀
+
+DataFusion 53 continues the project-wide focus on reducing planning overhead,
+skipping more unnecessary I/O, and pushing more work into earlier and cheaper
+stages of execution.
+
+**TODO: Add performance chart and short discussion of the benchmark setup and results.**
+
+### `LIMIT`-Aware Parquet Row Group Pruning
+
+DataFusion 53 includes a new optimization that makes Parquet pruning aware of
+SQL `LIMIT`. If DataFusion can prove that an entire row group matches the
+predicate, and those fully matching row groups contain enough rows to satisfy
+the `LIMIT`, it can skip scanning partially matching row groups entirely.
+
+This is important because it turns `LIMIT` from a result-shaping operator into
+an I/O optimization. For selective queries on Parquet, this can avoid decoding
+entire row groups that would previously have been scanned only to discover that
+their rows were not needed.
+
+For example, consider:
+
+```sql
+SELECT *
+FROM tracking_data
+WHERE species LIKE 'Alpine%' AND score >= 50
+LIMIT 3;
+```
+
+Conceptually, DataFusion 53 can now do:
+
+```text
+Before:
+  scan partially matching row groups
+  scan fully matching row groups
+  stop after 3 matching rows
+
+After:
+  identify fully matching row groups from statistics
+  if they already contain >= 3 rows, scan only those row groups
+```
+
+For a deeper explanation, see the dedicated [limit pruning blog post].
+Related PRs: [#18868]
+
+### Broader Filter Pushdown Across Joins and Unions
+
+DataFusion 53 also pushes filters through more plan shapes. In particular,
+predicates can now be inferred and pushed through more join types, and filters
+can be pushed through `UnionExec` without being redundantly re-applied.
+
+This matters because the cheapest rows are the rows you never read. Better
+pushdown means fewer rows flow into joins, repartitions, and later operators,
+which reduces CPU, memory, and I/O.
+
+For example:
+
+```sql
+SELECT *
+FROM t1
+LEFT JOIN t2 ON t1.k = t2.k
+WHERE t1.k = 1;
+```
+
+Now DataFusion can often transform the physical plan so the filter is applied
+to both sides before the join:
+
+```text
+Before:
+  HashJoinExec
+    DataSourceExec(t1)
+    DataSourceExec(t2)
+  FilterExec(k = 1)
+
+After:
+  HashJoinExec
+    FilterExec(k = 1) -> DataSourceExec(t1)
+    FilterExec(k = 1) -> DataSourceExec(t2)
+```
+
+Related PRs: [#19918], [#20145], [#20192]
+
+### Faster Query Planning
+
+DataFusion 53 reduces planning overhead by caching `PlanProperties`,
+reusing them when plan children are unchanged, and making immutable pieces of
+execution plans cheaper to clone.
+
+This work is important for applications that generate many queries, repeatedly
+optimize similar plans, or build plans programmatically. In those environments,
+planning cost can become user-visible even when execution is fast.
+
+Conceptually:
+
+```text
+Before:
+  rebuild plan node
+  recompute PlanProperties
+  clone large immutable vectors
+
+After:
+  reuse cached PlanProperties when possible
+  share immutable plan state with Arc
+```
+
+Related PRs: [#19792], [#19893]
+
+## Major Features ✨
+
+### Nested Field Pushdown for Cheap Expressions
+
+DataFusion 53 adds a new optimizer pipeline for pushing cheap
+"move-towards-leaf" expressions such as `get_field` closer to the data source.
+This allows nested field access to be extracted earlier in the plan rather than
+materializing wider values and trimming them later.
+
+This is especially important for nested data such as structs. Instead of reading
+an entire struct column and then extracting one field near the top of the plan,
+DataFusion can push the field extraction down so only the needed data is carried
+through the pipeline.
+
+For example:
+
+```sql
+SELECT id, s['label']
+FROM t
+WHERE s['value'] > 150;
+```
+
+The optimizer can rewrite the plan roughly like this:
+
+```text
+Before:
+  ProjectionExec: get_field(s, 'label')
+    FilterExec: get_field(s, 'value') > 150
+      DataSourceExec: projection=[id, s]
+
+After:
+  ProjectionExec: __label
+    FilterExec: __value > 150
+      DataSourceExec: projection=[id, get_field(s,'label'), get_field(s,'value')]
+```
+
+This work also lays the foundation for deeper nested-data optimizations in
+future releases.
+
+**TODO:** Verify whether we want to describe this primarily as a planning /
+projection-pushdown improvement or emphasize the nested Parquet pruning angle.
+
+Related PRs: [#20065], [#20117], [#20239]
+
+### JSON Array File Support
+
+Prior to DataFusion 53, JSON support focused on newline-delimited JSON (NDJSON).
+DataFusion 53 can now read JSON array files such as `[{...}, {...}]` directly,
+including streaming inputs from object stores.
+
+This matters because many real-world systems export JSON in array form rather
+than NDJSON. With this release, those datasets can be queried directly without
+pre-converting them into a different layout.
+
+Example:
+
+```sql
+CREATE EXTERNAL TABLE events
+STORED AS JSON
+OPTIONS ('format.newline_delimited' 'false')
+LOCATION 's3://bucket/events.json';
+```
+
+Conceptually, DataFusion reads the array in a streaming fashion:
+
+```text
+JSON array file/stream
+  -> streaming array-to-NDJSON conversion
+  -> Arrow JSON reader
+  -> RecordBatch
+```
+
+Related PRs: [#19924]
+
+### Better Support for Programmatic and Extension-Oriented Engines
+
+DataFusion 53 continues to improve the APIs used by downstream systems and
+extensions. Notable changes include support for extension planning for
+`TableScan`, new parser support for `JsonAccess`/`:` as a binary operator, and
+additional FFI improvements and fixes on the release branch.
+
+This is important because DataFusion is used as a library and planner inside
+many other systems. Making the planner more extensible and the parser more
+customizable lowers the cost of building specialized SQL dialects and data
+sources on top of DataFusion.
+
+Example use cases include:
+
+```sql
+-- extension-specific table syntax / planning
+SELECT * FROM my_custom_source(...);
+
+-- variant / JSON style operators handled by an extension planner
+SELECT payload:'user_id' FROM events;
+```
+
+**TODO:** Decide how much of this section should stay in the main post versus
+move to a shorter "notable improvements" section.
+
+Related PRs: [#20548], [#20717], [#20722]
+
+## Other Notable Improvements
+
+In addition to the major features above, DataFusion 53 includes several other
+changes that may be worth calling out in the final version of this post:
+
+### More SQL Surface Area
+
+DataFusion 53 added support for set-comparison subqueries, null-aware anti join,
+limited deletion, and a large batch of Spark-compatible functions.
+
+These improvements make DataFusion easier to adopt for engines and workloads
+that expect broader SQL compatibility.
+
+Related PRs: [#19109], [#19635], [#20137], [#19699], [#19702], [#19811], [#19829], [#19845]
+
+### More Visibility Into Parquet Pruning
+
+DataFusion 53 can now report the number of matched Parquet pages in
+`DataSourceExec`, making it easier to understand whether pruning is working as
+expected.
+
+This is useful both for users tuning workloads and for contributors developing
+new pruning optimizations.
+
+Related PRs: [#19977]
+
+### Continued Work on Execution Correctness and Stability
+
+The `branch-53` release process pulled in a substantial number of fixes for
+correctness and stability, including fixes for spilling, hash join edge cases,
+FFI behavior, filter serialization, parquet leaf index mapping, and a regression
+in `count(*)` with right anti / mark joins.
+
+These are not as headline-friendly as new features, but they are highly
+important for real deployments.
+
+Related PRs: [#20726], [#20791], [#20792], [#20882], [#20883], [#20884], [#20890], [#20891], [#20892], [#20895], [#20898], [#20903], [#20918], [#20932]
+
+## Upgrade Notes
+
+DataFusion 53 also includes several breaking changes, including updates to the
+SQL parser, optimizer behavior, and some physical-plan APIs. Please see the
+[upgrade guide] and [changelog] for the full details before upgrading.
+
+[upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading.html
+
+## Thank You
+
+Thank you to everyone in the DataFusion community who contributed code,
+reviews, testing, bug reports, documentation, and release engineering work for
+53.0.0.
+
+[limit pruning blog post]: https://datafusion.apache.org/blog/2026/03/20/limit-pruning/
+
+[#18868]: https://github.com/apache/datafusion/pull/18868
+[#19109]: https://github.com/apache/datafusion/pull/19109
+[#19635]: https://github.com/apache/datafusion/pull/19635
+[#19699]: https://github.com/apache/datafusion/pull/19699
+[#19702]: https://github.com/apache/datafusion/pull/19702
+[#19792]: https://github.com/apache/datafusion/pull/19792
+[#19811]: https://github.com/apache/datafusion/pull/19811
+[#19829]: https://github.com/apache/datafusion/pull/19829
+[#19845]: https://github.com/apache/datafusion/pull/19845
+[#19893]: https://github.com/apache/datafusion/pull/19893
+[#19918]: https://github.com/apache/datafusion/pull/19918
+[#19924]: https://github.com/apache/datafusion/pull/19924
+[#19977]: https://github.com/apache/datafusion/pull/19977
+[#20065]: https://github.com/apache/datafusion/pull/20065
+[#20117]: https://github.com/apache/datafusion/pull/20117
+[#20137]: https://github.com/apache/datafusion/pull/20137
+[#20145]: https://github.com/apache/datafusion/pull/20145
+[#20192]: https://github.com/apache/datafusion/pull/20192
+[#20239]: https://github.com/apache/datafusion/pull/20239
+[#20548]: https://github.com/apache/datafusion/pull/20548
+[#20717]: https://github.com/apache/datafusion/pull/20717
+[#20722]: https://github.com/apache/datafusion/pull/20722
+[#20726]: https://github.com/apache/datafusion/pull/20726
+[#20791]: https://github.com/apache/datafusion/pull/20791
+[#20792]: https://github.com/apache/datafusion/pull/20792
+[#20882]: https://github.com/apache/datafusion/pull/20882
+[#20883]: https://github.com/apache/datafusion/pull/20883
+[#20884]: https://github.com/apache/datafusion/pull/20884
+[#20890]: https://github.com/apache/datafusion/pull/20890
+[#20891]: https://github.com/apache/datafusion/pull/20891
+[#20892]: https://github.com/apache/datafusion/pull/20892
+[#20895]: https://github.com/apache/datafusion/pull/20895
+[#20898]: https://github.com/apache/datafusion/pull/20898
+[#20903]: https://github.com/apache/datafusion/pull/20903
+[#20918]: https://github.com/apache/datafusion/pull/20918
+[#20932]: https://github.com/apache/datafusion/pull/20932

From ea2756d9ed95b30cb02a6d62d008fdd65f52cad2 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Wed, 25 Mar 2026 17:46:11 -0400
Subject: [PATCH 02/22] Updates / cleanup

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 51 +++++++++-----------
 1 file changed, 22 insertions(+), 29 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 66a1f72c..c68692ed 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -45,19 +45,20 @@ stages of execution.
 
 **TODO: Add performance chart and short discussion of the benchmark setup and results.**
 
+**TODO: Also mention we updated the clickbench benchmark to use DataFusion 52 **
+
+
 ### `LIMIT`-Aware Parquet Row Group Pruning
 
 DataFusion 53 includes a new optimization that makes Parquet pruning aware of
-SQL `LIMIT`. If DataFusion can prove that an entire row group matches the
-predicate, and those fully matching row groups contain enough rows to satisfy
-the `LIMIT`, it can skip scanning partially matching row groups entirely.
-
-This is important because it turns `LIMIT` from a result-shaping operator into
-an I/O optimization. For selective queries on Parquet, this can avoid decoding
-entire row groups that would previously have been scanned only to discover that
-their rows were not needed.
+SQL `LIMIT`, which is described in full in [limit pruning blog post]. If
+DataFusion can prove that an entire row group matches the predicate, and those
+fully matching row groups contain enough rows to satisfy the `LIMIT`, partially
+matching row groups are skipped entirely. or selective queries on Parquet, this
+optimization can avoid fetching entire row groups that would previously have
+been scanned only to discover that their rows were not needed.
 
-For example, consider:
+For example, consider this example from the [limit pruning blog post]:
 
 ```sql
 SELECT *
@@ -66,27 +67,17 @@ WHERE species LIKE 'Alpine%' AND score >= 50
 LIMIT 3;
 ```
 
-Conceptually, DataFusion 53 can now do:
-
-```text
-Before:
-  scan partially matching row groups
-  scan fully matching row groups
-  stop after 3 matching rows
-
-After:
-  identify fully matching row groups from statistics
-  if they already contain >= 3 rows, scan only those row groups
-```
+Conceptually, prior to DataFusion 53, the engine would scan both partially and
+fully matching row groups and stop after 3 matching rows. In DataFusion 53, if
+the engine can identify fully matching row groups from statistics that already
+contain >= 3 rows, it scans only those row groups
 
-For a deeper explanation, see the dedicated [limit pruning blog post].
 Related PRs: [#18868]
 
-### Broader Filter Pushdown Across Joins and Unions
+### Improved Filter Pushdown
 
-DataFusion 53 also pushes filters through more plan shapes. In particular,
-predicates can now be inferred and pushed through more join types, and filters
-can be pushed through `UnionExec` without being redundantly re-applied.
+DataFusion 53 pushes filters through more join types, and through `UnionExec`,
+as well as has expanded support for pushing down [dynamic filters].
 
 This matters because the cheapest rows are the rows you never read. Better
 pushdown means fewer rows flow into joins, repartitions, and later operators,
@@ -293,9 +284,10 @@ SQL parser, optimizer behavior, and some physical-plan APIs. Please see the
 
 ## Thank You
 
-Thank you to everyone in the DataFusion community who contributed code,
-reviews, testing, bug reports, documentation, and release engineering work for
-53.0.0.
+Thank you to everyone in the DataFusion community who contributed code, reviews,
+testing, bug reports, documentation, and release engineering work for 53.0.0.
+This release contains direct contributions from 114 different people, and we are
+grateful for the time and effort that everyone put in to make it happen.
 
 [limit pruning blog post]: https://datafusion.apache.org/blog/2026/03/20/limit-pruning/
 
@@ -335,3 +327,4 @@ reviews, testing, bug reports, documentation, and release engineering work for
 [#20903]: https://github.com/apache/datafusion/pull/20903
 [#20918]: https://github.com/apache/datafusion/pull/20918
 [#20932]: https://github.com/apache/datafusion/pull/20932
+[dynamic filters]: https://datafusion.apache.org/blog/2025/09/10/dynamic-filters

From d9d80f9d0bd30106a485f62028408e818a410fb5 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Wed, 25 Mar 2026 17:55:24 -0400
Subject: [PATCH 03/22] more shouts

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 152 +++++++++++++++++++
 1 file changed, 152 insertions(+)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index c68692ed..f6a75389 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -135,6 +135,30 @@ After:
 
 Related PRs: [#19792], [#19893]
 
+### Faster Built-in Function Implementations
+
+DataFusion 53 also includes a large amount of low-level work to make built-in
+scalar and array functions faster. This release contains optimizations for
+string processing, array operations, and other frequently used expressions that
+show up in real analytical workloads.
+
+This work is important because these functions sit on hot execution paths.
+Small improvements to heavily used functions compound across large datasets and
+complex queries.
+
+We especially want to call out [@neilconway] for a substantial series of
+performance improvements in this release, including [strpos], [replace],
+[concat], [translate], [array_has], [lpad]/[rpad], [trim], [initcap],
+[to_array_of_size], [array_has_any], [array_position], [array_agg], and
+[array_to_string].
+
+Many other contributors also improved function performance in 53.0.0, including
+[@theirix] on [left] and [right], [@lyne7-sc] on [array_remove],
+[array_repeat_func], [array_union] / [array_intersect], and [array_distinct],
+[@kumarUjjawal] on [ascii], [chr], [regexp_like], [atan2], [round], [signum],
+and many other scalar fast paths, [@pepijnve] on [case_when], [@zhangxffff] on
+[in_list], and [@UBarney] on [hash_table_lookup].
+
 ## Major Features ✨
 
 ### Nested Field Pushdown for Cheap Expressions
@@ -251,6 +275,23 @@ that expect broader SQL compatibility.
 
 Related PRs: [#19109], [#19635], [#20137], [#19699], [#19702], [#19811], [#19829], [#19845]
 
+### More Spark Compatibility Functions
+
+DataFusion 53 added another substantial batch of Spark-compatible built-in
+functions and behaviors, continuing the work to make DataFusion easier to use
+for Spark-oriented workloads and downstream engines.
+
+This release includes new or improved support for functions such as
+ [collect_list] / [collect_set], [array_repeat_spark], [slice], [substring],
+ [date_diff], [add_months], [base64] / [unbase64], [from_utc_timestamp] /
+ [to_utc_timestamp], [unix_date] and [unix_timestamp], [negative],
+ [string_to_map], [json_tuple], [arrays_zip], [bitmap_bit_position],
+ [bitmap_bucket_number], [bin], [abs_ansi], and [array_contains].
+
+Special thanks to the contributors who drove this work, especially [@cht42],
+[@CuteChuanChuan], [@SubhamSinghal], [@kazantsev-maksim], [@unknowntpo],
+[@aryan-212], [@hsiang-c], and [@davidlghellin].
+
 ### More Visibility Into Parquet Pruning
 
 DataFusion 53 can now report the number of matched Parquet pages in
@@ -290,27 +331,138 @@ This release contains direct contributions from 114 different people, and we are
 grateful for the time and effort that everyone put in to make it happen.
 
 [limit pruning blog post]: https://datafusion.apache.org/blog/2026/03/20/limit-pruning/
+[@neilconway]: https://github.com/neilconway
+[@cht42]: https://github.com/cht42
+[@CuteChuanChuan]: https://github.com/CuteChuanChuan
+[@SubhamSinghal]: https://github.com/SubhamSinghal
+[@kazantsev-maksim]: https://github.com/kazantsev-maksim
+[@unknowntpo]: https://github.com/unknowntpo
+[@aryan-212]: https://github.com/aryan-212
+[@hsiang-c]: https://github.com/hsiang-c
+[@davidlghellin]: https://github.com/davidlghellin
+[@theirix]: https://github.com/theirix
+[@lyne7-sc]: https://github.com/lyne7-sc
+[@kumarUjjawal]: https://github.com/kumarUjjawal
+[@pepijnve]: https://github.com/pepijnve
+[@zhangxffff]: https://github.com/zhangxffff
+[@UBarney]: https://github.com/UBarney
+
+[abs_ansi]: https://github.com/apache/datafusion/pull/18828
+[add_months]: https://github.com/apache/datafusion/pull/19711
+[array_agg]: https://github.com/apache/datafusion/pull/20504
+[array_contains]: https://github.com/apache/datafusion/pull/20685
+[array_distinct]: https://github.com/apache/datafusion/pull/20364
+[array_has]: https://github.com/apache/datafusion/pull/20374
+[array_has_any]: https://github.com/apache/datafusion/pull/20385
+[array_intersect]: https://github.com/apache/datafusion/pull/20243
+[array_position]: https://github.com/apache/datafusion/pull/20532
+[array_remove]: https://github.com/apache/datafusion/pull/19996
+[array_repeat_func]: https://github.com/apache/datafusion/pull/20049
+[array_repeat_spark]: https://github.com/apache/datafusion/pull/19702
+[array_to_string]: https://github.com/apache/datafusion/pull/20553
+[array_union]: https://github.com/apache/datafusion/pull/20243
+[arrays_zip]: https://github.com/apache/datafusion/pull/20440
+[ascii]: https://github.com/apache/datafusion/pull/19951
+[atan2]: https://github.com/apache/datafusion/pull/20336
+[base64]: https://github.com/apache/datafusion/pull/19968
+[bin]: https://github.com/apache/datafusion/pull/20479
+[bitmap_bit_position]: https://github.com/apache/datafusion/pull/20275
+[bitmap_bucket_number]: https://github.com/apache/datafusion/pull/20288
+[case_when]: https://github.com/apache/datafusion/pull/20097
+[chr]: https://github.com/apache/datafusion/pull/20073
+[collect_list]: https://github.com/apache/datafusion/pull/19699
+[collect_set]: https://github.com/apache/datafusion/pull/19699
+[concat]: https://github.com/apache/datafusion/pull/20317
+[date_diff]: https://github.com/apache/datafusion/pull/19845
+[from_utc_timestamp]: https://github.com/apache/datafusion/pull/19880
+[hash_table_lookup]: https://github.com/apache/datafusion/pull/19602
+[in_list]: https://github.com/apache/datafusion/pull/20528
+[initcap]: https://github.com/apache/datafusion/pull/20352
+[json_tuple]: https://github.com/apache/datafusion/pull/20412
+[left]: https://github.com/apache/datafusion/pull/19980
+[lpad]: https://github.com/apache/datafusion/pull/20278
+[negative]: https://github.com/apache/datafusion/pull/20006
+[regexp_like]: https://github.com/apache/datafusion/pull/20354
+[replace]: https://github.com/apache/datafusion/pull/20344
+[right]: https://github.com/apache/datafusion/pull/20069
+[round]: https://github.com/apache/datafusion/pull/19831
+[rpad]: https://github.com/apache/datafusion/pull/20278
+[signum]: https://github.com/apache/datafusion/pull/19871
+[slice]: https://github.com/apache/datafusion/pull/19811
+[strpos]: https://github.com/apache/datafusion/pull/20295
+[string_to_map]: https://github.com/apache/datafusion/pull/20120
+[substring]: https://github.com/apache/datafusion/pull/19805
+[to_array_of_size]: https://github.com/apache/datafusion/pull/20459
+[to_utc_timestamp]: https://github.com/apache/datafusion/pull/19880
+[translate]: https://github.com/apache/datafusion/pull/20305
+[trim]: https://github.com/apache/datafusion/pull/20328
+[unbase64]: https://github.com/apache/datafusion/pull/19968
+[unix_date]: https://github.com/apache/datafusion/pull/19892
+[unix_timestamp]: https://github.com/apache/datafusion/pull/19892
 
 [#18868]: https://github.com/apache/datafusion/pull/18868
+[#18828]: https://github.com/apache/datafusion/pull/18828
 [#19109]: https://github.com/apache/datafusion/pull/19109
+[#19592]: https://github.com/apache/datafusion/pull/19592
 [#19635]: https://github.com/apache/datafusion/pull/19635
 [#19699]: https://github.com/apache/datafusion/pull/19699
 [#19702]: https://github.com/apache/datafusion/pull/19702
+[#19711]: https://github.com/apache/datafusion/pull/19711
 [#19792]: https://github.com/apache/datafusion/pull/19792
+[#19805]: https://github.com/apache/datafusion/pull/19805
 [#19811]: https://github.com/apache/datafusion/pull/19811
 [#19829]: https://github.com/apache/datafusion/pull/19829
 [#19845]: https://github.com/apache/datafusion/pull/19845
+[#19865]: https://github.com/apache/datafusion/pull/19865
+[#19880]: https://github.com/apache/datafusion/pull/19880
+[#19892]: https://github.com/apache/datafusion/pull/19892
 [#19893]: https://github.com/apache/datafusion/pull/19893
 [#19918]: https://github.com/apache/datafusion/pull/19918
 [#19924]: https://github.com/apache/datafusion/pull/19924
+[#19951]: https://github.com/apache/datafusion/pull/19951
+[#19996]: https://github.com/apache/datafusion/pull/19996
+[#19968]: https://github.com/apache/datafusion/pull/19968
+[#19984]: https://github.com/apache/datafusion/pull/19984
 [#19977]: https://github.com/apache/datafusion/pull/19977
+[#20049]: https://github.com/apache/datafusion/pull/20049
 [#20065]: https://github.com/apache/datafusion/pull/20065
+[#20006]: https://github.com/apache/datafusion/pull/20006
+[#20073]: https://github.com/apache/datafusion/pull/20073
+[#20097]: https://github.com/apache/datafusion/pull/20097
 [#20117]: https://github.com/apache/datafusion/pull/20117
 [#20137]: https://github.com/apache/datafusion/pull/20137
 [#20145]: https://github.com/apache/datafusion/pull/20145
 [#20192]: https://github.com/apache/datafusion/pull/20192
+[#20120]: https://github.com/apache/datafusion/pull/20120
+[#20243]: https://github.com/apache/datafusion/pull/20243
 [#20239]: https://github.com/apache/datafusion/pull/20239
+[#20275]: https://github.com/apache/datafusion/pull/20275
+[#20288]: https://github.com/apache/datafusion/pull/20288
+[#20278]: https://github.com/apache/datafusion/pull/20278
+[#20295]: https://github.com/apache/datafusion/pull/20295
+[#20305]: https://github.com/apache/datafusion/pull/20305
+[#20317]: https://github.com/apache/datafusion/pull/20317
+[#20323]: https://github.com/apache/datafusion/pull/20323
+[#20328]: https://github.com/apache/datafusion/pull/20328
+[#20336]: https://github.com/apache/datafusion/pull/20336
+[#20344]: https://github.com/apache/datafusion/pull/20344
+[#20354]: https://github.com/apache/datafusion/pull/20354
+[#20352]: https://github.com/apache/datafusion/pull/20352
+[#20374]: https://github.com/apache/datafusion/pull/20374
+[#20385]: https://github.com/apache/datafusion/pull/20385
+[#20364]: https://github.com/apache/datafusion/pull/20364
+[#20412]: https://github.com/apache/datafusion/pull/20412
+[#20440]: https://github.com/apache/datafusion/pull/20440
+[#20461]: https://github.com/apache/datafusion/pull/20461
+[#20479]: https://github.com/apache/datafusion/pull/20479
+[#20459]: https://github.com/apache/datafusion/pull/20459
 [#20548]: https://github.com/apache/datafusion/pull/20548
+[#20504]: https://github.com/apache/datafusion/pull/20504
+[#20532]: https://github.com/apache/datafusion/pull/20532
+[#20538]: https://github.com/apache/datafusion/pull/20538
+[#20553]: https://github.com/apache/datafusion/pull/20553
+[#20528]: https://github.com/apache/datafusion/pull/20528
+[#20685]: https://github.com/apache/datafusion/pull/20685
 [#20717]: https://github.com/apache/datafusion/pull/20717
 [#20722]: https://github.com/apache/datafusion/pull/20722
 [#20726]: https://github.com/apache/datafusion/pull/20726

From c4691f3eaf424d1af9e32beba4b7005c864500fc Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Wed, 25 Mar 2026 17:58:01 -0400
Subject: [PATCH 04/22] clenaup

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index f6a75389..02cd65cf 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -281,12 +281,9 @@ DataFusion 53 added another substantial batch of Spark-compatible built-in
 functions and behaviors, continuing the work to make DataFusion easier to use
 for Spark-oriented workloads and downstream engines.
 
-This release includes new or improved support for functions such as
- [collect_list] / [collect_set], [array_repeat_spark], [slice], [substring],
- [date_diff], [add_months], [base64] / [unbase64], [from_utc_timestamp] /
- [to_utc_timestamp], [unix_date] and [unix_timestamp], [negative],
- [string_to_map], [json_tuple], [arrays_zip], [bitmap_bit_position],
- [bitmap_bucket_number], [bin], [abs_ansi], and [array_contains].
+This release includes around 20 new or improved Spark-compatible functions and
+behaviors, such as [collect_list], [date_diff], [from_utc_timestamp],
+[json_tuple], [arrays_zip], [bin], and [array_contains].
 
 Special thanks to the contributors who drove this work, especially [@cht42],
 [@CuteChuanChuan], [@SubhamSinghal], [@kazantsev-maksim], [@unknowntpo],

From 52725fad99c577d0d8a4361e2dd12a8bf0ca556d Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 08:35:11 -0400
Subject: [PATCH 05/22] Edits

---
 content/blog/2026-03-25-datafusion-53.0.0.md  | 107 ++++++++++--------
 .../datafusion-53/join-filter-pushdown.svg    |  68 +++++++++++
 2 files changed, 130 insertions(+), 45 deletions(-)
 create mode 100644 content/images/datafusion-53/join-filter-pushdown.svg

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 02cd65cf..ecb231c1 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -39,9 +39,9 @@ making this release possible.
 
 ## Performance Improvements 🚀
 
-DataFusion 53 continues the project-wide focus on reducing planning overhead,
-skipping more unnecessary I/O, and pushing more work into earlier and cheaper
-stages of execution.
+DataFusion 53 continues the project-wide focus on performance in general and
+this release features reducing planning overhead, skipping more unnecessary I/O,
+and pushing more work into earlier and cheaper stages of execution.
 
 **TODO: Add performance chart and short discussion of the benchmark setup and results.**
 
@@ -54,32 +54,19 @@ DataFusion 53 includes a new optimization that makes Parquet pruning aware of
 SQL `LIMIT`, which is described in full in [limit pruning blog post]. If
 DataFusion can prove that an entire row group matches the predicate, and those
 fully matching row groups contain enough rows to satisfy the `LIMIT`, partially
-matching row groups are skipped entirely. or selective queries on Parquet, this
-optimization can avoid fetching entire row groups that would previously have
-been scanned only to discover that their rows were not needed.
+matching row groups are skipped entirely. 
 
-For example, consider this example from the [limit pruning blog post]:
+<figure>
+<img src="/blog/images/limit-pruning/pruning-pipeline.svg" width="80%" alt="Pruning pipeline with limit pruning highlighted"/>
+<figcaption><b>Figure 1</b>: New limit pruning inserted between row group and page index pruning. </figcaption>
+</figure>
 
-```sql
-SELECT *
-FROM tracking_data
-WHERE species LIKE 'Alpine%' AND score >= 50
-LIMIT 3;
-```
-
-Conceptually, prior to DataFusion 53, the engine would scan both partially and
-fully matching row groups and stop after 3 matching rows. In DataFusion 53, if
-the engine can identify fully matching row groups from statistics that already
-contain >= 3 rows, it scans only those row groups
-
-Related PRs: [#18868]
+Special thanks to [@xudong963] for implementing this feature. Related PRs: [#18868]
 
 ### Improved Filter Pushdown
 
-DataFusion 53 pushes filters through more join types, and through `UnionExec`,
-as well as has expanded support for pushing down [dynamic filters].
-
-This matters because the cheapest rows are the rows you never read. Better
+DataFusion 53 pushes filters down through more join types, and through `UnionExec`,
+as well as has expanded support for pushing down [dynamic filters]. More
 pushdown means fewer rows flow into joins, repartitions, and later operators,
 which reduces CPU, memory, and I/O.
 
@@ -108,36 +95,30 @@ After:
     FilterExec(k = 1) -> DataSourceExec(t2)
 ```
 
-Related PRs: [#19918], [#20145], [#20192]
+<figure>
+<img src="/blog/images/datafusion-53/join-filter-pushdown.svg" width="80%" alt="Before and after diagram of filter pushdown around a hash join"/>
+<figcaption><b>Figure 2</b>: DataFusion 53 infers additional filters are from join conditions and pushes them down in the plan</figcaption>
+</figure>
 
-### Faster Query Planning
-
-DataFusion 53 reduces planning overhead by caching `PlanProperties`,
-reusing them when plan children are unchanged, and making immutable pieces of
-execution plans cheaper to clone.
+Special thanks to [@nuno-faria], [@haohuaijin], and [@jackkleeman] for
+driving this work.  Related PRs: [#19918], [#20145], [#20192]
 
-This work is important for applications that generate many queries, repeatedly
-optimize similar plans, or build plans programmatically. In those environments,
-planning cost can become user-visible even when execution is fast.
+### Faster Query Planning
 
-Conceptually:
+DataFusion 53 improves query planning performance,  making immutable pieces of
+execution plans cheaper to clone. This helps applications that need extreme low
+latency, plan many or complex queries, and those using prepared statements or
+parameterized queries.
 
-```text
-Before:
-  rebuild plan node
-  recompute PlanProperties
-  clone large immutable vectors
-
-After:
-  reuse cached PlanProperties when possible
-  share immutable plan state with Arc
-```
+Special thanks to [@askalt] for leading this work.  planning overhead by caching `PlanProperties`,
 
 Related PRs: [#19792], [#19893]
 
+** TODO get sql_planner benchmarks for 52.4.0 and 53.0.0 **
+
 ### Faster Built-in Function Implementations
 
-DataFusion 53 also includes a large amount of low-level work to make built-in
+DataFusion 53 also includes a significant number of low-level optimizations to built-in
 scalar and array functions faster. This release contains optimizations for
 string processing, array operations, and other frequently used expressions that
 show up in real analytical workloads.
@@ -201,6 +182,9 @@ future releases.
 **TODO:** Verify whether we want to describe this primarily as a planning /
 projection-pushdown improvement or emphasize the nested Parquet pruning angle.
 
+Special thanks to [@adriangb] for designing and implementing this optimizer
+work.
+
 Related PRs: [#20065], [#20117], [#20239]
 
 ### JSON Array File Support
@@ -231,6 +215,8 @@ JSON array file/stream
   -> RecordBatch
 ```
 
+Special thanks to [@zhuqi-lucas] for implementing this feature.
+
 Related PRs: [#19924]
 
 ### Better Support for Programmatic and Extension-Oriented Engines
@@ -258,6 +244,9 @@ SELECT payload:'user_id' FROM events;
 **TODO:** Decide how much of this section should stay in the main post versus
 move to a shorter "notable improvements" section.
 
+Special thanks to [@linhr], [@Samyak2], and [@timsaucer] for these
+extension-oriented improvements.
+
 Related PRs: [#20548], [#20717], [#20722]
 
 ## Other Notable Improvements
@@ -273,6 +262,9 @@ limited deletion, and a large batch of Spark-compatible functions.
 These improvements make DataFusion easier to adopt for engines and workloads
 that expect broader SQL compatibility.
 
+Special thanks to [@waynexia], [@viirya], [@askalt], and [@cht42] for key
+contributions in this area.
+
 Related PRs: [#19109], [#19635], [#20137], [#19699], [#19702], [#19811], [#19829], [#19845]
 
 ### More Spark Compatibility Functions
@@ -298,6 +290,8 @@ expected.
 This is useful both for users tuning workloads and for contributors developing
 new pruning optimizations.
 
+Special thanks to [@nuno-faria] for this visibility improvement.
+
 Related PRs: [#19977]
 
 ### Continued Work on Execution Correctness and Stability
@@ -310,6 +304,10 @@ in `count(*)` with right anti / mark joins.
 These are not as headline-friendly as new features, but they are highly
 important for real deployments.
 
+Thanks to [@jonathanc-n], [@alamb], [@xanderbailey], [@haohuaijin],
+[@friendlymatthew], [@fwojciec], [@Kontinuation], [@nathanb9], and many others
+who helped stabilize the release branch.
+
 Related PRs: [#20726], [#20791], [#20792], [#20882], [#20883], [#20884], [#20890], [#20891], [#20892], [#20895], [#20898], [#20903], [#20918], [#20932]
 
 ## Upgrade Notes
@@ -329,6 +327,25 @@ grateful for the time and effort that everyone put in to make it happen.
 
 [limit pruning blog post]: https://datafusion.apache.org/blog/2026/03/20/limit-pruning/
 [@neilconway]: https://github.com/neilconway
+[@xudong963]: https://github.com/xudong963
+[@nuno-faria]: https://github.com/nuno-faria
+[@haohuaijin]: https://github.com/haohuaijin
+[@jackkleeman]: https://github.com/jackkleeman
+[@askalt]: https://github.com/askalt
+[@adriangb]: https://github.com/adriangb
+[@zhuqi-lucas]: https://github.com/zhuqi-lucas
+[@linhr]: https://github.com/linhr
+[@Samyak2]: https://github.com/Samyak2
+[@timsaucer]: https://github.com/timsaucer
+[@waynexia]: https://github.com/waynexia
+[@viirya]: https://github.com/viirya
+[@jonathanc-n]: https://github.com/jonathanc-n
+[@alamb]: https://github.com/alamb
+[@xanderbailey]: https://github.com/xanderbailey
+[@friendlymatthew]: https://github.com/friendlymatthew
+[@fwojciec]: https://github.com/fwojciec
+[@Kontinuation]: https://github.com/Kontinuation
+[@nathanb9]: https://github.com/nathanb9
 [@cht42]: https://github.com/cht42
 [@CuteChuanChuan]: https://github.com/CuteChuanChuan
 [@SubhamSinghal]: https://github.com/SubhamSinghal
diff --git a/content/images/datafusion-53/join-filter-pushdown.svg b/content/images/datafusion-53/join-filter-pushdown.svg
new file mode 100644
index 00000000..84a51472
--- /dev/null
+++ b/content/images/datafusion-53/join-filter-pushdown.svg
@@ -0,0 +1,68 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 900 320">
+  <defs>
+    <marker id="arrow" markerWidth="10" markerHeight="7" refX="10" refY="3.5" orient="auto">
+      <polygon points="0 0, 10 3.5, 0 7" fill="#555"/>
+    </marker>
+  </defs>
+  <style>
+    text { font-family: 'Segoe UI', Arial, sans-serif; }
+    .title { font-size: 15px; font-weight: 700; fill: #222; }
+    .stage { fill: #f0f4ff; stroke: #4a7bcc; stroke-width: 1.5; rx: 8; }
+    .highlight { fill: #fff3e0; stroke: #e67e22; stroke-width: 2; rx: 8; }
+    .label { font-size: 12px; fill: #333; text-anchor: middle; font-weight: 600; }
+    .sublabel { font-size: 10px; fill: #777; text-anchor: middle; }
+  </style>
+
+  <!-- Before -->
+  <text class="title" x="225" y="28" text-anchor="middle">Before</text>
+  <line x1="40" y1="38" x2="410" y2="38" stroke="#ddd" stroke-width="1"/>
+
+  <rect class="highlight" x="150" y="55" width="150" height="42"/>
+  <text class="label" x="225" y="80">FilterExec</text>
+  <text class="sublabel" x="225" y="112">k = 1</text>
+
+  <line x1="225" y1="97" x2="225" y2="128" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="stage" x="150" y="132" width="150" height="42"/>
+  <text class="label" x="225" y="157">HashJoinExec</text>
+
+  <line x1="188" y1="174" x2="125" y2="220" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="262" y1="174" x2="325" y2="220" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="stage" x="55" y="224" width="140" height="42"/>
+  <text class="label" x="125" y="249">DataSourceExec</text>
+  <text class="sublabel" x="125" y="281">t1</text>
+
+  <rect class="stage" x="255" y="224" width="140" height="42"/>
+  <text class="label" x="325" y="249">DataSourceExec</text>
+  <text class="sublabel" x="325" y="281">t2</text>
+
+  <!-- After -->
+  <text class="title" x="675" y="28" text-anchor="middle">After</text>
+  <line x1="490" y1="38" x2="860" y2="38" stroke="#ddd" stroke-width="1"/>
+
+  <rect class="stage" x="600" y="55" width="150" height="42"/>
+  <text class="label" x="675" y="80">HashJoinExec</text>
+
+  <line x1="638" y1="97" x2="585" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="712" y1="97" x2="765" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="highlight" x="510" y="140" width="150" height="42"/>
+  <text class="label" x="585" y="165">FilterExec</text>
+  <text class="sublabel" x="585" y="197">k = 1</text>
+
+  <rect class="highlight" x="690" y="140" width="150" height="42"/>
+  <text class="label" x="765" y="165">FilterExec</text>
+  <text class="sublabel" x="765" y="197">k = 1</text>
+
+  <line x1="585" y1="182" x2="585" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="765" y1="182" x2="765" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="stage" x="510" y="221" width="150" height="42"/>
+  <text class="label" x="585" y="246">DataSourceExec</text>
+  <text class="sublabel" x="585" y="278">t1</text>
+
+  <rect class="stage" x="690" y="221" width="150" height="42"/>
+  <text class="label" x="765" y="246">DataSourceExec</text>
+  <text class="sublabel" x="765" y="278">t2</text>
+</svg>

From c03f0e0a1d22aad4e643d9410c80ea6468d9deb4 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 08:45:13 -0400
Subject: [PATCH 06/22] update

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 32 +++++++++-----------
 1 file changed, 14 insertions(+), 18 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index ecb231c1..55bfdb86 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -118,23 +118,18 @@ Related PRs: [#19792], [#19893]
 
 ### Faster Built-in Function Implementations
 
-DataFusion 53 also includes a significant number of low-level optimizations to built-in
-scalar and array functions faster. This release contains optimizations for
-string processing, array operations, and other frequently used expressions that
-show up in real analytical workloads.
-
-This work is important because these functions sit on hot execution paths.
-Small improvements to heavily used functions compound across large datasets and
-complex queries.
-
-We especially want to call out [@neilconway] for a substantial series of
-performance improvements in this release, including [strpos], [replace],
-[concat], [translate], [array_has], [lpad]/[rpad], [trim], [initcap],
-[to_array_of_size], [array_has_any], [array_position], [array_agg], and
-[array_to_string].
-
-Many other contributors also improved function performance in 53.0.0, including
-[@theirix] on [left] and [right], [@lyne7-sc] on [array_remove],
+DataFusion 53 includes performance improvements for many built-in functions.
+DataFusion, like all real-world SQL engines, includes [235 built-in
+functions], and improving the performance of these functions benefits a wide
+range of workloads.
+
+
+Special thanks to [@neilconway] for a substantial series of performance
+improvements in this release, including [strpos], [replace], [concat],
+[translate], [array_has], [lpad]/[rpad], [trim], [initcap], [to_array_of_size],
+[array_has_any], [array_position], [array_agg], and [array_to_string].  In
+addition,  other contributors also improved function performance in 53.0.0,
+including [@theirix] on [left] and [right], [@lyne7-sc] on [array_remove],
 [array_repeat_func], [array_union] / [array_intersect], and [array_distinct],
 [@kumarUjjawal] on [ascii], [chr], [regexp_like], [atan2], [round], [signum],
 and many other scalar fast paths, [@pepijnve] on [case_when], [@zhangxffff] on
@@ -273,7 +268,7 @@ DataFusion 53 added another substantial batch of Spark-compatible built-in
 functions and behaviors, continuing the work to make DataFusion easier to use
 for Spark-oriented workloads and downstream engines.
 
-This release includes around 20 new or improved Spark-compatible functions and
+This release includes almost 20 new or improved Spark-compatible functions and
 behaviors, such as [collect_list], [date_diff], [from_utc_timestamp],
 [json_tuple], [arrays_zip], [bin], and [array_contains].
 
@@ -494,3 +489,4 @@ grateful for the time and effort that everyone put in to make it happen.
 [#20918]: https://github.com/apache/datafusion/pull/20918
 [#20932]: https://github.com/apache/datafusion/pull/20932
 [dynamic filters]: https://datafusion.apache.org/blog/2025/09/10/dynamic-filters
+[235 built-in functions]: https://datafusion.apache.org/user-guide/sql/scalar_functions.html
\ No newline at end of file

From 2f5a6f24f81b68437b516b10fdf7a62aeedb9ada Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 08:58:31 -0400
Subject: [PATCH 07/22] update

---
 content/blog/2026-03-25-datafusion-53.0.0.md  | 91 ++++++-------------
 .../datafusion-53/field-access-pushdown.svg   | 58 ++++++++++++
 2 files changed, 84 insertions(+), 65 deletions(-)
 create mode 100644 content/images/datafusion-53/field-access-pushdown.svg

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 55bfdb86..02d8a4ea 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -123,33 +123,25 @@ DataFusion, like all real-world SQL engines, includes [235 built-in
 functions], and improving the performance of these functions benefits a wide
 range of workloads.
 
+This release includes performance improvements for 42 distinct built-in
+functions, such as [strpos], [replace], [concat], [translate], [array_has],
+[array_agg], [left], [right], and [case_when].
 
-Special thanks to [@neilconway] for a substantial series of performance
-improvements in this release, including [strpos], [replace], [concat],
-[translate], [array_has], [lpad]/[rpad], [trim], [initcap], [to_array_of_size],
-[array_has_any], [array_position], [array_agg], and [array_to_string].  In
-addition,  other contributors also improved function performance in 53.0.0,
-including [@theirix] on [left] and [right], [@lyne7-sc] on [array_remove],
-[array_repeat_func], [array_union] / [array_intersect], and [array_distinct],
-[@kumarUjjawal] on [ascii], [chr], [regexp_like], [atan2], [round], [signum],
-and many other scalar fast paths, [@pepijnve] on [case_when], [@zhangxffff] on
-[in_list], and [@UBarney] on [hash_table_lookup].
+Special thanks to the contributors who drove this work, especially
+[@neilconway], [@theirix], [@lyne7-sc], [@kumarUjjawal], [@pepijnve],
+[@zhangxffff], and [@UBarney].
 
-## Major Features ✨
 
-### Nested Field Pushdown for Cheap Expressions
+### Nested Field Pushdown
 
-DataFusion 53 adds a new optimizer pipeline for pushing cheap
-"move-towards-leaf" expressions such as `get_field` closer to the data source.
-This allows nested field access to be extracted earlier in the plan rather than
-materializing wider values and trimming them later.
+DataFusion 53 can now push cheap expressions such as `get_field` down the plan
+and into data sources.  This is especially important for nested data such as
+structs in Parquet files. Instead of reading an entire struct column and then
+extracting the field of interest, DataFusion 53 pushes the field extraction into
+the scan..
 
-This is especially important for nested data such as structs. Instead of reading
-an entire struct column and then extracting one field near the top of the plan,
-DataFusion can push the field extraction down so only the needed data is carried
-through the pipeline.
-
-For example:
+For example the following query reads a struct column `s` and extracts the
+`label` field, for rows where the `value` field is greater than 150:
 
 ```sql
 SELECT id, s['label']
@@ -171,53 +163,22 @@ After:
       DataSourceExec: projection=[id, get_field(s,'label'), get_field(s,'value')]
 ```
 
-This work also lays the foundation for deeper nested-data optimizations in
-future releases.
-
-**TODO:** Verify whether we want to describe this primarily as a planning /
-projection-pushdown improvement or emphasize the nested Parquet pruning angle.
+<figure>
+<img src="/blog/images/datafusion-53/field-access-pushdown.svg" width="80%" alt="Before and after diagram of field access pushdown into a data source"/>
+<figcaption><b>Figure 3</b>: DataFusion 53 pushes field-access expressions closer to the scan.</figcaption>
+</figure>
 
 Special thanks to [@adriangb] for designing and implementing this optimizer
-work.
-
-Related PRs: [#20065], [#20117], [#20239]
-
-### JSON Array File Support
-
-Prior to DataFusion 53, JSON support focused on newline-delimited JSON (NDJSON).
-DataFusion 53 can now read JSON array files such as `[{...}, {...}]` directly,
-including streaming inputs from object stores.
-
-This matters because many real-world systems export JSON in array form rather
-than NDJSON. With this release, those datasets can be queried directly without
-pre-converting them into a different layout.
-
-Example:
-
-```sql
-CREATE EXTERNAL TABLE events
-STORED AS JSON
-OPTIONS ('format.newline_delimited' 'false')
-LOCATION 's3://bucket/events.json';
-```
-
-Conceptually, DataFusion reads the array in a streaming fashion:
-
-```text
-JSON array file/stream
-  -> streaming array-to-NDJSON conversion
-  -> Arrow JSON reader
-  -> RecordBatch
-```
-
-Special thanks to [@zhuqi-lucas] for implementing this feature.
+work. Related PRs: [#20065], [#20117], [#20239]
 
-Related PRs: [#19924]
+## New Features ✨
 
-### Better Support for Programmatic and Extension-Oriented Engines
+*  **JSON Array File Support**:  DataFusion 53 can now read JSON arrays such as
+   `[{...}, {...}]` directly as multiple rows, including streaming inputs from
+   object stores.  Special thanks to [@zhuqi-lucas] for implementing this
+   feature. Related PRs: [#19924]
 
-DataFusion 53 continues to improve the APIs used by downstream systems and
-extensions. Notable changes include support for extension planning for
+* **Variant extension planning for
 `TableScan`, new parser support for `JsonAccess`/`:` as a binary operator, and
 additional FFI improvements and fixes on the release branch.
 
@@ -489,4 +450,4 @@ grateful for the time and effort that everyone put in to make it happen.
 [#20918]: https://github.com/apache/datafusion/pull/20918
 [#20932]: https://github.com/apache/datafusion/pull/20932
 [dynamic filters]: https://datafusion.apache.org/blog/2025/09/10/dynamic-filters
-[235 built-in functions]: https://datafusion.apache.org/user-guide/sql/scalar_functions.html
\ No newline at end of file
+[235 built-in functions]: https://datafusion.apache.org/user-guide/sql/scalar_functions.html
diff --git a/content/images/datafusion-53/field-access-pushdown.svg b/content/images/datafusion-53/field-access-pushdown.svg
new file mode 100644
index 00000000..e0c7b90f
--- /dev/null
+++ b/content/images/datafusion-53/field-access-pushdown.svg
@@ -0,0 +1,58 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 980 340">
+  <defs>
+    <marker id="arrow" markerWidth="10" markerHeight="7" refX="10" refY="3.5" orient="auto">
+      <polygon points="0 0, 10 3.5, 0 7" fill="#555"/>
+    </marker>
+  </defs>
+  <style>
+    text { font-family: 'Segoe UI', Arial, sans-serif; }
+    .title { font-size: 18px; font-weight: 700; fill: #222; }
+    .stage { fill: #f0f4ff; stroke: #4a7bcc; stroke-width: 1.5; rx: 8; }
+    .highlight { fill: #fff3e0; stroke: #e67e22; stroke-width: 2; rx: 8; }
+    .label { font-size: 15px; fill: #333; text-anchor: middle; font-weight: 600; }
+    .sublabel { font-size: 15px; fill: #555; text-anchor: middle; }
+    .tiny { font-size: 12px; fill: #777; text-anchor: middle; }
+  </style>
+
+  <!-- Before -->
+  <text class="title" x="245" y="28" text-anchor="middle">Before</text>
+  <line x1="40" y1="38" x2="450" y2="38" stroke="#ddd" stroke-width="1"/>
+
+  <rect class="highlight" x="145" y="55" width="200" height="56"/>
+  <text class="label" x="245" y="79">ProjectionExec</text>
+  <text class="sublabel" x="245" y="99">s['label']</text>
+
+  <line x1="245" y1="111" x2="245" y2="132" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="highlight" x="130" y="136" width="230" height="56"/>
+  <text class="label" x="245" y="160">FilterExec</text>
+  <text class="sublabel" x="245" y="180">s['value'] &gt; 150</text>
+
+  <line x1="245" y1="192" x2="245" y2="213" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="stage" x="115" y="217" width="260" height="72"/>
+  <text class="label" x="245" y="241">DataSourceExec</text>
+  <text class="sublabel" x="245" y="261">projection=[id, s]</text>
+  <text class="tiny" x="245" y="280">reads full struct column</text>
+
+  <!-- After -->
+  <text class="title" x="735" y="28" text-anchor="middle">After</text>
+  <line x1="530" y1="38" x2="940" y2="38" stroke="#ddd" stroke-width="1"/>
+
+  <rect class="stage" x="655" y="55" width="160" height="56"/>
+  <text class="label" x="735" y="79">ProjectionExec</text>
+  <text class="sublabel" x="735" y="99">__label</text>
+
+  <line x1="735" y1="111" x2="735" y2="132" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="stage" x="645" y="136" width="180" height="56"/>
+  <text class="label" x="735" y="160">FilterExec</text>
+  <text class="sublabel" x="735" y="180">__value &gt; 150</text>
+
+  <line x1="735" y1="192" x2="735" y2="213" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="highlight" x="560" y="217" width="350" height="72"/>
+  <text class="label" x="735" y="241">DataSourceExec</text>
+  <text class="sublabel" x="735" y="261">projection=[id, s['label'], s['value']]</text>
+  <text class="tiny" x="735" y="280">field access extracted near the scan</text>
+</svg>

From 1aeed35cdaad2fb12333d0daf7b50197a394d8dc Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 10:19:39 -0400
Subject: [PATCH 08/22] figure

---
 .../datafusion-53/join-filter-pushdown.svg    | 56 +++++++++----------
 1 file changed, 28 insertions(+), 28 deletions(-)

diff --git a/content/images/datafusion-53/join-filter-pushdown.svg b/content/images/datafusion-53/join-filter-pushdown.svg
index 84a51472..68bd1101 100644
--- a/content/images/datafusion-53/join-filter-pushdown.svg
+++ b/content/images/datafusion-53/join-filter-pushdown.svg
@@ -6,63 +6,63 @@
   </defs>
   <style>
     text { font-family: 'Segoe UI', Arial, sans-serif; }
-    .title { font-size: 15px; font-weight: 700; fill: #222; }
+    .title { font-size: 17px; font-weight: 700; fill: #222; }
     .stage { fill: #f0f4ff; stroke: #4a7bcc; stroke-width: 1.5; rx: 8; }
     .highlight { fill: #fff3e0; stroke: #e67e22; stroke-width: 2; rx: 8; }
-    .label { font-size: 12px; fill: #333; text-anchor: middle; font-weight: 600; }
-    .sublabel { font-size: 10px; fill: #777; text-anchor: middle; }
+    .label { font-size: 14px; fill: #333; text-anchor: middle; font-weight: 600; }
+    .sublabel { font-size: 13px; fill: #777; text-anchor: middle; }
   </style>
 
   <!-- Before -->
   <text class="title" x="225" y="28" text-anchor="middle">Before</text>
   <line x1="40" y1="38" x2="410" y2="38" stroke="#ddd" stroke-width="1"/>
 
-  <rect class="highlight" x="150" y="55" width="150" height="42"/>
-  <text class="label" x="225" y="80">FilterExec</text>
-  <text class="sublabel" x="225" y="112">k = 1</text>
+  <rect class="highlight" x="145" y="55" width="160" height="60"/>
+  <text class="label" x="225" y="79">FilterExec</text>
+  <text class="sublabel" x="225" y="100">k = 1</text>
 
-  <line x1="225" y1="97" x2="225" y2="128" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="225" y1="115" x2="225" y2="128" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
 
   <rect class="stage" x="150" y="132" width="150" height="42"/>
-  <text class="label" x="225" y="157">HashJoinExec</text>
+  <text class="label" x="225" y="158">HashJoinExec</text>
 
   <line x1="188" y1="174" x2="125" y2="220" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
   <line x1="262" y1="174" x2="325" y2="220" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
 
-  <rect class="stage" x="55" y="224" width="140" height="42"/>
+  <rect class="stage" x="45" y="224" width="160" height="48"/>
   <text class="label" x="125" y="249">DataSourceExec</text>
-  <text class="sublabel" x="125" y="281">t1</text>
+  <text class="sublabel" x="125" y="267">t1</text>
 
-  <rect class="stage" x="255" y="224" width="140" height="42"/>
+  <rect class="stage" x="245" y="224" width="160" height="48"/>
   <text class="label" x="325" y="249">DataSourceExec</text>
-  <text class="sublabel" x="325" y="281">t2</text>
+  <text class="sublabel" x="325" y="267">t2</text>
 
   <!-- After -->
   <text class="title" x="675" y="28" text-anchor="middle">After</text>
   <line x1="490" y1="38" x2="860" y2="38" stroke="#ddd" stroke-width="1"/>
 
-  <rect class="stage" x="600" y="55" width="150" height="42"/>
-  <text class="label" x="675" y="80">HashJoinExec</text>
+  <rect class="stage" x="595" y="55" width="160" height="48"/>
+  <text class="label" x="675" y="82">HashJoinExec</text>
 
-  <line x1="638" y1="97" x2="585" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
-  <line x1="712" y1="97" x2="765" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="638" y1="103" x2="585" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="712" y1="103" x2="765" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
 
-  <rect class="highlight" x="510" y="140" width="150" height="42"/>
-  <text class="label" x="585" y="165">FilterExec</text>
-  <text class="sublabel" x="585" y="197">k = 1</text>
+  <rect class="highlight" x="505" y="140" width="160" height="60"/>
+  <text class="label" x="585" y="164">FilterExec</text>
+  <text class="sublabel" x="585" y="185">k = 1</text>
 
-  <rect class="highlight" x="690" y="140" width="150" height="42"/>
-  <text class="label" x="765" y="165">FilterExec</text>
-  <text class="sublabel" x="765" y="197">k = 1</text>
+  <rect class="highlight" x="685" y="140" width="160" height="60"/>
+  <text class="label" x="765" y="164">FilterExec</text>
+  <text class="sublabel" x="765" y="185">k = 1</text>
 
-  <line x1="585" y1="182" x2="585" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
-  <line x1="765" y1="182" x2="765" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="585" y1="200" x2="585" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="765" y1="200" x2="765" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
 
-  <rect class="stage" x="510" y="221" width="150" height="42"/>
+  <rect class="stage" x="505" y="221" width="160" height="48"/>
   <text class="label" x="585" y="246">DataSourceExec</text>
-  <text class="sublabel" x="585" y="278">t1</text>
+  <text class="sublabel" x="585" y="264">t1</text>
 
-  <rect class="stage" x="690" y="221" width="150" height="42"/>
+  <rect class="stage" x="685" y="221" width="160" height="48"/>
   <text class="label" x="765" y="246">DataSourceExec</text>
-  <text class="sublabel" x="765" y="278">t2</text>
+  <text class="sublabel" x="765" y="264">t2</text>
 </svg>

From 3935a305bb3355d8b454f265dabd473564b81f2e Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 10:27:57 -0400
Subject: [PATCH 09/22] More updates

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 81 ++++----------------
 1 file changed, 16 insertions(+), 65 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 02d8a4ea..8697d13e 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -1,4 +1,4 @@
----
+layout---
 layout: post
 title: Apache DataFusion 53.0.0 Released
 date: 2026-03-25
@@ -149,20 +149,6 @@ FROM t
 WHERE s['value'] > 150;
 ```
 
-The optimizer can rewrite the plan roughly like this:
-
-```text
-Before:
-  ProjectionExec: get_field(s, 'label')
-    FilterExec: get_field(s, 'value') > 150
-      DataSourceExec: projection=[id, s]
-
-After:
-  ProjectionExec: __label
-    FilterExec: __value > 150
-      DataSourceExec: projection=[id, get_field(s,'label'), get_field(s,'value')]
-```
-
 <figure>
 <img src="/blog/images/datafusion-53/field-access-pushdown.svg" width="80%" alt="Before and after diagram of field access pushdown into a data source"/>
 <figcaption><b>Figure 3</b>: DataFusion 53 pushes field-access expressions closer to the scan.</figcaption>
@@ -178,60 +164,23 @@ work. Related PRs: [#20065], [#20117], [#20239]
    object stores.  Special thanks to [@zhuqi-lucas] for implementing this
    feature. Related PRs: [#19924]
 
-* **Variant extension planning for
-`TableScan`, new parser support for `JsonAccess`/`:` as a binary operator, and
-additional FFI improvements and fixes on the release branch.
-
-This is important because DataFusion is used as a library and planner inside
-many other systems. Making the planner more extensible and the parser more
-customizable lowers the cost of building specialized SQL dialects and data
-sources on top of DataFusion.
-
-Example use cases include:
-
-```sql
--- extension-specific table syntax / planning
-SELECT * FROM my_custom_source(...);
-
--- variant / JSON style operators handled by an extension planner
-SELECT payload:'user_id' FROM events;
-```
-
-**TODO:** Decide how much of this section should stay in the main post versus
-move to a shorter "notable improvements" section.
-
-Special thanks to [@linhr], [@Samyak2], and [@timsaucer] for these
-extension-oriented improvements.
-
-Related PRs: [#20548], [#20717], [#20722]
-
-## Other Notable Improvements
-
-In addition to the major features above, DataFusion 53 includes several other
-changes that may be worth calling out in the final version of this post:
-
-### More SQL Surface Area
-
-DataFusion 53 added support for set-comparison subqueries, null-aware anti join,
-limited deletion, and a large batch of Spark-compatible functions.
-
-These improvements make DataFusion easier to adopt for engines and workloads
-that expect broader SQL compatibility.
-
-Special thanks to [@waynexia], [@viirya], [@askalt], and [@cht42] for key
-contributions in this area.
+* **Support for `:` operator**: DataFusion can plan queries such as
+  `SELECT payload:'user_id' FROM events;`, enabling better [Parquet Variant]
+  support via [datafusion-variant]. Thanks to  [@Samyak2],
 
-Related PRs: [#19109], [#19635], [#20137], [#19699], [#19702], [#19811], [#19829], [#19845]
 
-### More Spark Compatibility Functions
+* SQL**: DataFusion supports additional set-comparison subqueries, null-aware
+  anti join, deletion predicates. Thanks to [@waynexia], [@viirya], [@askalt],
+  and [@cht42] for key contributions in this area. Related  PRs: [#19109],
+  [#19635], [#20137], [#19699], [#19702], [#19811], [#19829], [#19845]
 
-DataFusion 53 added another substantial batch of Spark-compatible built-in
-functions and behaviors, continuing the work to make DataFusion easier to use
-for Spark-oriented workloads and downstream engines.
+### More Spark Functions
 
-This release includes almost 20 new or improved Spark-compatible functions and
-behaviors, such as [collect_list], [date_diff], [from_utc_timestamp],
-[json_tuple], [arrays_zip], [bin], and [array_contains].
+DataFusion 53 includes almost 20 new or improved Spark-compatible functions and
+behaviors in the [`datafusion-spark` crate], making it easier to use for
+Spark-oriented workloads and downstream engines. This release contains functions
+such as   [collect_list], [date_diff], [from_utc_timestamp], [json_tuple],
+[arrays_zip], [bin], and [array_contains].
 
 Special thanks to the contributors who drove this work, especially [@cht42],
 [@CuteChuanChuan], [@SubhamSinghal], [@kazantsev-maksim], [@unknowntpo],
@@ -451,3 +400,5 @@ grateful for the time and effort that everyone put in to make it happen.
 [#20932]: https://github.com/apache/datafusion/pull/20932
 [dynamic filters]: https://datafusion.apache.org/blog/2025/09/10/dynamic-filters
 [235 built-in functions]: https://datafusion.apache.org/user-guide/sql/scalar_functions.html
+[datafusion-variant]: https://github.com/datafusion-contrib/datafusion-variant
+[Parquet Variant]: https://parquet.apache.org/blog/2026/02/27/variant-type-in-apache-parquet-for-semi-structured-data/
\ No newline at end of file

From c7b0478604d28dc1db364a92d6ffb69e790b2184 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 10:40:30 -0400
Subject: [PATCH 10/22] updates

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 124 +++++++------------
 1 file changed, 44 insertions(+), 80 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 8697d13e..1270076a 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -1,4 +1,4 @@
-layout---
+---
 layout: post
 title: Apache DataFusion 53.0.0 Released
 date: 2026-03-25
@@ -61,7 +61,7 @@ matching row groups are skipped entirely.
 <figcaption><b>Figure 1</b>: New limit pruning inserted between row group and page index pruning. </figcaption>
 </figure>
 
-Special thanks to [@xudong963] for implementing this feature. Related PRs: [#18868]
+Thanks to [@xudong963] for implementing this feature. Related PRs: [#18868]
 
 ### Improved Filter Pushdown
 
@@ -80,27 +80,14 @@ WHERE t1.k = 1;
 ```
 
 Now DataFusion can often transform the physical plan so the filter is applied
-to both sides before the join:
-
-```text
-Before:
-  HashJoinExec
-    DataSourceExec(t1)
-    DataSourceExec(t2)
-  FilterExec(k = 1)
-
-After:
-  HashJoinExec
-    FilterExec(k = 1) -> DataSourceExec(t1)
-    FilterExec(k = 1) -> DataSourceExec(t2)
-```
+to both sides before the join. This is especially important with [dynamic filters].
 
 <figure>
 <img src="/blog/images/datafusion-53/join-filter-pushdown.svg" width="80%" alt="Before and after diagram of filter pushdown around a hash join"/>
 <figcaption><b>Figure 2</b>: DataFusion 53 infers additional filters are from join conditions and pushes them down in the plan</figcaption>
 </figure>
 
-Special thanks to [@nuno-faria], [@haohuaijin], and [@jackkleeman] for
+Thanks to [@nuno-faria], [@haohuaijin], and [@jackkleeman] for
 driving this work.  Related PRs: [#19918], [#20145], [#20192]
 
 ### Faster Query Planning
@@ -108,37 +95,31 @@ driving this work.  Related PRs: [#19918], [#20145], [#20192]
 DataFusion 53 improves query planning performance,  making immutable pieces of
 execution plans cheaper to clone. This helps applications that need extreme low
 latency, plan many or complex queries, and those using prepared statements or
-parameterized queries.
+parameterized queries. 
 
-Special thanks to [@askalt] for leading this work.  planning overhead by caching `PlanProperties`,
-
-Related PRs: [#19792], [#19893]
+Thanks to [@askalt] for leading this work.  Related PRs: [#19792], [#19893]
 
 ** TODO get sql_planner benchmarks for 52.4.0 and 53.0.0 **
 
-### Faster Built-in Function Implementations
-
-DataFusion 53 includes performance improvements for many built-in functions.
-DataFusion, like all real-world SQL engines, includes [235 built-in
-functions], and improving the performance of these functions benefits a wide
-range of workloads.
+### Faster Functions
 
-This release includes performance improvements for 42 distinct built-in
+DataFusion, like all real-world SQL engines, includes [235 built-in functions],
+and improving the performance of these functions benefits a wide range of
+workloads.  This release includes performance improvements for 42 of those
 functions, such as [strpos], [replace], [concat], [translate], [array_has],
 [array_agg], [left], [right], and [case_when].
 
-Special thanks to the contributors who drove this work, especially
-[@neilconway], [@theirix], [@lyne7-sc], [@kumarUjjawal], [@pepijnve],
-[@zhangxffff], and [@UBarney].
+Thanks to the contributors who drove this work, especially [@neilconway],
+[@theirix], [@lyne7-sc], [@kumarUjjawal], [@pepijnve], [@zhangxffff], and
+[@UBarney].
 
 
 ### Nested Field Pushdown
 
-DataFusion 53 can now push cheap expressions such as `get_field` down the plan
-and into data sources.  This is especially important for nested data such as
-structs in Parquet files. Instead of reading an entire struct column and then
-extracting the field of interest, DataFusion 53 pushes the field extraction into
-the scan..
+DataFusion 53 pushes expressions such as `get_field` down the plan and into data
+sources.  This is especially important for nested data such as structs in
+Parquet files. Instead of reading an entire struct column and then extracting
+the field of interest, DataFusion 53 pushes the field extraction into the scan..
 
 For example the following query reads a struct column `s` and extracts the
 `label` field, for rows where the `value` field is greater than 150:
@@ -161,7 +142,7 @@ work. Related PRs: [#20065], [#20117], [#20239]
 
 *  **JSON Array File Support**:  DataFusion 53 can now read JSON arrays such as
    `[{...}, {...}]` directly as multiple rows, including streaming inputs from
-   object stores.  Special thanks to [@zhuqi-lucas] for implementing this
+   object stores.  Thanks to [@zhuqi-lucas] for implementing this
    feature. Related PRs: [#19924]
 
 * **Support for `:` operator**: DataFusion can plan queries such as
@@ -169,59 +150,39 @@ work. Related PRs: [#20065], [#20117], [#20239]
   support via [datafusion-variant]. Thanks to  [@Samyak2],
 
 
-* SQL**: DataFusion supports additional set-comparison subqueries, null-aware
+* **New SQL**: DataFusion supports additional set-comparison subqueries, null-aware
   anti join, deletion predicates. Thanks to [@waynexia], [@viirya], [@askalt],
   and [@cht42] for key contributions in this area. Related  PRs: [#19109],
   [#19635], [#20137], [#19699], [#19702], [#19811], [#19829], [#19845]
 
-### More Spark Functions
-
-DataFusion 53 includes almost 20 new or improved Spark-compatible functions and
-behaviors in the [`datafusion-spark` crate], making it easier to use for
-Spark-oriented workloads and downstream engines. This release contains functions
-such as   [collect_list], [date_diff], [from_utc_timestamp], [json_tuple],
-[arrays_zip], [bin], and [array_contains].
-
-Special thanks to the contributors who drove this work, especially [@cht42],
-[@CuteChuanChuan], [@SubhamSinghal], [@kazantsev-maksim], [@unknowntpo],
-[@aryan-212], [@hsiang-c], and [@davidlghellin].
-
-### More Visibility Into Parquet Pruning
-
-DataFusion 53 can now report the number of matched Parquet pages in
-`DataSourceExec`, making it easier to understand whether pruning is working as
-expected.
+* **Spark-Compatible Functions**: This release includes almost 20 new or improved
+  Spark-compatible functions and behaviors in the [datafusion-spark crate].
+  This release contains functions such as  [collect_list], [date_diff],
+  [from_utc_timestamp], [json_tuple], [arrays_zip], [bin], and [array_contains].
+  Thanks to the contributors who drove this work, especially [@cht42],
+  [@CuteChuanChuan], [@SubhamSinghal], [@kazantsev-maksim], [@unknowntpo],
+  [@aryan-212], [@hsiang-c], and [@davidlghellin].
 
-This is useful both for users tuning workloads and for contributors developing
-new pruning optimizations.
+## Stability and Release Engineering 🦺
 
-Special thanks to [@nuno-faria] for this visibility improvement.
+The community spent significant time this release cycle stabilizing the release
+branch and improving the release process. While such improvements are not as
+headline-friendly as new features, they are highly important for real
+deployments. We are discussing ways to improve the process on [#21034] and would
+welcome suggestions and contributions to help with release engineering work in
+the future.
 
-Related PRs: [#19977]
 
-### Continued Work on Execution Correctness and Stability
-
-The `branch-53` release process pulled in a substantial number of fixes for
-correctness and stability, including fixes for spilling, hash join edge cases,
-FFI behavior, filter serialization, parquet leaf index mapping, and a regression
-in `count(*)` with right anti / mark joins.
-
-These are not as headline-friendly as new features, but they are highly
-important for real deployments.
-
-Thanks to [@jonathanc-n], [@alamb], [@xanderbailey], [@haohuaijin],
-[@friendlymatthew], [@fwojciec], [@Kontinuation], [@nathanb9], and many others
-who helped stabilize the release branch.
-
-Related PRs: [#20726], [#20791], [#20792], [#20882], [#20883], [#20884], [#20890], [#20891], [#20892], [#20895], [#20898], [#20903], [#20918], [#20932]
+Thanks to [@comphead] for running this release, and to [@jonathanc-n], [@alamb],
+[@xanderbailey], [@haohuaijin], [@friendlymatthew], [@fwojciec],
+[@Kontinuation], [@nathanb9], and many others who helped stabilize the release
+branch.
 
 ## Upgrade Notes
 
-DataFusion 53 also includes several breaking changes, including updates to the
-SQL parser, optimizer behavior, and some physical-plan APIs. Please see the
-[upgrade guide] and [changelog] for the full details before upgrading.
-
-[upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading.html
+DataFusion 53 includes some breaking changes, including updates to the SQL
+parser, optimizer behavior, and some physical-plan APIs. Please see the [upgrade
+guide] and [changelog] for the full details before upgrading.
 
 ## Thank You
 
@@ -401,4 +362,7 @@ grateful for the time and effort that everyone put in to make it happen.
 [dynamic filters]: https://datafusion.apache.org/blog/2025/09/10/dynamic-filters
 [235 built-in functions]: https://datafusion.apache.org/user-guide/sql/scalar_functions.html
 [datafusion-variant]: https://github.com/datafusion-contrib/datafusion-variant
-[Parquet Variant]: https://parquet.apache.org/blog/2026/02/27/variant-type-in-apache-parquet-for-semi-structured-data/
\ No newline at end of file
+[Parquet Variant]: https://parquet.apache.org/blog/2026/02/27/variant-type-in-apache-parquet-for-semi-structured-data/
+[#21034]: https://github.com/apache/datafusion/issues/21034
+[upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading.html
+[datafusion-spark crate]: https://docs.rs/datafusion-spark/latest/datafusion_spark/index.html
\ No newline at end of file

From 482ae210aa0a9644025fe83e4ffa4ea41d924e04 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 10:44:14 -0400
Subject: [PATCH 11/22] grammar

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 54 ++++++++++----------
 1 file changed, 27 insertions(+), 27 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 1270076a..fc004d2f 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -39,13 +39,13 @@ making this release possible.
 
 ## Performance Improvements 🚀
 
-DataFusion 53 continues the project-wide focus on performance in general and
-this release features reducing planning overhead, skipping more unnecessary I/O,
-and pushing more work into earlier and cheaper stages of execution.
+DataFusion 53 continues the project-wide focus on performance. This release
+reduces planning overhead, skips more unnecessary I/O, and pushes more work
+into earlier and cheaper stages of execution.
 
 **TODO: Add performance chart and short discussion of the benchmark setup and results.**
 
-**TODO: Also mention we updated the clickbench benchmark to use DataFusion 52 **
+**TODO: Also mention that we updated the ClickBench benchmark to use DataFusion 52.**
 
 
 ### `LIMIT`-Aware Parquet Row Group Pruning
@@ -58,7 +58,7 @@ matching row groups are skipped entirely.
 
 <figure>
 <img src="/blog/images/limit-pruning/pruning-pipeline.svg" width="80%" alt="Pruning pipeline with limit pruning highlighted"/>
-<figcaption><b>Figure 1</b>: New limit pruning inserted between row group and page index pruning. </figcaption>
+<figcaption><b>Figure 1</b>: Limit pruning is inserted between row group and page index pruning.</figcaption>
 </figure>
 
 Thanks to [@xudong963] for implementing this feature. Related PRs: [#18868]
@@ -84,28 +84,28 @@ to both sides before the join. This is especially important with [dynamic filter
 
 <figure>
 <img src="/blog/images/datafusion-53/join-filter-pushdown.svg" width="80%" alt="Before and after diagram of filter pushdown around a hash join"/>
-<figcaption><b>Figure 2</b>: DataFusion 53 infers additional filters are from join conditions and pushes them down in the plan</figcaption>
+<figcaption><b>Figure 2</b>: DataFusion 53 infers additional filters from join conditions and pushes them down in the plan.</figcaption>
 </figure>
 
 Thanks to [@nuno-faria], [@haohuaijin], and [@jackkleeman] for
-driving this work.  Related PRs: [#19918], [#20145], [#20192]
+driving this work. Related PRs: [#19918], [#20145], [#20192]
 
 ### Faster Query Planning
 
-DataFusion 53 improves query planning performance,  making immutable pieces of
+DataFusion 53 improves query planning performance, making immutable pieces of
 execution plans cheaper to clone. This helps applications that need extreme low
 latency, plan many or complex queries, and those using prepared statements or
-parameterized queries. 
+parameterized queries.
 
-Thanks to [@askalt] for leading this work.  Related PRs: [#19792], [#19893]
+Thanks to [@askalt] for leading this work. Related PRs: [#19792], [#19893]
 
-** TODO get sql_planner benchmarks for 52.4.0 and 53.0.0 **
+**TODO: Get `sql_planner` benchmarks for 52.4.0 and 53.0.0.**
 
 ### Faster Functions
 
 DataFusion, like all real-world SQL engines, includes [235 built-in functions],
 and improving the performance of these functions benefits a wide range of
-workloads.  This release includes performance improvements for 42 of those
+workloads. This release includes performance improvements for 42 of those
 functions, such as [strpos], [replace], [concat], [translate], [array_has],
 [array_agg], [left], [right], and [case_when].
 
@@ -117,11 +117,11 @@ Thanks to the contributors who drove this work, especially [@neilconway],
 ### Nested Field Pushdown
 
 DataFusion 53 pushes expressions such as `get_field` down the plan and into data
-sources.  This is especially important for nested data such as structs in
+sources. This is especially important for nested data such as structs in
 Parquet files. Instead of reading an entire struct column and then extracting
-the field of interest, DataFusion 53 pushes the field extraction into the scan..
+the field of interest, DataFusion 53 pushes the field extraction into the scan.
 
-For example the following query reads a struct column `s` and extracts the
+For example, the following query reads a struct column `s` and extracts the
 `label` field, for rows where the `value` field is greater than 150:
 
 ```sql
@@ -140,24 +140,24 @@ work. Related PRs: [#20065], [#20117], [#20239]
 
 ## New Features ✨
 
-*  **JSON Array File Support**:  DataFusion 53 can now read JSON arrays such as
-   `[{...}, {...}]` directly as multiple rows, including streaming inputs from
-   object stores.  Thanks to [@zhuqi-lucas] for implementing this
-   feature. Related PRs: [#19924]
+* **JSON Array File Support**: DataFusion 53 can now read JSON arrays such as
+  `[{...}, {...}]` directly as multiple rows, including streaming inputs from
+  object stores. Thanks to [@zhuqi-lucas] for implementing this feature.
+  Related PRs: [#19924]
 
 * **Support for `:` operator**: DataFusion can plan queries such as
   `SELECT payload:'user_id' FROM events;`, enabling better [Parquet Variant]
-  support via [datafusion-variant]. Thanks to  [@Samyak2],
-
+  support via [datafusion-variant]. Thanks to [@Samyak2]. Related PRs:
+  [#20717]
 
 * **New SQL**: DataFusion supports additional set-comparison subqueries, null-aware
-  anti join, deletion predicates. Thanks to [@waynexia], [@viirya], [@askalt],
-  and [@cht42] for key contributions in this area. Related  PRs: [#19109],
-  [#19635], [#20137], [#19699], [#19702], [#19811], [#19829], [#19845]
+  anti join, and deletion predicates. Thanks to [@waynexia], [@viirya], and
+  [@askalt] for key contributions in this area. Related PRs: [#19109],
+  [#19635], [#20137]
 
 * **Spark-Compatible Functions**: This release includes almost 20 new or improved
   Spark-compatible functions and behaviors in the [datafusion-spark crate].
-  This release contains functions such as  [collect_list], [date_diff],
+  This release contains functions such as [collect_list], [date_diff],
   [from_utc_timestamp], [json_tuple], [arrays_zip], [bin], and [array_contains].
   Thanks to the contributors who drove this work, especially [@cht42],
   [@CuteChuanChuan], [@SubhamSinghal], [@kazantsev-maksim], [@unknowntpo],
@@ -172,7 +172,6 @@ deployments. We are discussing ways to improve the process on [#21034] and would
 welcome suggestions and contributions to help with release engineering work in
 the future.
 
-
 Thanks to [@comphead] for running this release, and to [@jonathanc-n], [@alamb],
 [@xanderbailey], [@haohuaijin], [@friendlymatthew], [@fwojciec],
 [@Kontinuation], [@nathanb9], and many others who helped stabilize the release
@@ -212,6 +211,7 @@ grateful for the time and effort that everyone put in to make it happen.
 [@fwojciec]: https://github.com/fwojciec
 [@Kontinuation]: https://github.com/Kontinuation
 [@nathanb9]: https://github.com/nathanb9
+[@comphead]: https://github.com/comphead
 [@cht42]: https://github.com/cht42
 [@CuteChuanChuan]: https://github.com/CuteChuanChuan
 [@SubhamSinghal]: https://github.com/SubhamSinghal
@@ -365,4 +365,4 @@ grateful for the time and effort that everyone put in to make it happen.
 [Parquet Variant]: https://parquet.apache.org/blog/2026/02/27/variant-type-in-apache-parquet-for-semi-structured-data/
 [#21034]: https://github.com/apache/datafusion/issues/21034
 [upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading.html
-[datafusion-spark crate]: https://docs.rs/datafusion-spark/latest/datafusion_spark/index.html
\ No newline at end of file
+[datafusion-spark crate]: https://docs.rs/datafusion-spark/latest/datafusion_spark/index.html

From b0225764033a7b7378bda7b4bfd9b7b0dbe95957 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 12:06:42 -0400
Subject: [PATCH 12/22] clean

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index fc004d2f..1576c15c 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -45,9 +45,6 @@ into earlier and cheaper stages of execution.
 
 **TODO: Add performance chart and short discussion of the benchmark setup and results.**
 
-**TODO: Also mention that we updated the ClickBench benchmark to use DataFusion 52.**
-
-
 ### `LIMIT`-Aware Parquet Row Group Pruning
 
 DataFusion 53 includes a new optimization that makes Parquet pruning aware of

From a980b7bcc19216d8ad1b7a933ecd796914bf204e Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 26 Mar 2026 16:45:11 -0400
Subject: [PATCH 13/22] update planner benchmarks

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 1576c15c..e78e6c85 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -92,12 +92,11 @@ driving this work. Related PRs: [#19918], [#20145], [#20192]
 DataFusion 53 improves query planning performance, making immutable pieces of
 execution plans cheaper to clone. This helps applications that need extreme low
 latency, plan many or complex queries, and those using prepared statements or
-parameterized queries.
+parameterized queries. In some benchmarks
+overall execution time goes from ~4-5ms to spends ~100us.
 
 Thanks to [@askalt] for leading this work. Related PRs: [#19792], [#19893]
 
-**TODO: Get `sql_planner` benchmarks for 52.4.0 and 53.0.0.**
-
 ### Faster Functions
 
 DataFusion, like all real-world SQL engines, includes [235 built-in functions],

From 859f29c481c539066c4ecd8017f2b809528ffd3f Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Mon, 30 Mar 2026 13:40:17 -0400
Subject: [PATCH 14/22] Add performance results

---
 content/blog/2026-03-25-datafusion-53.0.0.md  |  18 ++++++++++++++----
 .../field-access-pushdown.svg                 |   0
 .../join-filter-pushdown.svg                  |   0
 .../performance_over_time_clickbench.png      | Bin 0 -> 57883 bytes
 4 files changed, 14 insertions(+), 4 deletions(-)
 rename content/images/{datafusion-53 => datafusion-53.0.0}/field-access-pushdown.svg (100%)
 rename content/images/{datafusion-53 => datafusion-53.0.0}/join-filter-pushdown.svg (100%)
 create mode 100644 content/images/datafusion-53.0.0/performance_over_time_clickbench.png

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index e78e6c85..9fe4aa3a 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -39,12 +39,22 @@ making this release possible.
 
 ## Performance Improvements 🚀
 
+<img
+src="/blog/images/datafusion-53.0.0/performance_over_time_clickbench.png"
+width="100%"
+class="img-responsive"
+alt="Performance over time"
+/>
+
+**Figure 1**: Average and median normalized query execution times for ClickBench queries for DataFusion 51.0.0 compared to previous releases.
+Query times are normalized using the ClickBench definition. See the
+[DataFusion Benchmarking Page](https://alamb.github.io/datafusion-benchmarking/)
+for more details.
+
 DataFusion 53 continues the project-wide focus on performance. This release
 reduces planning overhead, skips more unnecessary I/O, and pushes more work
 into earlier and cheaper stages of execution.
 
-**TODO: Add performance chart and short discussion of the benchmark setup and results.**
-
 ### `LIMIT`-Aware Parquet Row Group Pruning
 
 DataFusion 53 includes a new optimization that makes Parquet pruning aware of
@@ -80,7 +90,7 @@ Now DataFusion can often transform the physical plan so the filter is applied
 to both sides before the join. This is especially important with [dynamic filters].
 
 <figure>
-<img src="/blog/images/datafusion-53/join-filter-pushdown.svg" width="80%" alt="Before and after diagram of filter pushdown around a hash join"/>
+<img src="/blog/images/datafusion-53.0.0/join-filter-pushdown.svg" width="80%" alt="Before and after diagram of filter pushdown around a hash join"/>
 <figcaption><b>Figure 2</b>: DataFusion 53 infers additional filters from join conditions and pushes them down in the plan.</figcaption>
 </figure>
 
@@ -127,7 +137,7 @@ WHERE s['value'] > 150;
 ```
 
 <figure>
-<img src="/blog/images/datafusion-53/field-access-pushdown.svg" width="80%" alt="Before and after diagram of field access pushdown into a data source"/>
+<img src="/blog/images/datafusion-53.0.0/field-access-pushdown.svg" width="80%" alt="Before and after diagram of field access pushdown into a data source"/>
 <figcaption><b>Figure 3</b>: DataFusion 53 pushes field-access expressions closer to the scan.</figcaption>
 </figure>
 
diff --git a/content/images/datafusion-53/field-access-pushdown.svg b/content/images/datafusion-53.0.0/field-access-pushdown.svg
similarity index 100%
rename from content/images/datafusion-53/field-access-pushdown.svg
rename to content/images/datafusion-53.0.0/field-access-pushdown.svg
diff --git a/content/images/datafusion-53/join-filter-pushdown.svg b/content/images/datafusion-53.0.0/join-filter-pushdown.svg
similarity index 100%
rename from content/images/datafusion-53/join-filter-pushdown.svg
rename to content/images/datafusion-53.0.0/join-filter-pushdown.svg
diff --git a/content/images/datafusion-53.0.0/performance_over_time_clickbench.png b/content/images/datafusion-53.0.0/performance_over_time_clickbench.png
new file mode 100644
index 0000000000000000000000000000000000000000..74f23c18d65c4058896f851cdf536d33575786cc
GIT binary patch
literal 57883
zcmeFZby!s0`UZ@M0R}oCAV_zPAR&#?F@SUq-6h>1qT<lqFo-a8cMH-WHFPT7$Ph#G
zZS?(}^Pb~-uJ5n!zmMyhi#>bqwb$CudY*edaj$_X%1hqEA;!VLz_=$ZC8mUdaf1;9
z1M}YPo4}c&n3Dz!jB6qmqN0k@qM{Uv4tAy%)+QJjL^0ms3e4|b;kWx5Nl@e8DUCoV
z8cQwAdGx*vB=*0{yrW|z@eMw8+8Al~9ZzF|iodSp6@kzjs_w~dwQnERoVV|qK3F+)
z@)1PWR3ZK7SPC4kQRuziYMr)bsA+tNBeakeM&LnI68-t}?$KMZqs_TIudx38saJbr
zJEy*R&dgg3Yj)|BnKp!8*IRZ4)kz2jO?nhtZhd&hCB%`OyX($X7Wold?~ZG3eg&-;
z{S$evqlBA%D=s4)cTG<YoeZ2OoL+&0XkbspqcepaqDFLrsS>@dT{^?tXld-qlb=}p
zaK~Y2zkNrNh!B!=ele{*!#w+?H{pxLOd0o&7QgL_S{}E%qzbQd%1yNOruJj5_Z^C5
zwA;{{3+>m(aoxTzy41V}l}%OWy>>~h6Ek;>SzoKNo;Pa|5UHAphO{XJg24nF-^REW
zV1a=J99;uG#J~pw<9g%=j9b8e5bzO8!~DOiHyG2d|KBm@y{j8vD2qx<1OJtc9864X
z9nI~W?ouxf0*@NAP|<MGfXML~+1aoe7~2_|u({d1zG{LY;Km0W+L$;QP`KGx+dA^O
z2~z#KgAX{qIt`|x_;rhul^~S{M3F+&&cTF&o9!vvQz{`G3JMAV2V+w{B{7M=ngibi
zsmz_6Uh{#$uCA_ZuFu%)9L&J%yu7^NryO7o4p!g}R!4VRCj&QDTSw~O5Bb06h?zJV
zIas`QvaqwIxO%REp`EjnAQjbBNB{c!Jx&uhi+}fI>-g7e0WS!?x&mfrdkX&7vw@}p
zS7-SYE!<43HN`A!0Pz5Q2yt<63H-YMk1PM~@oz0P{@s%O83+5{oBr+6e>YWiG;t8M
zvjO^a68iVl{MGpH7yoJ~0KR(lzlq|voPV7K1TBOk0RGpc3E{|_vx#G1h+s&Iy-;zx
zwwZD(S!Lw7X-?C_uH@63H-W@f_>Zo&F$AWwGo=fE9&Bx5Ezu7m3ejzT$uch6szFt0
z{!v2wUf|nyeF#<dn>4CGOseAHluIw%)ZOjP-py%7eiL|H|5J3T=epx|^2BKhIvFnq
zTEbyTghhdI?Y0O8=7TpF|9R*~^)Q=BsRsYfKVJsUet9U8u_*E1?f!LppbSPtY4#E0
zlRuyR`yQY*+5g}AzXbpPDmC*&*0&>h>Qk<=jTP&Un5`k3ME?IZb^n?I;#+=*Z;6_D
zv8)BUr}sx$T$1xlXqDKY<pXMkIu;uRRjD>ad2*_c?kCi^V69R9vnD9AVOd&QT8sC(
z+c&?xF`6!ZXJb6aQC>AnrAXuPy{NBacN$A8ZKkw3LP@{UP_zA~7X51|6pDUW1YoGO
zY#dt!c|n!WXu+K&Oh#N8tnPfxs0aT;MIK3t^aT&y%_KFodK?t3<*%e%znj)ez_s>2
zlvXhTE90>YfAoj=5EB+9%cv&>FJLeJLAQU8N{WVJOOe=+JGMkBN4H+Pox7{~e>@iu
z3=JsUh2b7MNsmT`4xZzPG2xE&gQ%PTW85(cB9osuTdnggxM5P)|1E=m3s?sBC1115
zdcGA_D8sisk{5TdJu}s?@qF%ZEel@DQs=x8>pWW2aF{4t_l<i>Bj7%%oO@pi{}?H=
z+#&q(tfTS#P<vyxp)Q97I5XPh>x(8k6kej#E;W`GuwdKzXr^RvtI^l1Y<+>0yRpAY
zW9R(v93n5S;C(TgJVZ|?ks(LZ!ut<XrHL1gL4e=nCGj4=8S3`qeYMD1Pn2R#ob}J|
z&6WK^r$@59O`hz7W-6VP#~i5L?KTKl>E>Sab0*4i@6$thntRZmrRT4zdpL_uuwK9w
zKJE3FljSNrwXHlrZCM8Ie2fk$8MAoyN5^RxSfI8#{fD6TdRP6Y0%Hrtm?nm)vtK#L
zUz!+_&~A5?K4cQab$@jtKm)o7Ilj54rUPdq0?(9;d-@G!p-q@lCTaF;cA-gZ;g^9g
zzo20|o^!r@xQDupOk??wJK7D_nLULj9ix?oyC}Yn=I6&uJn^_xCgC?b{>%KX6u=nr
zB}hlp>;Z?t>p=8~x?T#$f%{6ziU2Fu7XMhz9F%iE)#1fXP4puvd&-3IXrfrR!u-hh
znGETp_$U?3IKr0bfD^?ceU03D66NY0y9!>0dAh*SnrYnBEG+4rtz6fG_rhSYCMmB?
z9*1Ix@4k)p#ggk`lrp*)-QHjuG2UboHZ_6>@t&Bp9d16kJJpN`O(i0H=Iy<A5Gs`b
zj$Hgxqs)!5KsOSdzQuJ$%B)JqvCVJ^r%ZK35G9zu*i|Bnp-*vAGK@}7@wBTcR?hdX
zmBV5NKQWv4UfWY9j``7l=UQiwd$VWygQdmDtRkbQxY(^P*!V^pw#3EAc{H_rSx4$n
zKgpsRo8G=P$#tsv(BrR!xx_=>TUC)aR=d#ltjT4f&+Uv<`*6gy(s0lrNQjs<UL>c5
zV|Ipr>}C|%pGM->P4)dov3?%Syp(=~)Pf#LU~jnrBP#R2p2Hi?>u8F2<rugk-Ymy~
zji-LM#UVUA#)3~Tj^grU^81ZyvB82yr;s1D+q(9<IwPi}%)N7C6Ap?(Aexx46WI=r
zwX$fd*@R(>i8EUHG?i%xmCd|#ZWDAllosl9FuSt*^W_r?C4#q1?l1p924c!<Fh-~f
zb|5dPv3_8FZqJW?Fsigs;A^|O4=AkVHY|_?zhr1cnx*tdgasd%ldjM)p-^5&R$ip#
z)3!~+sn8)>;-?mBcumw>s8P=!t*uzC!q&x|DFgRBSFgCaB%ifqL9$s43-aozpx&gP
zvNO=FoE~s_p3_t@ml4I^kG7qKaT(RRQa$YZ2Qcmod_y?>R@sd<T6%gHY3PqBr}iET
zWZ~Uq{jNiVzyi0yEkmlA`vxBE9kk6p6xMvyY-EpZ1?|L%>O;%YS=4_ari0KOGs%}S
z4m1}VvpoC8uX@qEU$eM#I@EM36f3g!^~|%l_ayXO5aDGkG^fw^eytXI;lXWkWA84R
zRa7#%qro7^l<l=)M$_T9{o_OU>0CK2oRwpvG!>U^oS(|1c_5PgPt&%482);`0qxb}
zT&PYd2G)~6?D*fp?c=l*Z^d&giQ`G#A$*?$S2~cS6f6Gl_Hl>=D+tD^V(-o<c}}(a
zYNs(Q(&Py~(<3mPDzkfIqB4rL^oDcbY`N+!ZF9X*Y_J#xl?i2bB>kTqKLkn9mS%uj
z9mHt{cSHHNKbql9)6kQ}fAu4LOY`7{;6-#Mx?7fjZLli~qF&!Bs>wI_l)Zt-)2`!0
zfIX&Yy^wSLdYigd1L%|{zGz+V_Fivv)e6Dgo{C;ZWyE5lc=x^`U8)EA?I)@Yjpd>>
z!QdQ)*4dRPwt{&EpK*?%rL`i9K(t+dLo>YA6>=<30A*GpQ+U9*cRMQo&vhmw;mv4I
z)EiK8?D`EbjQ^M$f{s<p(XETx+Dr*Uwcxx=z=_UHYGu?M=kVf-t4_V7kzi_su3s-u
zU@Vbcxkzj|uMsO2yH=J;Mcii6FP98DeeT9;tw-lie<4IZ$I~^hszpR)l6k<e<@blT
zfDe1qN(|EnY1DV7=SqzaR=gY+I^A3-jZxHuA3qm@c<-Jc=xcjv<f5P2H@qQH!{1+1
z8lfcG-uTS%*86+I^GE7q59s}eBgs7}$DXMIE0Ru;S7H-A4(0YrS?Y1`od^y%XV;xS
zU|^sVTj~r&m59~z;U`wq>MGA*0<R-tN=|Zx!(GeLkrb?Wf-Zh(LRrhj;9`>4)XEIj
zH;I(6_a<L`o-jZ_*Cf~7OCv9o<!+_~<+fLm6lR^tA{5@;_g<Me*|-n{aI0W?cFqI+
zRf}XDKlkEe^m{7V#y{ZR4~PD{cu}kAO~FG@(Xk`g_+(FKj#g$pM$T*hZ6+bZxo2th
z!H8(OO0hTS^DqHeYa~v{ZN_uXJz@aPEN>VuJjO*1CPMsFp#w|n?O1OlGVQd3QW?}v
zT=M41);!_Rops3r*-{_K=0fY9zEI5*^c0$Bcd;v!R^{f`g@`@3H&O|b@y0j)a1zYr
zu<DN<RenPivqX<a=CVP<_c2X!-45P+zUaOaQ}fZsr@I1fS->S2Q#9@-G^^YnM@`w2
zB}<!}Y3YyJKTvoS@N@ow6s-#a2Q1HeN1KVt#NarR|4%DP$BvF#U3xF$EcNrb^{pFI
zPXz8KDTsE&5ls#ds7c+^ZhE;wo*WC+(`>Fbn0nz=CUp=PGbkt0du&{Ju!f|M%5MEr
zq5g^|h$;OZKQq+`gLr<hbyr=*r}>}WK1H}B=ErFFchYKqOweC~5pl!Jmwuf3nDH;u
zsWkmYTz9%%XM=?BR8We3cus|rS6Em{z^bLOxBQJHc<5+%!Db{^)uorcgHt0Rfq>I$
zl*RLSDbD7~iLJEJ=HISf$amftO`xNP3Sy-XW`BbV&&9IoS_*pXn-828rsgg=xlfdu
zs3|KOZB176Fv4Hb`s}t5Zha0DDutgd(pUDd)zPq;9;}ZX9<RXbcJ&($M#~TotILV=
zv8Ia=0yZ7<4szdmS@_n`F1qP*CdP3dMz2K!lGciIz)pT<7J9xwep!6&`b`b?QMsOI
z<VI1GtzE-G?3&OB9YFx8i-x?BZ~UouTi_#vr|g1x6di=4z#-d>X`%6S`&kHBlH9wT
zq)yaC#Q1mkCK8Kfd07!{LPzsgI6wm;;xrF0(5)N2A$+_THJT|K=Y}tcagR)3+-1^&
zzV62dlKCG?PH6$Hx5&K)e9jigJH)%QQa!pWR?__Mm(HDUyWSg37QQ_0r$Z!sjBfUM
zb>kp>s=`X0AJs{{WhQ*83Ejwl|5Y;di6(Qu=zU(Sh*E-xP5JV?l^AuX#pTIZ(>(my
z5)HdtuQImTL`<w_F5tMX9Dd`O8%`R$Nn))ZCU_XXCKT}<+$=Ni!kF-`Lzb<kHN@f)
z?R%-7W!W#V7d+!KWmD<4JFhiXIqe|Jzds;q({z4VmYd6j47=0a6BIoWN**SBQSE)U
z+u_oJO`q|Ik=$dIQoTq&bN|>5i3pi`K0mCiP+|Hl@dlXZsKjS@pgNb&gjtEsD$9p+
z^7YqyHmd`lT)z9=N_4^R8?~)nqx0M!lw9aiRmUDUVe4_(RH&Z=gok(9qjzEGxm6Lj
zRXJt5Tk7;0P=u*yM&(UZ511!p<$ZB1t9C8<?KAJ>=b?n@a;SH|G4>+{6)pPOFPp<p
z{-WzTqbi#p>H48JDE>x~$n{5Q0s1x^V`XMX;^`#WfxCu2CrCJ--2k}hY^c2WpkfSa
z<?&SLd>^@B?;u!xFsiLfpswp2bFeYi?@k(%Yo=3ZW7D}0_dZM@_1rA!^HfV;O!eHh
zAmw){cb)SdoAug*@Skp$ZDmBsJCG^f(=2i_>3SxQPOO?JGppMR;hyYte{=F0d%1M0
zYHk)V9F>mKfez;7x^1%w)~abkQa-0_)tItAzMca=O!Ta$-NsB^mDS4WPLmxnOMxeI
z#wbiM`L;T<!Dhz5LV$p>2{D{|%K91lSg+I=*09&df7ptrFixoFWiY1W7!su1cX{m6
z{q@yZ)9E(<>3Uv~_4ag)e8xx^ALymOaI(oH6nh24K<fbYJwK>yBhY?sH;yZGM3Ge@
zyz8T`C$%}c!v88mB>{BC@4B6AQBfA?I{Zp<!gHs=OhNd(9B)jO?dET&VBowa*8LKX
z&iBeLLgL@lr*qy@#pOj23|&`8&{39!D-+&{@V&e^hZKD%)>#J^rOSdFIH3?2epZT0
zqUX`YNRf}KS)fg&NtdkkLI_W#-}PIyrJx(6H%yaU>rs!N;HL+_TugYOtP@>g)WPRh
zYB$%&?XkbArCbm9y>MSK$jrRG$TgayoF!KVAqMMh39KYLS}yH0%-k0~dyQlsq)9Zh
zN*Luux;1;4uVG^Ipgs;ake0_mJ-V=?3$!X64CAp%(<}n3>*tg-MMKC@@nqcA8ui4W
zhNXv0Mbk><)+f?i;jV_oDWD4)kX1eZQnd24lj(sfhpg&yz5AYy6ZQbKJW3<pVlK0r
zr}w;YM`_19DSzAGtI~+{v6NGs<k=&2ogZ$GQ2e~s4KUnaQj3uWk~JKsF=|)gG6Obd
zBBm><i+(De-DtCqf2aHoTQ!AE!(wJSqhnFyNp9Cy$#ld!0&KHHGDc*%DMWVKzJqj&
zu<7W15(nd;2$=`Snyqe=qw3;^o$sl3TwHw|r!(YO>f<SZgEm!jK6OcCIXR(;E^OD@
ziU_l~=A%1$6P~5q3auAhvUmL1e2c%o+zCl}xd*?j<_LQ}e>-lc6^5sK9?7M(ltVUK
z-k9=yO?N|924m2_yngX5T^Nh|PlnX>qP1nMez}f`$@~19`cpPlDR_E|sx^Qj@`Kf}
zbuIk`zH?^2QvTT1tq_Cn3h|Eq^6B6KSApKar1O~WjFVUi(gSw6b9pYTUU80|n=6<>
z;K5W-vnY4wZ#65a+E$i3)L{aKpc%HBMNrhr%pGvOJ_y_Ad_%A9B-<MH0EOXwIANCJ
z8X=&`HPWbEW!r~@q(>rVx{Y?ygV=Jp-m4%3RRfBeE+!$YY9w5=>d=qcHWjK*vWhw$
zn<X1)Nh&I{w$-Z_LD~An*kzn!!Wm)%vl^$(;NEiR_j+N<HW89>^>xO%=&M*16-=1C
zC}MS(<8@HAjclz@K+m}9p^uMOgfG{`-an+47=MS2C#MV>Q2?pGrE6YZPO@`B=dIHt
zRUBr5kO+t)Ty*9#k&1cy!f`c$W`o52+p=Y1H(SJ%=fX{4^gOEz+~*``Lw!V6IeX~`
z;zEUQ=5y^1jmizysLC4)^kj2x3n6S$MoQ_h1UY3RydCWKK@CFM7i5YW1E3CEUr6ek
zb!zu6ORq=<x*a~&owQTFgq)N_f_!pVpTM3Rk33CKRxyNoqE&ZDFor?H!~uT7t_~eP
zk+TerTzgi)J{LyF_Vf@g6oo7B^NodNRd=*mR{PlX`_I2f+Cx*((~e=MezF@A-O3th
zhxv07h45xSjq}|wU(SM=oV8uoY5Ra4Bz?ioY<N{hd^acC`Ht_UW5+zR?a4OEsEO^U
zUy5!);-`=A=~mUuih{d&@X$#dhxf{wCj@y~akqv1;%9&sv=fxDG?T~vQs-PF+c~P?
zeNNO+e$&OtcG<zSV}Ff(+vWGgy#><WK5e)i;Jh5u$@u(oUKE4gjIItj*}<ex7X#7l
z#t1I5g8|k-ne>nKt@{t2+fF~{fed*jGNiVW!}vioG^<_p8q`1J+|9kY6Y6=L3j>}B
zF3O7FeCX*>PSTT_j#zu4Iv*@x%{}Mkh%To8w5Q;E5`{P2Cw%Ee2%hjwOwFzBdL~V0
zKx?nd*Y+yK+xw9M?kCWyF=&h?S7x*BXUm-#(Go1SiQydO1uNZkKW*I0OF$wiY1sio
zI=!Cw(i+3ptKRWdh#%*IIb$?h;!ea^Xw@vbKh3W;YrK5R1A7WmHty~xmwgx8#Vm>2
zV3<}v5{G&j^(H^SL5~nd)U7uwfO?BU9C^5natTiN&^!$cXrb8-{PfQD{YK{gJR{{u
zQnl=A<VS`VI%wDlbd8I2*t7fC<7}Cy$zIyo*HR{@g_hWS0R(oqbt>hZ<FTHKJf~*Y
zc<JBz5XQ*?`<k0nZHZI$u2Kd5unmnF&TS=Ap|$|_J(e9)Uw@X+ieo=0K?I`;8*SL@
z#aDk@fN;+(YK3gH2f~JO!_Lol0QFAt`WS7Q)hc0k;aF7L?WC96-ztxOL&)k~z0UMP
zhN=lpkYq9Cl>plTO~(}Z5QUo)aLz2$&fynk;wx!#*cp=Z*q<r7758;(do%h$TFOvE
zE1ZL|V7={@HU)I_dQu-}Xm=~qDr4N5_j-p8tc@e2Cbz29>EIiXJlbkG(0i&-A>Qk>
zv+)R@@xVR8f<>{8D^%ve7^m|&MP$L6B;C^IZ~EoCAypJBD(HrGn-O)y>0p=@Uu<~3
zCo=3ja6y<lA(*83)O)ErMqWprMbe~`=4CF6Z{br|tj}Z25XNp%_jP{Nxu%m{aRP)q
z+lx)%bCFNGTJs!KnOQBrU4HyP5#&n~x&MoQ9)_D!3`*kWqMi(2H{^(VP8%V$QDGsI
zA@&rT7WDzepCco;kFfGM5(1p$<ah>qcJoOaRcJCOPw4EYD1EpLOXNxl+2$j*ej@h~
zbG{@(M%o8m{v8TS%BkJjNdAMgW${FP&74r$f%PRwdL|YHHfgHA%l^Sj?P484IT`)%
zTQOoXFz*XW6)HHoSj9t7Z2bkVHl8c?I}YqJ(9HgWe2ztL2KHA=mlx<J*B1Z!JOOa#
zE6=kJ`14t}jIq5u(qa_pi=^$=@7PCwp}98cso8Z$?QKcx(*x~+^3>!Lxea2=6{m)J
z4Ek%LXZ!kWQ9N-15sR965c2Zl4`S->b24(TprHFEOp@mh@J|;7bNQsgW5OSD&e)-#
zN>i=6>A~BKOL?Bk-f4{y7B<4*(XJ#VGCz<GO|eXx%BgBB;Q-B9_o<QrNLe`n3n!s9
zQDs^D2^h5ojm<8om5Ek`cdCG~r41R_mFN^+cvmi-XZDASF}INjc~xsuE5Id3X0%Hi
zcgJvc3wOn=Y}ZrOXK3iGbRq%@1c+#z7<JIGLJGwESJ9sh>|iYH35UBLePX@4NtXb_
z%zKv0y6xHx9p6q-1$40-W=;KCE7c%PhV*mwIjd(ZJ8e&bvpnS%3Q!S8Sr@UzL{Fwl
zYZR95<&-+ZW%rR9d8e{V<^qJ&GpC?yBRm(97}%bmXn%yK5**^I*X^lenA=+2x=R_)
z7G+eR%II<;u<DpkPQ+2eQ*N8l*fEStJhn8kW`kPnx+w1R^j?rK#B}fKHc}@0f%1_b
z4QS>(fj%Tp*!J|rwRFG6KCH1Qt_Nq6OBuHHzE!{5P9ns+^E<xtd;3s>$F)#@BI>V%
z(U)YYv;cWQPWK-)5kZLluKn*TS<&kTW1gRmHiS}IIbnDnX;EfLh*=Nh^=wqc1I8$+
z&-MFhw7YEqRx(8`^&(l-fmuf$KcgPd;VQz+Q1_!1<5nN%a3&;Tv0!_Eyy_(b5pGNh
zurg5UzDUs+qJ+`I<6~S^`ny{&&p`T6Xqrlm#}|1zwT!%|NlHqkmiX3d*c3QXofLy|
z%Kj#x*D+1&)IU0c1W_J1iLC<IVjwKv{m<lWZ2|gEd|CXI=@it#it9S_A6QD@j}W(B
z$57INrs|8ctPIy~$fV#Bb4uB5|CrsmpI>|W8qCQU>L!~uEXWw~g6d=IqXfe1GhOJ*
zwNz9~V6_{r`pyz#Kdrl#bn-rCW~5AXL{@%%0?k18;#PL!{&~S#cd1w<+tZ(Luxzq&
zLVcT)f|Qri1xM4i(LmI2Su7Ew^0agmUCoIlk1ebY2Wj@Q?rzUw8)3PVH(qdD(Sl9Y
z%qX_lx^jXL4+pFj)a?x<sA18i)|1Pin-@{>V5E1vxsQ(Q%8jIWM5CiO-NE(1ZSwPh
z1<1nZ6J_CKEXXd`frfX6&;dM0{|TwAT-?w6)30F_mQ(g6<X&gq<cC5d=ZH)Sr=m_L
zUzyuc$NkamN(o{-m%KwYh$Q;J=o1}JKPCG`UY0$pm$gwWmq@|+Q|a^AFi&b-)pZh{
zq0DcKiS=v10uy<NnK`tkle`oIKca@l^TFc8Hq%y2=?<vw|IT4hkYGZ}q{WE;ihhYO
zJsyl+FB6zuoQbuzkr8B(d3x8wBIoYT;HqGZSB=MOTqDrqDC}Sm3#}BOLwTdBQCUhH
z$OCIR%y>kPw9dW2TTpzrr5JTfSXeX1V!`qhJS@bW#uO_{KbOV*Mz^E#au$o_w|--U
zMbQh{igxJJP~pR^5taV5uDw_3nPSsodo7ZNZl#Vpk8-$gehaV)5wZjM&=%4$D`V<<
zFK2Dbh{YZY4V7=JEH9X!zij$o@U)Q7>8(Mt(dF?^Wr+XSS;=H!oR9+aUF4ViQ!V5>
z=S#-vJF$~e;>$M_u{EI4`|c;`)<#?;k5HIb<EcHxuH)KQZ(lQHWvhqBGRPYA&2weF
zED^l-Ho0|6&gWw|eMY^<h%4xmQUX^0_YWMzmy2t82Wm?{vbpFS8x&M4#oQ7uy*eFV
z(e>9CFZ&Msbbk+l6V3T9AAFV++74+wL#EJ&lB~^}I(!SuX?Aw<=9}R^u^oxu?l~BH
zyK|w%)Ch83de7>|KU3S=EWrEWcfe~biV?v<Vwv$PCFD1N#VM$fv?7%qMW~s+_CqRN
zCjIzUx(coWZ1-!b=fjRH<nk>Kjj`*M+#TWZbk|;D=8Z1*w3j0?1^Q@K%}-2=<*;9J
z6nx88=Rn^7Hj78M${59BOJ@KQ7$$3Q-I^S6pTZJ<0G{*!kG2X_42oAGa=vrS6*YT#
z#bdqPc}Q5F+GQVZ^VCOGbt|U?lzeu-24xuSnv0MQa?EwPwTHAf)I^y^TJU)=?VYnK
zJO@9&x!_^XN<<)Vs9eKE$`j*m&%9cI@EVcuz9?`0P(N0l@rK)BH#JAAN0IJfpQk!q
zY=rz9NbuBy6`1$;p1ZUCI6ab6^$!x`t5E3K7{f$x&!P`b>{8Z(O!lz%=KC!DWaEyL
zPZDJDzNb@ja>B9MDK)cMO!OsoEtT{dW5&JyV_%<m?l?hCM_uKCDR{a>?|>b}t*Ijn
zJ>7jq)DbX(KG)PsL3Igh2#5Yv9)MvzV8nyHzW*zRyFr;2K;iyc7o<%syTL-2_)f&V
zeATZ<p+oJ-&eFVV8bmH`$FV@1q3-hHbRkuQZ)|_oT26Vf{6ngomB0Io&Xpx``U1S{
zgASzx^ejqQUb(G@Mh7=}O@L#Gi+YCya(RD&HbHt~T@u#NJy#Tx?F5ooTw@B$vA)qQ
z*NwKH^>UmcPS(<vw`y|Vn}uEo^wx>tCZMflV%5$I4y~?}3NHz3)~gt3+ci|w(S%7g
zEn6B+2ax-W-#D(msI6b2QIdIwb{;9$vTGD}D$1%O^geU&=$D)-&wnQef8qZ924o=)
z3yiHlDZu!AMwqC-$dBl*tn=nZ^0p5SV#LXc^%cAQycm6X4>QN}*qaM<WT*Yf0<RCK
zUlgJ(3@`Y`!br5hYFoQ#y;8ULca^8gFAC%^8+skt+D@TpS?}eI9l`JZyXUwBC+tjw
zdvpBI{a#w|&|fv8p@;gAZYQVe-zf%n+B97|+RC*TT!S=|oiu4!-|n#8TWVe!o5Asj
z?PO+5S<xzPY&H^nU{&_GKsoz>G^(OmVGzj|4h=tWm#X&Mu5q?<J_%7OSJN3_mQem#
zKNh7Bw^8ToHn#s=4C!AyI@mkvB^*XB+^9yNRV0rI(~~S6ZNp6VIzOnjSc2is)RceQ
zNn;KxZ_T_BU34;LcacFlh{waPB#fAPBT7_$=0~(nHXtDbqxyiw5d+>!P2Z+Dgl`MG
z-?n;&$0|&*(+#AeM)X|X#<ucccTU;}4$>5k@{kI$LvoU3^Lr|6X7^O#HDXbnD`O$U
zlVx23OiOFo_k8PDpoV0y6;yrf)7*>Og-LmC>jACyDsSt!b%)MB?aBy;u0D%_StP6K
zU_(zHQjeW-Y9?qYsJQGjpliu#-`by@WH%Rkk;{sVBCNHtROUp9qfCygc~H9Q!VNP)
z!D}e=Lg?t@LugE3(XIrvqBiWz0c#B<?F#7;IO59;>frhd>I<<-9q)8is0@i-74Sqp
z)1=)xM#}TEu5GtXN-ZUrrP^D4Uh|VmPbQ<|I#+B7tUn=c*{3)M@)*^(G$3IklO`yk
zJ2?IcqoO282Jl3>c+m9rx8`yX;34K5P*j}e0;X;Ng9kFROXH3LL1GaeZuyH~GCnR{
zDAWx^g)bS~zveb6tzEZj`IW+^<rCNTu#e5%&mm?ibJlHQ*dtc~D4uQ~-v&!u)#Q&h
z^Co}$aAnHShTGKk8s5~`WNnaicoT_YEbh#fE5=*-`9^}{8$xQcAa7@duCK317R?v%
zX2!I@$}5vJ;#RSvD+Y{VysY9BMS7Xasai;767!=F!{Gp^%AHC{WbaPx2HO}Mm3}vH
zH;_I;maKnLSL1kNFU^^NR&gCVkXX9Ly@egCYS~d!`3>BWcJB+#ZMwu0Wy90p$Iqru
z3VoijxPQul*T)t5)OQX8A`)0x%Uc`o)+76<CG>JOk{DdSy0fS3YblLwBmcSNW$FfK
z`7PZ;?erQStU+9c+LZ&gy^R|y3qto|bBt^jQ9uG>BF0t6I*QEZ1GRyQQOG+zR_sUx
zBreo2t<u8%R=?hSJ-05hu1aU;9$cYuw!-U8yOa0LioNRghwdI^j=k*8fdeZ#3m@A#
zym?j~3&=%iZuab_YzUfc(DYG=paT)>J3#*XaVs6pUv2;o5PR3`Qgi+K`i>?oK(8Q%
zekbj_f7Cl%82x0Y7!*27Gie@02yVH7JbIo`PanuG6w*HiKMhbF+4&Se=FmK8GAg`C
z7Yk?6BVQ&DjOhY0{Ods<n`AqOoGb;4^UjSrLJKMk!q%5vD^#i6%`IFVmNvu%x6E0J
zs^kO_(F%r1j|e!ip_X+B@os<F^dKI{#6H=l#uS*Gmresc6j>uCV;}oED^BjMU-zCe
z;iKh^wVqe;Jmj$~zHPf>-e(#meO}c98cWid&PwO_zV1&fSS4%{@q2kx1-+iYvv9&Y
zb1N`9LP*SjEYDo+OQTDxETblg46$46;U7W4)RZkdL}Uu_iy|*$gs5xDX=G&XqUrki
z-FLQ2d)S1sN9$U-y!(s;o;XMqSn)+XC|n|v2q88B>In{Da%uSWCT0gFIw4!{ozxWH
zeoh}l0r+eaJrsSQ;)bx1?fnYRyWIDOVQFLQ!BS0Xah-&7?6w<EZPC|*e2jMy<WL%l
zk*aqufBFoS+aHw02Hp-0LSgNb^}}q&${i0rVKE<k;%*z{%M(g%5Bolz!7WV`SG*Mm
zS4;t%5O4Qp*AV;3r!sI}>{*V32%q}!<H}RctK5$;?qOC_?9ukZdIxSpw8VmJ5wt;$
zM9QLfpTy$0JYkkZ$WBiI80f99-n#w<f$Dl>fuBz<e&G7PhO3^hQX*Frs4^9H_Qtqu
zqG=u~8^xVK5F0qSIO{@?c)l)CWp48U3Ur_EU%|*r&_{sFlrY#|<0zD{#f{5gY|=^c
zJGDcBiGoI>Be3%1cziqQdEJRDT(u!R(N8~xSP^$~R_Gx2X}G1U_i;0tQEd|GIapui
z#**(Ka2;88m6huc3Z1{9G*c0yR^EIS$RIs~_1Lg#AEqX;YwloLp?Wjz5A{one%YQ<
z(8R84#?WnN$UXH_Jn<sA?$HJ&AQ|)dMArOUoW*+f3DiB;2l~BJ{i3axeyUPhr*1Vp
zNRB@p4lE@WNfbNJ4nkU~XRg;AhHR@s_C{ASDLdM{rn6vaoT=taW$ipAjG3OyQ=Zf%
zA;)}UZbkhy@77M*>D?7&-i!)n?WQ#z9gMu169<hReC@oBD-gban%yW+{oAPsiF%X#
z#KU^>nbX2UMp0_&-_?v?1wIiY+O)MYt@amWIX;44+PmMnsIy&zA+=6k^mv)CSL+eh
zIB3iD{HoVcz_CgnF-|<Ehc}2o|5oCA1o=`l<#w``H8!!Cn@%FXK>Wy$#J0K_J6h|`
zBfrvD&VnKD0nbuV1t|V$s+{5f9@FnMn>-fhf$j#ArB(rA-m;eCrk@N|8sm=Etf<u{
z_b|darZuec_b?B!;<r--`IY|Ud;C>PyoM=<zW?R>|6-v%U?9KhUdUS<@(-+xtD2@_
z1mK4;mug$!{|2Q$h$Y<t088O`{}I>K+x(ja_OC+`agazxey45btG@~7?}xw;K1MGg
z)rax_D5L*=+@q@mK2=?e#Q)&CrFq?^P>NO1d!_t)75!rzDl|aZBydxW_|H=Gud=Nl
zit>%r|JOn=*NP~x%m~R{n{2Swf^yl}P7KEc0`vBs{>zc7TYEHypA6j>BBibd*(nMP
zKOO9+63a5-`+yT53UE=a8HVnu$n$>mUD~Ie5Sfu+2hIm$ED%-!IvalHT6?(8Y6Jo1
zgKQ_v;E0&nujROyF)!Y%&BhFdm1F)7shY?FQnjuN4gBS5{2?=uG!8(jY7)YxV}Xgp
zW#GoA+GTWQ`*~^!-{5uxAp~?jrZu1|;=>m{OyWAuP!q#oy#z%df+fcFvuolyMfLPf
z^$<n!G8+Q#xTNsCF4NS}(DOD+NQ0SSaKx`WDZ;M-^KzFW#wdD~HZEb*C;x!&XuwT5
z9b9kBrEP1(#lL0AHRGwd?-wvXw*I75tZS9rt~e3GQ<ezdG~}AvG3MSX;2WI3X0XtK
z=WwMf6o>p$f(vm*D!MjYr<-Pj^l+IlmQ6Q&Zl9B(8k}pl@%=<5*%$L}s6EpUAF-||
z7}*>jv%I-wGuUyNSQ~55@veARIj;B7l^Wfk40j3EsqD2pob9(fv9{qlzp3$!KwEqN
zZBXxMAp{y%Yy=%Y>XM}ev^HBG(BPQKniBAv&hOs&>lJanZZkeTwWIB1+n<|wN2pek
zO;jDtG#69(Sx@RBOdW+G^3-%v5=Ez<RRxc$Rlg-zMEnTQ{szm+IMl}>5q8gc2M8@B
zMx=I^xt+$NO;rt97d8<<KBB<z>t)WlC~jg=3=Cimg-{MAj?b}w&9|H+QR6yqBH&%6
z2uk(!)7-TCoI&6scbi1m;uF>tIxtZ9IT;y4HXl1&p66Sx=cyg6SqBNchKuiz#!bGU
z-Ha#X_PNF$4)_|1iGan+mg%9C^X`-OdK{kps2G%Bn_IO~PTBBTtkz59vLQaDA~qal
zUWOdE+>!1H5jw3GKO{1*l+`OU{l?<ZQZxp^yPUh_4uf;zVNzoFwL>V{j%XPo=2GXe
zZl3zc$=(XwWhNype3AARJA+u1d;AMk7TnVI%zUnoDuBGwmT%$oQX0pforaO%s<{Jw
zxMi^Q*eGwQM%cC~BvHdS3Nl)uvKvKYWDH`!0#Jb$Fp*wtJq!BKM((zB>b}xd6ZQCp
z-6C?PJZD;?^b9ULGTb2R#)OG_4L6R{Gsm*o2H$GX>2*DV6u2R4R`ITHX9D%rtcQwp
z@5-w4J8RKD<+M{1E+}4jHGXt3_J#<sw}-bgNJTLZVDhl@*UTHFr*#Yo?@94m+7;C3
z46g7DM=ouaorC);o5zxQ-thr80k73NO*m`h7{f2&qLi=!gk*{b&@$6y+hs~~vGd19
zYv*`+2Uoz|vekE4V-;#aZB-V!bxW+;iL9DR{Jn5SpFFh->lx3fO7?#L9m{~@vf^ED
zzS%3N{&=MvET4W2&|pK<>$XVRPb>~nQAB_7s(4(!C&tbYt$RaDD7L`@Jvc`3>gC>3
zhj$U>th6WyX`;QJ@7yUH$%`7J|Bu0mAaQK9Kig_87K>U|UG6{o)-G)5GNAaZW&^AZ
zJv|A0WJ+SEb2YHW!dnpjh6)?PyT?DCTc43DIbZ?y_hl$l&*1nSX;@1;t~U8RfGjuY
z!v~2_Yi4SDL3&!oP2Q{y%a3z5@PBEhbC=FlGDYzL=Hrlvsk3n^*BRT0_-qYi6?|g*
zoP}dE0HU$*=nnt#`w(*Q$LqiwmOS_pRa&yYFF~TkS*e{=c8`ZMi7RguY&f`v{8roE
z5|z*@+lD?|s<oJnxHkqlJ#5}Z7w;}$^IpMx_&br;XEF}FHj3lB4Ay7m?;Rihst3c2
zI)iZwOjr9^P{;-rhjVA!nTqIxyZm(TLg1zPY*!1dIRz%n$hcSo7PwJlhgfQf)Es9A
zTb~)5fPko3<*F!;vo*V_`gv^2tIi{2qHd?n#iXvE%Ky-*AY`*K$ulhUv}_`gi{QMY
z%(R5@M}+sGCZW)QbKb-bM%hH!SQ~k1kvM&W_2F`pQHv?dg)TZ@8`pNDpfqp?v`Y@G
zs($!@sq3*xTTcfn@DC9Y#=pfkIg@!2qko%|4AY_`99O6~vEi_!^UE~?-UFVJ5DX*S
zS@t3s?dpMRhIcIl(Acc+0gna%v@<j(dfMDUMzf+j9k;`DzAagm%G=eYnuwZYLsztp
za^N-KP|Uo}Cn?WZ%`4K}U)EkQkqCW0O+}wlyj|~X6%<_8k={XSHy8wVHi*b8e7Z4f
zGvf?d8fA;~2}H{a+be&vyniv<CR|BoG9rpjQuQvf4+tSIdUWf`Le@c2{61~zk_v(|
z>At6RvP6P=*2p|dl<=6%&qpo-I3!KB&e(uT$OMbjveTRlb`(|)HVIjHxCh<k@1^3=
zz5FUacXn@a@F(CasSp8pmf^A=3ibUg4n7jCABm6!b5A+PIj<O8+5#X7WdiyH-ul7S
zW{^bbg%vN3sgwP+C~1{>e3%GiY+H@XZW}if470H=`$_U_t39Q9H^i_&mt<;?zw$vT
z!iJ|uKThYo`NEl{sgUHHLh(XTYvwcl?#6JLt;|bcP9!AK`l=P@jvmD-1nx4Ja5K)2
z<(PHftp<#flg5t^4rven(5j#b3<8FCn@d8%2=ygh+;pz(iF&O|h#<Y`_p#{t9YnN-
zQfrZNfAr(m*8pvF#(ZYlTTGU8cvtt^U5VrY_sBGr+vQ<4)}Pgxg3gspXSk)ED{QM>
zFy?65mie5GlHP~FS#Vo`(X>maH3EPT^LIcf#v&5M>4}>LFcd`>ZW|md*30J#MR%9S
z{1y??;{h(vsppjoc&ICN{a-%Ye|iIov%rGTQ+Jsy_>T@A23B_}2C|{2@s1Ub$cCH#
z)l7FsU@WNf9uhl$5LtVmf0r?(F^vuo#M(psV3r1<PZGefE+#F#QelYAG6+`S|7yuK
zv`~A+trKAPnB5^{D`(TK)tDL&MTX2++9V>!4F-Gw^UuHi)F(#O^_AxB=qG^B6S)Cm
z0#GOV!L{~I(E8n6%xkwZXcd*HOq`u7?vV>kUlAClC<C5Emd^LYvX<$c%@YC?t&$W(
zQ&q#^WX~1Zf!-EigO<4fx$zU(T{H}jzWvYvW^?iOe|VZ0n6#p4#z1{95JdQ_Br%bA
z3Lpw?7Jt28W;;_mks%$mIUfYB1mcVVmkATXAc@bTG?k7A@QeK{t%(Zr!SqpH`oDX=
zSE%&c#RcTCiya*-<=)Mh^&Gbwcxok-383bc0URY}GP@`UH^4&ScL@%V&e#NY$ebl#
zzilill3o={@r42?(NAk1(@nbVEkG^ny%l;W6$`P5ssOHAHqY<pr@h9Kk;vGbt|_xJ
zp|g7B@x*bZuOfiVqmxb{brTU_SJFms5ed+9y8C!no<yOI=Ni3J;!KbTTI95wATNUr
zW=|a0RrYqVoQMTlnRirvW5;qW@NjD?&gdY=5&NM=ty2z{)o8ExT#RBxw#<zO8Vw#5
z7K~)=IXJ7mAnUiIu~&Sxq;r+GBcEl&>CzG$g9yRgm%$o~>xu%uM#%ZVuT*U-mPOML
z#H60DVX(K-&vVXRaI4~q$@SsxV?$JaGLKzq$m3`2vmP6)b(^I<(<-mA<712?#^#C<
zm#*dFu2UnUXG?xEmRf^Tr2v5v&@G*s(pBeAm@(oazFr$`4JM=cE6atB>irklrF=Dh
zTGY)X{vwpOSNV&FJ96lFLMw7fY)ca#W=%~#lV`@diDG$hZASYfNg&w&rC;vLN?Iuq
zbU?k$F+U@^Z_qJ4mGB@rd!<dLbopD-vxS04zUWF33W>CQ@ot;SiFVtD1GUq`DZ8!j
z{<xJ_#KRBwNKE-Ri~T*1$;gC#>Sw(6z9EDwmmoQ4nNqqEuzf;fCv$;K+crJ>2)XoY
zRu_e6wc}c_13N2un!r(S;pg#9&v>uC(z-mNtkMOw!W;)u-*qe3*Y0>=P&CTy8R#^$
zFpw*7XZ^&3TX@W|;qomy4mA|Uvd(`O6NYuxpw_*hV3;C%P%ph%hIpShy;9JU>ObL_
zuCl|oe#37QJGw$@EotVH=q(=I#U60e{)>uPo^k!_u7zM;SahP6(kp_g5*^=J3(~$o
zEA0sWz_nTN?Ke1Vv%>Ma2_djMbLeL*rFB0Nwo^!XltvW3D4^Mv`b{~x{UwA+Ea)J6
zGN-477i>%$L?>j{Be{<Nu&5=xqwWqHLzS~G?Pk3$=(p<|9DkPHF6}*QD6*HtTXSWA
zozJeFAL)P)!Ycb6#d3z}!lTCk(_?Eu6t_aN%yexS0ZB3i(nfIU=;63H0a~sD&y^Hc
zm$9pCl3o4o2W`o<AaG+b|4u#9<8ad=pLzA?&#T4sS3%*w%ups!q{~eyl!(1$)_sNa
z)(RrYmZO12%rd~%9mb+jOwmXO6SxUbJcXx#G!<z;j&8H-jPndJi546M@DgEqNaC2!
zl3}56$A5hKRiDxK{URSS0vpMaBD=pg`$NEXT8^mqgwSH~Ac6#!o12prc}dF)MPW~P
zhiE*yk)^tIE}J+j%~wm^l?U>7sJKE6KM;)~2bGBix0YKSV9ks-q##?{gP$eGt^Z=Z
z{m}wgm7e-VYDSg%O+oDc;p(qmL|_j@GFT!)(}@51j!aSg<=zBtlWu(VPXPRBH+s24
znX}H-Q{9&n@=?Pdkj!fNolG`4bWBrgvza*M3<v8~NvhwHw4>7*uflr~0GXN`fHLK6
zPAFzXSwCmP(_gPAB~O7GcMGCv5;K2{_k1J2Y!l$pmS6F7T?Ss<im{jNy~^LxW0HCO
zY<2-qs=Cq($X;KuTPOW-h0JH_T)~#77svg+lgvf+hIKoQbpXO!9iGlz&Zc^-=S0im
zekj-kD9hHLB|=I8wpE(^kipjvH4SR(Iv;PG9d!s-vR00_jBIt$!{rDvS-t}6UFlRd
z=RasmNXVC=OxYBXiAHZv`q-E{shkU|+4>r*o{|nSD`49pw96%j+rDurRy$ub;Ql?O
zj;oxHT7lM5ckTX=Yyq%G&}BU*t9Ea+K)cQrRuf|kM2^Er<v0FeiUVn%UR8b8yFS}3
zc19c?jOk5<`Rr4#gt!63@QDWWtko5zax0E7t?x?bnbix+t%@2(p+3hZQD2-_KLt!i
zNRqA1LxHu(cHu``5aASwUA=+i6Z>~Pu4hNPYPz~x-BCo`hOM`ebnS_UUV!yXUgd|y
z%ApM&2?#P_fRtP~4eX5TWotTfNMbko(Bpda^9Q%n>OizN3JJd$p7lAksRMY!n^!bI
zJcZO4bshWncxJTb-5Wp4k@979ki>C31%V$!wg9*Ja385KWtwDT3KYhwjmm&rckPDN
zSYZWl!=P5Vxxp3rEveMIlQN+7hVc7LORHjzt978nV=+R~x-Xg61>hjw0@p3vB6Ir^
z^BH|rUW0Uy3AhzYznL-r9Ku2zQp?AHTN3j#zMFf(IO-_#ubrEa{Dv-n)ZFDomC@&?
z_S<t!O*wNTaX@-8D<@?g-7kDOee*7nKB{GeZa7ymm+#~(Xg5?!^X_^y_MfIFB8^AH
zwY#3bj$ZFow{h4!2x^qnM9crkVz!%dCgK0t(Iwu1jYk(RR^)Sj!n^yE{PHM(6v+Hn
zE`;(Yij@;W-6(}_=%jxVs{@GM^u#_#KLX^9qXw>u^L$|<uo%FUxnOV<ezZOVdjSA)
zwkH(`?f!&kJaq4LZLfHrTfnp*jvIy5<cv#_I!E7o@_gU|AQJKWDHhYHr8*5sSB|J@
zPXNYa4TR#Q03g(r&(u20b}Jw7F9biER#{&N6P~5%ys-s1D%=J4+<&|$oaCAFT4<GX
ze|E91-sE7o4!AMN6t!bePtCDH-IUuo6=ZXsJA!rJ-`+|7m{Z#+yBDo2Z|__WFm;m0
z83z>rqQIW9&An04^!YeFpEayTKu_oT1kpOKn&}1W)c`f7qv0j3m&^l|3np!7gu0%0
zMb71R-Hs4FD5&B@VcwFP&QHsGWedRS*O*J#^vU1Xvr#lML2Cd8uWw&FC=)Y>Xey{m
z!u8l^A2J5LNa6U0OeKQ_CbB8n*4LZo^x%Tb%Q3~wU|d3I_mWIDaWSu`u{2ERq!XqG
zIs{aRoSs`cBO1M!_QF4TB@OY0d3%2isjCCCilSF=F^qxd)&m8%iM@W|omP%V2u`g9
z4s#IKobRP~)OUK^_A7pQ59&KG6*I}-v)%wo%*Iv@_?efbw6d}Bqsr0J8B_ZH_cq%I
z^~!-c?|V;ss$iHU_7ajc$`>;WQ0YytD5_NHlgv>82r+J0GzTmgrG<|-0W4DsvoUd4
z6Dh8<6J)+;_H~PqvdJ2`L8zxAwW+yqfoIAUBPP*{!n?CqoZc9~lm{Ea0EcMgh4b1}
zmEH8ltNEOIv9Wy31uUp8|6Z@Nm8BJRs<n7G&4|OxLZH7}Tj7+V$sY?G-PJr4CY+qD
zC(GTG40dG*DDVCWKEQ%S;m67|KGZkoYMh<e6~tkF-2?Y_s;N`GAp}L5)DV9V*L2pl
z^x0wrwHZ~F-pY;bkHS|uD(e^8Xyte3T(8{(_|ZvsT<J?L-ztYpm0K+M&(tA4=ZkLq
z-){kM^=tq!Zk5aYCxn{SWVxKl#ip6CV+ZUq$2vjaR)h3@q@otQYfn+Rn@e{psXOut
zX@ojSn*bQyb;SMvO$)wt^ZwbsE5!bAwA7^MsBe62a(yAoBQ7glKo(eNN#^7v(+d&W
zR)Muu_jslUfNcPa2;H5g%L}3Ev>l%r8NH8vTwB_)g`sUag!mCdXo9~OT>okI-#sx`
za4(uagN+y_GP<5RLLO}+v%<9fm88oFG7(2iI>^kb2~!fD%;O^8^E^E2hLv<e>WX6-
z#Z6h&e8fY54drC8Wyy+kva*}hob*&Zd3ssMCp|Fi5nv#QTAZl?R*J;Y(S3kaPw%?=
zULz%YQ|#492kt3UPaC;J|3VE}uOM`6%%eKn5mS{r`hA#lLo6yjU&!iq6rlx^ta^uv
z1u>OZ3fNjSTI}&ZFWR55%@eEY=gJL25jdxDtxpTARvWM?+yl8y2BWPYMzPZ2>l0;J
zA;etW?n4$cS0UIKPzKx1I2iB%)bsTuL`S)zfd|oo9yh`TU~^>?aBN+XJr(RhJlG)|
z?hh?<E1@z)e5QsVZQ+Zf1veE~wARL}4Xf@uipYL0h;GuC#jopDR=FH%`5Hg!g=mSb
zlqqetKgW$`4QoZ^79mnLbdomBb>$PdD_@;+u|j{W=NW3$@)3L%YLpXT6vY0O@A_{i
zNCX=+pkfF48bc2qpi#1{EfJHn@%rpo(3b7|z<>4wD)8(p`Nema{lhh2tA)Pr>Zt8L
zJ&3D<F%<E1BU3k%OK{S2Svne78;<%u%w?1P`J0t7-xz`u^>o<dBZs@1B|S=Id2hto
z-9MMqM_2m2uJed$pR!6_M9-ql@zmVKC@Hf|ZO$y4QsiZ{gvExHRk3Ze(TIZb-~kBN
z4(Ks$O4{p3%X`~f4&;Rg+$QVf1#4@rH^<r~CbWu$gu;vsnyoHsUHAc9BX1(wi2#1>
zz&Ab`j<|VMrJD{!iv3V3xh>~CV|ky%He%*fbFpN%EmvGTHc#A|%gsWhog!QTQ>$OR
ziK)_VP6$2`YXarc9rHF&2RyCV1F%cx!A7OeHQT{-@shlk$B!gLpafdNxzQattMMC<
zqRt7kTJ`tNFT{By$%FuEvp@4qKQT;{z>tS+3css-eXF|^=AeP3hEumno`=A4!H8W|
zcn%B-!&<zSUGb3C4d7E~`<iw0E2aj&K$@u-#u3)j1U3TxR-flnA_WU|z}Z1$o>n(+
z7=%@L)}(%;eYZTNRKXSJu-}_?FIYvdi4B2@8ruN&b4(|uwT%)hjj+eLFc*kzUgbSf
z?xjY5iio&!c}09~%cbnqR=`-HI`Z7^qgL)kwgk-Qiw*aUZ~x6k{n|xD(&Jse_O(t{
z+E{tePEOf1*{r~Q8<SqL4Js=)FKxw}qm+8HEcOy!YUWlhE2a8;lx^2~Ym@#nNtU-J
z1ESiYgX*Par7~gLm(u3Mgz+OD<x(-L6_)3YV+;YKOb%jrR~_Y;Ib=Mb7>rb*+|TU7
z=3jf?_XbOx@^xy3<5Fm%TZKu*bN5%4kWYU%LGxrR&KWH$L(AN<cb!+bUJGo>du}kW
zUJsPSdYSv?jp+;!zWL@2dfulmG*vO}>6xu*-ipX$og9o8%rX-2K2m(sdfj8i%x_&w
z)$=ETE8qIDbC5K(%4qsm?7#b_GJ$VmH^zN0&t-GvRS>@lHzLC8%@;eBw%z7N4@qtj
zykwAe0GoT?tT8kzdthhia^s;Z=uX$d@fT=hV^2>vjNMA<;<eq#IA4SkC2{}0$MPb!
zn*8~(sMYhge@jyQiae}=x{#@8LF-U*lbv;3KA|lKwrhW~mzG+E=v}GMkPyd6M@oFm
z_Y}p8lj7W*%l8V&weNpyu(JM|SNb5M3-3SI8XzSE1RM=Q*t*oPC@z;M*?@5(Vsh-O
z0Eg<QQ?I|0m9j^ejOx8lrSJGY=KdE5Fg)<f#axoD{4I08{M%B|PGs{X!}9sTu-HF-
z{zZiw$kTRj>uKErlqZb^TPl0FHkt!FW~+b8c&U29OS2_w=<l-!gj|6lK#lD7z#i1N
zX(r!tRwqb+TWCi!it;J|dx<x2l&PL>ag0--yvEZ##21r@6M7Z4{=}j(gV$`^rH%9d
z=2SUjuRb0;PkH2N+DZ`~{N?S9JDUKL2c|dFcz1HKtYiY%AY{`W&A4m2k|avC;`ZY4
zli}gv_;n=F-H1~5eGZ$e9e0ZDSmpAx$flm^>p6V?o_Z1J?PS+kckb8UMUprzlnTz|
z`POI%FRoT1H8(nXa^JI(D_j>Z0QQfbA5Qf{y>{E=B0mH!{6B=f1z4187d9#*D2ON}
zY{eK+x&#RUfl*QE4gm>8x<Npi0UJTOW039!X$%kt6r|fAloljJ>a3U4??3-J`?|J!
zU*pUhPpr7tz3yk+z2CmGw$%-?sO$l|?;Z9x9$z=bd3&t=igSR^%3BV+I$bZAC;8?f
znceZWk|#Mq@}k{3E|YwnTma9>$Y|PpH;-1f(U$T?l3SIN1l3N5=*CoZKDW^C=Brq1
z!l{a1a&*wEDSdWk++%%N_vJyQw_ukZ^gEAdbyT`6QghpqT0;ILGT)}`Fp>AJ2R;we
zG9;4GHa_%3wHg{Z0WeVI{M{nWRRow|S6WK!*hlC<vaskW%+s<gtrfca!#}+6+eZo(
zwrpt=a%5)kXVLMtdO2$&$!_mD(t`EU?AiD@j{mJBgLzMW&n5F}^}RRr`!D)0vLbzl
z$#OBK5StamIP|=qs$A6Rh5kD;C$Vb@iY}05PJD&J)DbqdF38<u&b|C}jV>a)->yky
zq5jPC@QeEx9y?{Er{_cR*}_@z^D@(8mH}we)xJpb46+rRFW?IF0gpvy-co%`P=4J#
z^gbvHG9S4cy>5&BZVvVX9?%`RVYW*89ip`Fe!6O(-B2+5lr1|KV)Is&0fPO<^^gue
zK9iQWAt{Am;3`F+d9t|MFsd6K%<{~2ze2cFr*<drJd#eG<w>84q~A=XsylT(`irI^
zC$=+~2B?xz@toocsqKx5dwDU?)@47&TE)sl_wh-DMgDun8QfOVlm*>dGr36EjC8wo
z8XY(e1poQz&!2}@)Ba3C(^z&4WC&SqYfInx^=kGAL&}~B$xC*%ayz#UhU5ek<Q0q*
zHY%y)67n@U%7VPbqQgTGIYO_LlwVb*82o9w`8QzzGeY=+JCD1t_N`<qo8hPH3p~9m
z&CMF(LRJPH`f-V)LiS^KAfL8Z4q{-gn#Hp@tgkM1QnCzZF$c(94tbj$WmVNZl;t?z
zuHFmnVG|I(6kIth(suUP8gwFCzlh_Kq3?tsu`f;hgWX5LcTQ>_k)#T(vyoI$U4`~1
zGo}lNj`F-E@xZGqmon%i^E~k|ine*U3fWc3zey8Iz#QDp=5{jSTiw+$_2gN`+T_xv
zaPBwY0=bx;O5JbMd=A9GSZ(W9MA@RI85O+pg2U8W0E%V=y_K^uZch(coeirl%qkh^
zA<2Qp&~T|AGQ3Ukj?)7U5MGrvXwoU!_dy?8%;QMX?{8;YVx(o(pc^;lOX@vc*O>=Y
z%<>n$64EBSyFt(Q3BUsS+cpyP3pOuK*u16qXvf!H>7U(rTl-2PJB<^5W0<j-B*rmP
z7BAk+O=nUB{zV7DiK2YO_@n0EB=gK)&9Pf8Z)zMUOOvy%Bv@W$v7nX>3l)fYE1Grd
zC2Ylxr(B2eRc_05l{06@B%f&&82XApGCI{0VU*o_SMGA$MJlMkRUQ^wX&!)tiD@6c
z+-6~-dnOgSIS=7g9w7-4MB#TRm6RjKGDe3Yt1|6bD<%I6aZDBe1T`_HW-0$*a^?Uf
z#!o61L0v2ep4V4C8|e5Q&>wT88(Q1;J%gM~`&Fk%{CQ>j5ZdI~=B1UTU(gBeAV+N<
zk@#Ew^+<O4)L8Za4BwPnV>oyAp`~xpFMbrhqO@(->3(T5D06#;``z)ukHZ>Eb#zu2
z0(i7fMDcToNaoI&4*Ab+0fcA~f@zT_3K6kX>VIXM`DdGL&KJZ*pX9Wu32N!7O`b9Q
zDrsV-p3I?qIVxGzOkfPs#%j5nPPC@Sa+Bp;f*vyyD4z6!w&Lo;a?9B)b+h$B-{bFJ
zl`WSm`c<;{v~6zokIc!~NdD~6o`E42=3eLtck>r^A59AT_MP%zCFbbigU5fA$td?!
z7MXXq$U1eFR*2-3jvO2B$cnP4S{jk-l4(B)RIvtSMbvaxo_;^C>4Wb+1zcx(9kaBK
z^2TJu8H|qIcslMyue8r4P=CWRGqa&Gl{M7*uNv&U8AgNR@x(pN-+x$T2|5$!j<jSq
zLH{NQ)7ZW^m&(!-Ntw;k{)F-QWm(#6K<qS8Ny(JU_KbW-gij?G@s>dQdiiMbsS26O
z#bFSB`gRgG+tZfJy~Ja2;cnNhM}k-tCsToG83*z8iGrp+UMucSq0oWZSKZ%38p<mq
zlr_qPlBJXzf7*-Y*JctJY_U;n(O){#j{jcWCuFG~)1_klPl|f=G&vP3X`YE(E{BNA
z!rj&k?^|m(%w$CYud);>>`+7$2z*b9^k{cHWCMJsg6cbUHrd@kz*9cl%q1)XwSY1{
zE(Y{SLm_{X1LGoNfRdR=`oJuzjoPkV8M2jzv-j`?*6#xu77#ruL@Ua<AaoCVPLrtH
zpO?Pgg{9ARzC|AXIIo6&M?5+fuHQi<>BnSnDVN_mSy{Iv^36l7JB#i#PUfzAY2`Mc
zR@*ZQtwrSxIMac?Xr9|K_~@Sj|Dg;d(Ly86t!%an5mTz6zPOFx<A4NOXoA>BF2Ypx
z`Mdk6?&U{v&*ob)qed%6Tu&6fnScF&wEtt5s7z~-!ECN8^f3<#tC@vOh$|nRtq~qe
zP3W65l(p&bYmF3|Z0K~c9xFO5-^h0F<g)^h+PaV<UY(uz4M}BSP@&_ug(0Y97-?g6
zjJ^>|+W1AR{Zm-Rhr7HFv=8#yYb$?oqjQbc3q7E`%6mkcYtnLR^40!RcZHvPf8G6E
z^|MQQS&~%BZ&BGYm(!|$Za=)9uAaGaE8jgb64B34_S?*1K#u5H1r0?9F6q^O-WX95
z<4G!xL;YYwTn)8Cora|o=X@9G#(vx!2%+R+>z->Ub{|F376yT!;v-m3hqrB(fk_wT
zPAk!Tt*ciKn7tiv-RC4FHzHZ~D6pe-ysy;m8^gyd5KbQ?$$b4{JYo>g9$Qp`ZjPdC
zTcj_@WfVZv(6*eJQO!3j$jtTbA&-S=f{`^z!V*7sfTJ-WT>YP$m+Hk%MgLNKe%C26
zUB8UC^k!@FbrJr_qe7&~E|0EVLv8LG1C(aUY{c|~_wUc^7A<^t$XC~&{wK&hMn$kn
zo+#_!g-Yt-;u3!17ja?G_)7FcYGze2rwcTfaVi71TcAu(K@u$5bW@Nm?y_dOOT52<
zVUPg50e{<;ni_+wr#-85nM>;N-7s$>?iPY7>^S=LW-);EfoHOFT=ld>XS8C|bADN(
zcmG@1z|A47IGs3!4Ppy<{TTx2aySrj#Y&9%d1=s*e=klzJmd|;6jA2D{TqJ~%uCRb
z#y&a~Z~H$H@xg$=sLdw;GF`P?(qj5W^oZj7x#?@4=6NaqiS3#90xQa(SyBaP77$C%
zUPs&f5x8uy))h$g@6EC1B;v4mo*4$-`^Z#~B_AgmA<l4!ko95B>~KR?dOBXS*j$6#
z0<=>v=|N4j*u0y^3`&$HDybA5m%k<Qw+4_+h}0CmmlNy(XdZ!0LCMd@HdEA|^A;*Y
zJ*RkV%(O0zaw@?#yC;xAI#V`^zXL!{Tg!d+T;y1!i~+HbJxDfS82mXl9*sJ%1~}xI
z?~+BX&|;YkjPGp9yI`fzH9{dF%|3-ttN*JalS+;7zgmI{_ilWAnMyAP-&`;8KeZqn
z8YYm`D@10)?&6|}bL9*K=V|Q!4ugOo6ae$Uu)u1uk#|9#5#<AXd|rPw_ik9dnfn_5
z2R9{q6QQ&*@wkJ?<&NH^RNd0&*h-aH12NrCg3UTI->04_w3|XCD?rJi)Wq?^mPPd^
zfthZ+PigZ%>t`qoqwyrzZlo~}sg)yD@<QsV>UDP{m`951eivsY=y>itt^}bPlUD1^
zIYyJ8^}n#{()KpPJ8ZjeYA0vIUZiBE3sCYJ{B-?PjAAQ5g8E_F4iO@N`!9faLO=h&
zJ$<_KhBDVoo*-PJjZBL2fmBD@+?{)mkR7m_^ZV<f49+MG+-;RjR%dj8N{7(%1}La#
zJvwwE6GDM0k)T3@qef%nYG8G*K02Wo+K!}lm_bI|jC76?2w1SU$HwcBpY!o6b?^U-
z$#dKsj4{gQ^DmZ#*%lffBZ(ZaF=VAHy{g=ygcE-?LD3`f(=t@%4Y?}k0!<=+^L5^&
zl2^7dQ*o~zN&YRa;c<STY3YKsjQ&lwso3}XkI&XW*|Vso?V7YA*jxF1s`&SqmH8ZL
zXV;Ie!s)%Yg*~1hca8P!`|fB@*HUx3Pnzst1;vpR7il$o+=s4HkFx!y>}oz^{H~U+
zyz5fs6j6<K4z2yIe4O{IG@0M<5q>!})ezq%)!*)ChzO6dVDl+h3P$5#>34st^^uMG
z!LE^iHuk}5*7uPh`(vR`){=Z#nk^G4Ft6`dvbVS2z=fcqt)Aq2z5A~6yJ@PBy*@#X
zqMrbxyCon~_3257&`ItubLNpTCA?+Np@3xoB!DKZ*gJ#-*U`!AweD-|qc-UV6X&~l
z)HK{^)PD*nX!O54R_^5y7p6L#X6{Y?xFdMscPpBFRzac{YP%=@gw@3!?NsTy&|PZx
z{USL%DP-Zn#43dnOXZ)lepw$CcF5EQSGxd}3894}0^vngRz1}f@Jebn7HqRAR}ZiL
zd5~=&ErPm5UVn;U^<F+=2Tm=o@H@2rW?&1O8PAW?z6}+zd@zU(w}%UzS|>(>m@>`8
zDb_l#JgQrV$dfw$?$gXWh)v*Ml~BhIs_-so1al>c24y+j;K=AzL?v|uShXZdYP3h;
z`418}S&zSKuwClK-AcWqF2%bb;8iCn5PDwDaDRXbEGKUYW9qd}ki<IZFj5k27BZ0v
z=ZzE|b+5$#S+WXInC{2HfB1A;yYD!6mX<)LbAzAI7_)~d<Y$#&l0L|;E<uINqVJ_m
zM}d*NdzvrRw@piH%Ww|2xJS!&l3v*3a{wn>ma-sgu@1wQCd{rK`brY~;qLS+@BQUP
z4DD}WrFKLLSe=ZkT7_QyZosj=Wp&Oxra1HMJJb&Hv~6mS%-9{;+Eq*R=p3ntWRG!%
zU$>_2um8Q*kZp(4LH3(mBrl7BKIh9XsMdyFknkC=;P~DYXJye_Y`{Tv_@U#bn<OUS
zl5&opZI8&Iv5&!f@dj?%yUkZYE==u@=>}4I112^U>bBQ$2TA>YZ&*|RBZw(FK=<T-
zZhrxV;G;#|>XrQ8gStCX<l~$NDF{#@;-3ALY-s}085|Q++M25Y(>=x(_}^{QU6COJ
ztvsKHnZ+CyISEx~sj;?Q{ZX;3x+wTwm0Kq{pONGAL!8V8Tz=rXUfr{!9r~@Y@uaCY
z<nhgCx9xU46~U80_bVp^h@#2tXT?L!|9LE|6a#R^(?q1pH$KLt{$CBXgd8-E3esNQ
z3bzFj?+{H9_*b_97`J8#@Yei+e#sFVL!odS1?84phEC3>^6R4DVbl27{tplE8@q>o
z_%Qo6LPz}!^KH)RKHe*yaWm52UKzN9BoGPQN>Mb}oFRGClgkX_DxpSVN@o9f74$lI
zs9#ON$G0Apg#>(x6z|RL3vJHfF>b)ouJtYPZc2nay#n!y&zl&fopfR&T4_Osr_3$L
z=|ggrC7*V=*4-|Oh?LWeV8^kPrx80`XZ)nKS|cT&dw{4=kQVicyEA8*j(K4NLqq6c
z&e!gG7}dre-FWXL7{vTGm&bn)!|aK<N+m(f3cr0(9#yyT$&P<ufpJXUbLHHozre*4
zpVdmC%)vi=Auss1%l5>uAWVk`I=Dj{A2<IXIh(C4IJWfZp+ff$E7CnXZVOrdpWwUZ
zyxOu?HeL5|;<Iw=*HYOZ=_L65d(mtdKs$uqu<>(DEXmmeA(_e%Mm2X3&~bMU0Dny!
zzr|leK0c=(BQ>=dXKoKRjvL|q%T&s0ZQe8!aqt@Uv?2e-3$|!b+1q<D5aD>kPmg&a
zi=ZRToOmpl;GbIpCvj8X?m+af_vYiEbA?X@wni?(PBLr^zXj&K1YxmZ_n$Ndoj8ez
zhop!HEDHl*E7PDMG)YHwOx=*MBU{@HVbdY>FO3grmfR!~(QVnE^RbotyIPs#b0N9X
zggh$~Xf3sw#|+fbSm3K^+1v9Cst_$$u<)8dTGf#(OJ@dVmPO@}Q1E|$M__MFui0<x
zCo~nRWuFW5<^-h{eH*%(cr#ZIixkqQuv(?#&&fnQ8?N~sH$f1gcPsm`wr60mS&FEM
z&R<PrkE>!WimVd)r7HYo!TVdgkW15fi=i*S$XN!(FVxe*xVr|Njk^OA6{L8&kqVPI
zcXlJZP>4nPcz*Hp11%x@ypxiS-3LB?;9mf)8rTn5U@;WxdVYLL=+F5%{3JwkD^Hd3
z!F&@6(buUO$aWhoYe_#_gixCQ3RTDhF*@RS5QgeScx<bG=6&RS`ppYIn7@5Yku7w}
z{r9bUQ;IFIojusQgkIw+I#g6F8Yw&{II{m;+DUR~E_wsV{biN;DLFK5+&;b|rCRUj
zc^_Y}KW3QHEDIlxHTlv#EW!Hsf`dg?d)l8iz!nh3g7-m8y?~D0fqAm^IWD%o@{8ZF
z@jrydVgW*bwI(N$wpJ5sv}Z@6>+El)zH353`s1p_VkJ~MAx8$j^8#;ld!g<@g-GB`
zM333m9W{EXU1?Q5$=})#MTl2SxK^rt-R-}fsw#w(>#^m%@+Ex@Pl9Gs=Y%~QS}FgA
zWnttL4iV1MD;1lz<1K=Jfw)m`vV5Fn4@D_JzQylOy4;B3x4>if+HVE9!6M*n=v6}9
z)LZua{kFV9C?^UQks_q#G6(F~&5gi(2rCYbsf+q&ihQg>G+u%ApPAZ&JxFqgL&+Uf
z@)m}I<a=h`owAz6MGHGx;^_eh{qZ~dGs_v9(=MogbNq!Re<bgL_JFp#lT%Vzkzs#a
z5{cVc$LSeV4o$RwmR(;MSBZs8?H9^WL%<Jz1&!Iw`;WYQixd+Gh~E;6KRR*i6N_@z
z*v$3DKa=~YkoRGbMVkdEy%_quwjcoP!8hiw?~^P57M5uK17KfJOb$FmyY+TUOh}O6
zBD8RfZ^pFb8rb)dIBvnMdEEZa8UG@QQiR!2?Y=5AuIP6{nBRkso@smNX0O$_$mh!$
z4EP69{4+EPdr+#~BO0oz23cAq#S&+457zk>09&@>y=h0Lh68ZsG$5voF(ceIan0ul
z&#IV~mB`#4Ztgv78t%pNKLjVtL&AI-7Vjnb7@>ARSl5ZX`)qoPEs#10ztP8Np-$rM
z<%Ro7ERpKUSd9Hx(?x9!DQB{m{$Fbk`Cln4vd8X_X_Px;RRTCXr4+`g9e8zz*MQx3
z_De_FGm?-?Ia2rW`zH20{kN#4L|N}FD>eKQlmL6vRWojm%qF@*En<wi5L6nwp(3Hq
z6uq;){xANnjnI~oMINdzz=<mrfin0)!l~8z1yIzwC;1^tpadX~9*`<Kd3bQ!3t(H&
z{>&<fp>h!#G5B9GcjkTAXmR}4H;>AN5Y)Z{=w{<!Nx6G4n-K!6nLC#-mxpU~R%Ml|
zR*~w{6j394{@H&DgWzm4iMNt#+pXl9C4uusZp|8S<9YxQAA8#k1%D)Iwtr3HiAkgK
zPsw%2w*qz0H1&`_Ax8<adNXE}#D40L*FQj*Z~czuJ~SM*@9RiHdX}^vDnwA>pGsCZ
z&zpiosbO?__(_hehmd@{?Sb@876Qhsr;1U5Xa!Ij1=4RZ=)8xCyQLCttgicK&gs`3
z9M3apvVL+q&u%txE~yfVR1VNHx!fu=y}G<u0xPtGgz2vl`%=lopi)<h4^R2u_!oDY
zc*wwU8njQ}Soqn2xM7*lXJE?k-+E2B3<7HApv}OCGPp8#)?jOq+CrI#=jE?OSfF_B
z>w#MaecM>nKSS0O=r6V%0{%d+`|_lA8l=%SuZMu5q}MaQ3^dusO#ZjqD+vO?vnrxg
zy%;0CX@VJh)*{3I;{p8<(`IT>WXwC(5&Wr#2VAhD{Zw08kW1WoBe~PPn1SteJrq#m
z2{Pj`Ro*%w#34?EGXHAFVoso(G=gkb=Q|xw2I1YHGL#SHo!*`zGb<`4(dh$yrWsjd
z)G|AcC}{&yD)&e|@%OG~eS*=s$npq?$#fQl0fR5d?iSJwfqbGRLY_V0Hus79RlYU@
zcLK-eQ(><mfgYW#dh13r(e|TKhagw+SSX0ArOSJ@kFE=fVn(N2e`4o<e0qa`{TXUm
z2uJ$eV4`qy+YHzaRn__Nl+W(VS*q&ly$|Uu!#SmQ9G(AO0+NuuF7pDiOR*P!shW9M
zG6@{%bI}~}=ac*=z9b;NfSPB+YIvIV6s94}pan39JqQOANmH3=N&FCX*GVWBO3`_M
zEH0216@-7*?4A7i=5UGkA))%%d*7biiv$V6j(vxY-}qj1yY}h5<J;*Fd42SraTUXl
zPs6ugb7RebP_LkqlBwn*!a3bgnqwBc_w&o|!fD|Abpalr-y@L%q0)TE1fDI2_lZx$
z4Ov;=Nnji~01#ffGF|%pBtYd+H*dh8e2e<MGAi$p9-<BV$uvg2@qJro)=2JGn(oTD
zt0(Dnxj(!(4Q^!+jX=j{#_biQz6q}k2BfA%kSH}dtnpEZQ){aGQOd`Dr2}(Mqur8S
zLT*u|^1u{hy(k|a5`O(cIa#xP@Ok)rZx+rV067u?+DeCpB%NLn(S{SgMoNo7;^Qn?
z8v;~$GE~db24%1~)jx>{&|RY+W=V;@{ruwWW63+2O}t4fAU137WxTvF8FM4e<Qvea
z+%K9yqkSe63X60iuWg3yu;l#dAZ>aTQ~hYbUBU8b;IY301#;s##OyO%lLNDVNMud{
z39jDT<pV2rY6YpVuB|TjLX9R<IYsX1B|zi~s7{>M@ZsUU(tf3*+;RFhLzcG_(0{}4
zw=u=Jwt(-Us;=8fq)I9L6_y&`Q<?wYb|B~~X~-|eLMH*66%bjcDnb2Vp+R@8CA;d6
zsqjd+{shRnHYSV4x5~tc_FF<ABRu4PC0Zi=U}5WTAkWFzD~Y^&c2pmcGS&uGE$2Fr
z{_b<81XJ<)W-X<W%v0XZKtPs}DFNHSA2AsQXKKD{GLA}zVR(<($8|$!9BRoxmui?*
z&6Pr_!Fzvyb)7qhx6-pygARSx*IkCfA5HdzWfzE5*?LE<WhLo-9Qf6^ugrxx*`M&P
z?x>4380Gk<pMG~Jbg(b1_OYvV;X~Lg$OcWW+-i~*JXTO>>2~E=kW<>Zh{UtFJo?Yc
zHsU!_E|v`fDKfKWN0u}Bo?LDFZXx?Xgw0e9F-|Hp*vzX^)ko#Z9VWktcj$V!Il6+{
zY3-W>PCwzOjvn~e63l4=l&|S^+7RVYXD7O?=7lK&xu`#dmXoKRxILS7f9_h{k(~Te
zU9<OT|M?zsNso)_6VA!C`}!U?bb43oL}YgN-Z=M)zC5<$O?=%tSOcT5ODC1IU#=5%
zu&b@LbKTVMT(x_a(d6V37xgIm{fltIiak2`&)9&j&-T#F5vkF7umi)zD38yU#Vm%^
zaro-~M@UKXqg+-@XVyGOf^o&EeKC&PclZO_gy4mXWJ<7GAy_sGXc3rzm+;FuBM1uF
zw3+>-^Y2~=OujvFaq(+uh1qxOEdA~hE8W&);n9k=!iPyR9tR~8+cI9TFVrS{=3+=z
zz1jP;Pq8d#&dV%7_yx+R^LvPknf5#9by88I?*1QTVRQbv=I*o;j#cF;MG<~)@hk-?
zU9EYeUv4Z>`qL_>ty@s|x?OLuQXWjBWp4<(H<(_!en>pg{N`$`iCLwiVP17FPnH$S
zJD<8Wx#kO_qGllr$8X}lq|w1HI)aUEFGKW%<X-id2pfCvBePY`@xJxs-7NF0?lnp(
zKJ7-*C}<(*G0Q0$&o2LHasL-oXI_pD2~QRVQL(PJ*-f6!F5)Pn8Bx=+FhHoo96A+U
z2pwa*yMXDW^pj@6Ere<)z_TOmcvnWXi3NVEv|$S;IV?M#JpFH(fHfG?L7$YYY~~#;
z!yZ&b&#S!*0k%!I<WSo<6x{7zB;H)5Wxw&IgHGu_6l=R`Ay|*>RqTs-yB4HA3{>dr
zQDPodz4m>eAcp^xurB7}i2AOky6CYI7E#=)^q&+(*{36gU(OKMtKrak6r@KUC@H9Q
zv)3yz8_(p3W)|O&&bum}PXFR|JhOri@r%Ps%qW^WX)2*p5As7tdHM}!t3S7vatHM_
zo{fkaVqA*uWD6SN{2u$O%F)u&UGRIK<eFA21!fNcAvp$zK|OL7Xo(mUp-xrFkP-nL
zl7=|ieCuzKucVS=E4+h`T}jrhc8})<dmy5(bLQH!b0RCerftlvXOtt)-SVPKumyHX
zGWYrRvcUU$P{e3YNVBMZdIu7+`WSr-CEvUmaye!r?2W;Pz8d{rHt)Y(MJV*P*=z>a
z(MAOQXeoV0n=8+wM{mTRAXxV+?JnfCVq!B?FZFb$Grkv`I?(slN-?dUx6kZW+Lt6v
zt-u#txOm57bacMIROS{{1w$1aEjkorpGj)?x#avb=s2f$tk|cH@j;0pDLfG7G$A6_
zzIQ*_N5S#P=uSN%c>RqrKuZ%4F`SyU#3k%Ws!KcW)XUs?<YGg+#zAC&HogIWq>Fo^
z4MHZ2)1Pd-9wv>j??Q+UZZ<uE=_UP3&MyPuA!q;kIyt6>7A-L<{qxz@Xb>Gy$dE#X
z5HT>?{q;4x8hQ?$RL9x!-?g1zqgBy-3Qn1wt1j7{WYOEBxiO^e9OP2|N&=oQmqwx{
zQxp91p-Z6$T0t(QI0K=(7Yg4QTH!rsVDOO_W+SLeGp~I`_t*V^lXaOyX>HQVqX5bk
zO~79ja|{#Jb(s6{X#qTS7YNP|$$vY#{o>Y+zhmEA2suu&ardo1n(5&oW=pD9vk;Pn
z0|c0waiVUNFQGE>+@;dIJO3@O4l@56U$4TLE>1t`Q+Sv1b?2&>J=k<gVg`v)*Ckto
z_;dbsdwaV<^t*#w51ML;ZNo4Os;Di_U?z80V!ARq7(EQm$Pg*UjEsz^J*{5)VAZz_
zzOp3pW0XJ9h=|1pxBWL<=iZ+-l5s$W;49JHwH9xce$)XKewnrAhttlfB47I@j1nEM
zoX}KNTjUOqog7q7wyfMY8!Nf-z@w-?em}AcDyE+y6M`N?#S^k{RsFuL1-*R;gG@vW
zl<^UEN)L25>ZL^6KDD3h;Tt;?e0Xcn=+QM7LMXaOs0p@55#E=Q_&#M~A0W{PWf)0;
zBB$mm^<JR>PFb9IP0EFcZ{rBtZwGCSu|hT~t-sntQ6t+W5HOvs!6ZieS7K5oXf0eM
ziCB#q!@trzeK$zsQ`@71E#3N|CuSc35o-veD9?M55ohfhzDb1~;R__Yk3vAmDvu5h
z4HosOOEfn&&V_!(mqwKx-aG4Th!vi{^l9G3%?aGN?l2p<lr$a!y6!@g)`@PTIQ#xe
zWkJKHSPKLoi)nP(USc?H0$=4c&d=@{QdwV03e6o2Xm)lG>?LeIIn4W#_o<Z{9p_O7
zC~`f6^CjMynwe>-R=F%rWexn=xBaN~S8EPSQKmLt^)=XC10d6aRqzTfQ~U^^#I$9w
zvZ!iG;`W%yq0+EYb!qG=mDe@?`iZd?I+&Z~5r9q*uBi2D-R9hWv*t^%LnhGZjK4;h
zTyeub&6wkh@66EC3EegR@&Y-g!1t8cfMwN^VJ}E?w}y^(=ESwP-KRV(+Ym)UOs%8`
zc-B-G)MU+03tSmcWo~Y3y94-C!Jj{WSTjXb4=E${+Ky5?ql}CU((}Ff^$n*|lpa)$
z{UCV;gaw`63a0`-etz571KlAL#Rq<jL>{ALwGU@@CHMEVtS^J467f~&jTbdw&b9t}
zneKwI^+$%oN#%?0L>IUnvD}t@FKbZNgV7ioM*6M&zn@U4+=GqHLvw$=f$H`QZP;Ke
zhWQd6V2d*JFdSoHkL1%mTrYYp%PmOK>lnAj4%b4of=l$ca-uFRb52BVb=MZMsv<iP
zD#+$-sfr6BIj%`cF{!j$qbja~!AgnvRayj#%+%dn#4S6DJJQr%-nOH#r@S%Dd#{>i
zN)fScZaH`@FsCEWVkYa7+ZMDN@Yk)_RTaB-rU`kSA!cL+BJ{Itp#8v62=8t>PPg&6
z&IZK|)d{9CS6tt`0k)Y4UfOK$T4A%mwTD|PV`DtXBaxSe4cB?(2b@fBi2Xl-k0f_)
z(QiFAH3pU&kDLL4^?S<gB?t=^MFJalrK6<EfQ*dnsLxBlLH@E-7_>R;wt2m1e6GR!
zZ3BR;^Vi6JKhASU;+E9)?Cqz(>?h=v8y2@Rm#S7x%AeKJbU<sVUELj+Jy?ID*+WVQ
zfgEA=^$LEE)^8G7jONfT<JJEpCp^hJ4hyVqsqqJ{rXnS4=`A;n@V)R_!w(aBhM9ga
zNYv`M(&;@VuOdSt#I8TpqAyl6LAfggVhoZdH41(Qr%h4Nc|?p}+L#Mt4)RlO`L%J8
zMtqC?`ur2Pj`L2B;2PpG=xKD4S0F;&kAc)MnVU%$ijmJ$g+~*Eh?NZ6ZtdG1F-`DO
zf{VVT*;j|-%TH?lkPQ;s-dV_q-#P8?=s)}Tj9)Hv7-|5dZVsvg8s<;0ZZ9Nk3kkDA
z+WtY6P|H)f27fJ39+I{@k)H<U9+1-ZfU&VS<u<42;=^-ldtM$t^Q=p&fX$PAF`M+I
zfFZ*__DzZn?rZxYlCSu7?E%<Xb|z*u9ozN*dz|Pd;)ey(^Dg}Lun(x}g(Wmz(g&R1
z{!rl_IH9C6tXZZM<@RUDTT$zxr@2{HUvO#_H+JOcPFIt;BAs;YZRCf^-tMfXU>Q#v
zR_SX^54Z0Nk8aG+{1rW%6sFU|`UWRZw(R))q==UxH`(@EF_*$yMgQ=X9j!}sy%Wgx
z92LG36Kw2SKg=WVV93-g1om`8vGrTq!>k=Y>vZ7vB}r6>6Dt4MUr&32{D-yw?2m-!
z#X#mw+j~{7{SfIv;d)aw&Pi+jJILz<U-tZ^&eGOx$cVfyD)g{Z7W?7;`DU4nbCvf|
zc{5=VCpR7ftB38%jG!TntyMwXqAltm_0dx8%h;{C#077A*b+j<>&^mgit5(?4x1b-
zD+YnrEPZmPjkczmxpI3N(b=RIu=ti~YN?H_y77lX9?Hpaalop2b>azAZ0e2wSRcak
zZoF`!!+PjrF|reXEg5oRdGvH>@K?wdc9z$V1I)}3eJ}FY@F&ct=<E`-^rUt!0Ve0@
z0rEulH!;@&w(Z?Z)H7v`n^wiUUs7;G5wc`h#VPA=-Kt(0=JnToeZIeL-9v$v=srYV
zuszk-JJ`0V&wWe68jm`+w>~`i5EBwM0eiIsX@rGiGNW}_Y;PbXBn^(NVT5f!$8TP5
z>O`|+D)Wy>kg0%4Dz;@ucIMVxV{#EY2HlUnCt+I%1N(Wh;@8Il=;`f?jEeFu>B6Q-
zhqC9L@w@OE<m(3vImOmyp7+}#kqSb;uA1?+>OxvZM}!PS>;_sp`W!uLugBf&Xvh5|
zMk~aiggn4P;Q9!mzkbezUDRN#!wfwgu@N5lbK<5SPYE?Ff0RHT@&zaIM$gWF%oST5
zm3IKLW)E_3%;%+#j)-G$=+t%utrxS)-2Z`}hXrV&F<kWKkZ>0w4*?Cid2nu27qHY!
zk#lF%#`DRyU(w%Z&!Ec4i`x9{SB94-Bn5-2(g7`;1<=dc3k4AuB&mY#LkEPW-&-GW
zW)jH&1qSY&c}b`tryA>gC#dRAdrlANUVOtBLbQuOmqtL%{$QvYdTHD0+R{i&u6?)=
zB1n~w+P=RhCWQ1pO=56D*}aU6MYzRCop&3rJ<&RJH<}v7*oPwI+sI(`F*qw{G{CF}
zoVN&+j*zm<P#J5(PB5WC%B--|+|_Nl^yIk=PhDDLq#x1tB7sexD|hQZ=;Gd=O@y#T
z&EeIc55RHbG#`Ls@)6juzhVSo8usVB>C)kJa$PL{Tm%G((plrmdlz+V1X|O&Uw#sD
zfd)s@{13jqr9X4cg;57=-`(bR>?<r_E}*EFgzlFg9>oFc8X0B*Z|GC5Fz4l3Ql~T(
zUw%du(?!gcPxD*mb5mEknR-vy^nVue9M1S_1N`{xi?=JZzlz{FMvo^Avf|gHVk!o!
zJG|T1U(tL_F|dC;`n!dU)6Dd;#d$@weboE~o}!&|k7t*d>_!v#QVJHUlh^Zo2OR50
z4hYLTfQGZ1x=-kZq)U6Seu!JF7)#GE2P85%yZxAeEx=D%oEPrSHiri~Q=gXwI4st6
zRx(P|tqcy`YOpAy+$;BI&!Y0Ina8oF6SOr&gAa$u*=zFVb!pkx%mSL&R%R91`h=>*
z#&wpzuJ_*&Tj`~anh=?jbgMPu;_BN+8g3W-oTJ*@c|n)f-z$?8UE|KS$CT*K=qcJY
zJ0g{|Ayt2aB&V@*5pW@_(h$}pO})&T>Vs|%pZk99^Kn2WFCY931a(tr_%q6|wE+HT
z)<D=Q9T#z?OIl|~hh}Yi)>~mCi2pdhWMhO$u;+-{yu=T$8CAw#*3`?ZpC)cK()s=|
zCM)}+RZyg^Xy%csn#F7G?<VQ&SJiTrmL;`cH!!3LVEuBK*RASAl!-hIdXrkRr}4v>
z_XW|WtLfux+SP&N5=WX#uXC)wa9xw+uv7W`u5z6_hN*w?=*J+%e%WAB0zyr~%cHu`
zS0Ud8=Q=4VsAw~1^cDxk&JW0!d6!s*4yGW;?3=@4=1AZ1xMdY^hFY)CUfJ<5xEqM|
zV>c=G-Z3=_Q>ggD#0hBfgkq=`QPF?gx9^+TC;CQ%#z9AtG%XT^_3?iOod~T`TZ7xn
zkS4uw$(msfY6UdJ<jTW?lCf{iwhbi?hk7*SoXQ7TnBRO|it9JuQp;$3hF`3LS{5tm
z0r8Q)Hv2tPUL6_Tx-jk*-C1cGuKf5bkM?oSfa&ZdqP?kyNbe#?bYUCLQDAMUapjJJ
zQ$@4P1!d0T*YxM-=Z6*bq>gG}i_AJ(`i1jdm@4(CPE~$tTh(G2!%EF-ev)OoAfGgS
zJ^AiUTzveie)FKIffFT}Z;!Y2yTFnI%2AV4=1F_Rz_IG+LF*#Fb@P|#@uLuaA~x3I
zH?;WYU-I6UbmTVF=%(n;TI{3RaXGA%Ki?&);CLu9*%NG7GmO<0cLQ%MJ?VWO#-uuD
z6}AQdO>2FZQy1Sqr|*37;q<oG3pfIyWc}2!1An=o%UA_XCCHYN7&1{&*S75zUK$)x
zgPDh-=Jo(X&`El~B~YpM*}K67D6m0>Ir&-KXy#k(0yLS^RMD94B&5t&u!8zV<#{Zs
zcRPvt=n8fq8<+zO-vQ(*-V4BvTtXP8M#yP4!mzR|fT)0kR8b->4tzaI+g=<4IUnE*
zVULNnM*+YR43c6ZC;K2`I!NOTl45vGHlA;~>uAt=aW^!;<bti=SAOoub`*rcIU^9o
z00im|aN=7pQe*<qPW$pSRow6UE|ZSya*bbzc1<-Kfk!m{Yx&8s<mVL@!Lo_Vj%;TJ
zcETcJm2AXYCMPqFL&+|(1^m-=^5n}wKa*eK+k*~SH7BIJql_{;UV%K-6-hJU8A|%0
zneYkS$YTrG;rvdscW5gnI&*Y~r9<-{{)#OgsoKjf`e#Tf)XM@f`%WZ3R+9FEDOqE~
zO?_5BA@WN0%R`}cdr7w=aXj|peEy3v8{S=lwi&Xbm|-`gv&h^l(n408md6|w6=eaP
z8Vi5Mb9#YxYmSIrK&sLMjXasjjKLqo`vXCsb8COblbn+?Fb=-6lenW5_Mzg%54(WG
zOPQ0OW$>?vCf7PderdVyyXbeX5CH9%_wIZ8czb_~%fk{%D3NET*d5JVghCHOEFe~&
zfO;(%oAB)JkQh@mUkE9C;y3<hwf?MtDhJ0-@&*@Fd5R3PpxsCpurAL>Mnh{+2W-G+
zzap^Pw?j7<d!7nDtI%CcWw+}E@f)BHLM8=?3wf$$?dlLD(eVm45t7+K4s59uaJIkw
zub2;6Bv59to|@IvNZ0xs)WF)(zz8})2N2P*XLJG|sZ;yF+3)>+o?>TNp#_kIqF;#_
zrm||ZCy?$Z>9aD4(6VIr-FZMy3@_p3QX9GC^k|#9qYD7bbU}kzj0r{Zj!P#VI*qeo
zJrb>At6fvc&Gz>Y1ZQ9_c?Uxdon@y{cGslzkAmg))<ox{OhqJdJo)91ymsb~cnD2y
z)|bm2`++}~fPY&HV0Ope{Q+BEXv%T)`MWGIQ)?j2(gVTNM*AHUkN0Och!ehG&vjpQ
z&PYqUCDYD4oyfxx_3-tL63dwO&dx9sGG+%Agrqbssy{w9_Co0_UJ->Y<>;?km3_o5
zHw?9hth~H{U%}7ucB!a8+077pWMpNL@#*qp9K?={7$R+2^Ap_#Jgnjzu(PqOR5xB)
z$CDlBz3^T05f9@qv{ssF<?3DPsau3@FJWj~BvD9ye{RRf5F^X$O~F~F?NQi+ZP{@D
zKp*NOvklC^g$AwgM6$o$brHF!V=1;T1o)Ha5@rJ=E%o~pQgTOzkj!>-LzDYqnDv!d
z(cN%BrE=JL`}vOXwmnH@_p4WC8@>id-rO}nD5C#Tc>Etc*>x%uv_Gt+@#;EGwMc&>
z9MR_dg3kch9}}K?59Q{#4zmHB^3-)}kWq@NqbUXdLN9lDX8VHJS(<-V;2Q{zD^!_9
zwjazB6HFx37C77F;r_PH@5DGN_|2okC9U#e{X~q8$PR=Sd_ANS8;Z%-v)7K@i}`EU
z9^bZL{6A2kI{)TPiq!Sx3`tY<0yf_kEowXo96%?c2Oz}H>Fx++Eyr!r%mGm@0qvzx
zR0JF^Xp&)Y?=KI2157h}CRge10=|F7-%(o1{+A`hj6i?o=D?)ANNZRG6JtKwk_42K
z<I2(vP0;@O?||mhNSsKSy^%xr&$eSG89I>&3SI*JA8gu+gq4*@rIO6DOY5qjCXDoO
zg_$V+wRfl$02%U?PMdp6XP1Q}W!YQia0`Y%fO6W;0!_R|oq_pTd5lmQw@`lvrx@@Z
zu+Qq_X>{p{Tt1Zh`$rI4iC-IF^Q>x^?_4547ll+^iY|rc|2tJdu!x|zmBfr*dH30%
z8lf?{0Yrc#i(`3l`qayVRJ$hWDHm+<93f|~-k)!;Gj@wxrYf(*)X?mHL6c<2ZSc6S
z;8HD@`=|ceWzcykLArvepY;`*1aW8<*29Tfhg@K)CXmk6$jA!)_^bmByEr_w14B6S
zo`|Iq5C(U)dy+dT9~5M$kx@|XvKg$~z)V18Y60o_W%aPI$kK+cWyC82-2K<pjER{Q
z1YPS-^S6`-_=(9S)n%!a{F8mxQvXo!&xh2<;nP*xZlD5~0wuu190-WXCVTfEDg-I7
zex!x=cE<yYzw8t9m*>i9!tQJeX`V@CUb-{g4R(!BYInzhyj~Y~<&u;~0ix%2;@%hR
z)K~ezU6dc^O4!z}rLX*_zd9-^a^jme7i-)B1*KhckWZtRTc1DRMtySIN)yXt`m0yz
z#`=DD-J!GVWcg|!C%3zWi3?9-{PW3<8}2+~J0lL0RF|aUY&vr>quO@AqjHR?9j<Ek
zC$&e=691(1;yRc?6<D5*6SHsXO7l&lrR<5AcCH6&iFphD<$)ta0?1AX@kU+M*oS>e
zaiiM3vc0YC1~f{sCCjOtJ%*AUxM8)0@fuS<RtnCl0d%fk7L4r}zW5OJ#Yj|?fK_o_
z-L>TzG52qEX(}x3M{Xi+^R5aKc$PU%TNV`+o%rx{o=pK)TH39aqhkSGe3>;piVQGF
zxLKm@dR6I*!o<diXGeC-(HMoq$rL4l_D|Y1f4iXj9R7>_mB?uk%dmrkRGq3;C#`@m
z*q^T%`e6IVkhe47FvCAjKMF5%69b4=;QwXC+#-gJeY4Ow0AZ13A4F&@Y2)9(Gq<J7
zfYeX6{b)-q1cxoALRP!i#Y<xd-amYQsw|JC{e=>NbeTt3r)Xl~_H2NS1a#AV+U2u!
z+c*MQK;1@bmiNQTtHbVk2+m1o#h7=F>#ysn)tm}a0x&sv#l%31<-o2ZP_V@+)Fs}8
zGiV`VXUpP2#LP+toEs2Sh5tFp_8mcQn*}1f;nL5{eA}j)++d@CFuoy|Rz|92>kP;I
z_V)q?!R5r0v%9$QH#d#d&~s(MPP&|bLS#{-D1=bl%%FgRbWyv{XXth#K@3ny3i=`R
zKxD3ZpcEV>-+g2M_Ugu<U!&G8gdB@fk{nc#bDgUr4hw&=yY|FY5o9rl?JjbJw7!1A
zDyRSzKu?4Sx3qQ*;;_csSb{&CogMxw`n^lI!p(9tg-2-le&9~fJIn}Jxs>W35FWGR
zU`Cb$l1?MkU4E$kBNexX4~Z*y;b{}IDdC!L3G$rP`O3GU((6nRQ$2@1?TK+B4Za<X
z2gFz_2U}sZ5v^m&Jgd%>jk8Y+pFz;p$7LAfOLzc@Yls+*glDFu<qcH2^un2{R)|QC
z+&)=ZSrdJ&;*d+(4J9G8zjh~emKIW+mB|}a+oO`|fJ_MD&8Xxqys|7ErrD4gfVK*|
zY~*ObNte&du`K@KvPxkyTyCS%oX`E(*MYp$`}6CY){_Wk#;CQ6=aBy&2ptYjbonSd
zqE5)&|8@Dg*q;DWsXxhp)ntl5n`|-EAn7hx<HQ7iE&&DW91(4StOds@2)c@kJKSe{
zbPMs>@K3wTUpWQ!HHd6T2c>p_{Uju3nne!nG$ZJ)_E=YwTU&<a@7(q|oB-`jYb;Aw
zH}FMq-6pL)S~{2TaFltW^V{2bg|2@kAqxLtl^_ETkMw8TUaOv%c%sWDlHCc{4{Jf^
zpQeVc+q|kk78K%5ekn1fG9(Aq6*Mvym8^gOB;sn$6~*>Jx`O04(tQROd#Ap42cMyA
zxYs{=#9fuNWrV%F<NAv}1BJ&B-l^lC3LD~Us#<)0JWpxJGniUQTJ^o^e!Wt7(d9eq
z%Iq)wtVWbj@?BT^xvoDW0qR|^ydQR__C7CR8VEO=SD(@=X!58zbi!6*#MelWX(UAc
zi5ZUl6@6iNcAk*4`PEb*YKA)m*{|O0oo@=7TF-i2#g=}HavNkBMot#*JYWmePX}5$
zA%ZG6!PEg*d|FUQHU}0TYkFj{6CCOu?v;Xw`LQS`cR_PI0`S?>=7G&+Gj%#pnWt%$
zDO)5K;-decJ*(Maa(G(zBG5IHDz8R7$H<}F_0n2CUKc!F$X;R5aGqxAhtu~p4%w*{
zIJjzLg-4HrkcXIA0(bqt^{C~!p59Oo$0xF@%OCiMXVN(JoVsrfny-8fGw5>e4O-oE
zt-sDU`IYP46@hGx)&;3X-(b(|yu0R~T200+-xtSUOLVwRK$<XqR^;`Cb?V%3g8+Mh
zX%E@>i~Vu&Y4GBf08&O@!GB{;^e4W<g<z&DznTP!BBO!x)jJEKYyEb>Hq*xOEPuOS
zkm!}<{M!V){#3H)0bcvDw)bP><Et7WGZTUtmAY3uvfjEFLn@kWZv~!6`X(j6t_f-_
ztl+8eso}hY;$rn;ys*IcTLDe1W}qkti9&ho-lfbYA>uof)>yIW<cGInQ`Qz=PHWKS
zgf&2TQ|g%{s=J!Uc0ME_Yt+Wdtte@tCCR7VNlY#Xe;sjCbK2NYh2esxry5GD<ay5X
zV+u`Twd@|j#YYya*2gZUtqUCHmAab14mKBqRM@15-XU_N$kFf!$X8sq%;3_DyEHY`
z)9uF{p_Wh1GzYle@W_6aD~*wvyLQ2r9M~cURK~bF$5En26BEa6^%QqQ1u1HzZVXNh
zk@}!po=Xv%JSo}+Y=t}3%l*?62qmxDdb!m*wmrRmx-3EOg?puX19=imHKH*x`ABRf
zzMR4@xu|Fb;q)P5#pXGc3%1oYh_U9s_fs#UtHb)uq5Q}Y>DtRkG=`Uhx=8mV&w*!{
zM?}=A$ESQH1!s#DpV&b5i3$NE%UI6oU7fC>v;W2`coE$$Fl4M-KcwT8hK$5r!u)b+
zS>a(-as}-wu)x`B(pBqg%L_ibE3exOiYs{$Oa+Jp{V&g$hNN{`RV^iJ$!d2GxGie;
z{_qtZN+}NAIJ#KY&QJ7@F;XXI`H?om{g<=<M6Q6zzcu^Fq5TupO)Pb8^zR?OpO}v)
z?Ig|Q%w-s64i;5Y`*F40StvW%3Z%X-YJVZ-XW(B0-ET6_huS)3sTgI>5pjc3mBBHY
z&T6WAvwtdfFeY;5N?>nTh1AEX{b^A|?asz+5(FZ8aDvY?SleyBeF0#dF7@g^s#<xs
zujY-BzdS)CxgWTsd<6L4S5hJ>=bvLpw~NvMcF-s&v;dFChV}WprfT5s3P_hm7owB8
zkY2#fO07Xj)8lyHHs+V{N~>szVMb%)7N;_^^^(?^6(hRXw@swKk0&bRSeVi%$+V#m
zn58e4=|R#b*?o3$MujR=y#3+=2{fI2vkXn(>>OK@WzF<Fg_|p(j1&9d(Y;_7x3_iV
z9V)r>f%B<m=WdyXBAus>+aUxPl1-P*!OAC>q<N!dRha1-&CU?vm5@sKJFzvJagZKF
z;ef-C9QAnx)@Kn2Bx=dZrjqef^?rSzVIJqh5!?4cn9no(BG*xNJ?9M9z_1Ss_G87B
z^Te)(KJRR|?)Zfi3VBddR+vp$YS>{9L13z}0g4WGtC(!@rp##!B5sr3Be-6a`fSqW
zNd#?SBP$Uh%jBdHN?n#zXI~cGOG9ER>EplKVn&bLX_~*LaJns;o%JRzE$uz@2pa)Z
zdPGM#UELUgN;Nki>g??1DEf6)s^*^}zP`Q;BJxWdtXIYQ%ISVbhfaM_EoNiQFM~RG
z@<TF+T#smov9@z>K?U~heY!Y(%}fMM5trYAO-Vx`?SfiW0g`why6|O>BdvwD@g1qS
zKW!<-vJoUHfyw)cxEe10EWPzeUQ3qUP`m7IcUcO`D><Z)G);PjkXVVPsjQ_$+&{l%
z(Ry@@_5nJ{RIHOOP6==f@kJ}$7Zq_NN}-EuY<EXVlxQ@=S0IAJzCJKSm)KZdzKERP
z-g(o`w;xCw2vMmAB&T>fU`x8p{RprTJuk$34uT+2n`gC$P8#NEoEHw)*spf@031X3
zi$C6SOe5(sq3GUO88eO^r3=;CO64DX_l`-bFJ>;BBR#DLRHF+Iz+ikrwp<IL(z`U7
zskIYvJneIjeXAQqTGR}Yiju-GYUUG@)=HX_n%!p+!#9W0ZOPJ~BbXmloJ&DdwhPin
zE4M#$d7wYh+o2{cRd)W`>bb|(ODHc78r}~XmrtYOWE<jUdHhyO7_2d#t)oOTP1L^|
zHHC4pgrWr<`j?Q6i9J9C$g|AJ6kD&Nxx1;kWlB5MQQR>k?uL}q&jD0cg$uBI_p2qG
zu;*0AdVbMOuK8XU|NV{tqXDE=Z}eUF4M~0}xOrFzzX6MpKg5{gc-E9+C{hnz3B9<z
zSFtUmNZyag(T=uEl#L*@y}o;ibIXyG)Jo__HAbg3U6YBp_M)P*{}8mcc6?}O!A%wZ
zZafk&yb)mMAK64xG=VEbH}Q6%hdTU0sI1(WsN8NM{5Tdfa7oJde3}hQfh1GRUCK*%
zbKH%6*gG3fO%q2b8uZ1hFVOJU23rHSLPTxsT3_C}__>-}Nb#sgCjRaHP+BTwCc=u&
zm^aGG3g^!AeMz(T3YGrgv$(yv_n;<`7l{^ib6zejAcU1Gjb`?w<^90MhiyvK4tvFH
zLZY+5rU}N>5RvJ+cC?l8C`(KApSUO}5-_XtD?cVwwnlj=X`je>A|7S$*oaW&*xias
z3o`C%LI>xs?7^}Vd7<uIYOR<cc4ds1s?cqr+r01)E6!8LwbO{hvZ+jUb%wQdb_Nwb
znbd|!4)yr<fvHuE^d_ed&Pah=C4y>B=FG;DLJ(k;JS6@`3KCG=_2`1kv-Ltw*%&JX
zM_N618IfE7i@J2_BhF+1KYZ+W<_x|nX7O<fim)w+TU0sV9;`s(`)uTGUfjl!1{6P4
z-Ya=WTq0<y8AhREH97g}BmU?#FdM)vEr!$Py~ZMY@aJ1?BX{TLHA3WaLRc!nEXtV>
zD@R5~W>!@V2suuf10}v{D?y~Zrty9+MEdo*X!P%Q|HTnRR~|MN;jYh`r>wYw)G{O}
zB*^|GRmrpKc)?qGo;TLyCy;MH!}lX1Ej#T}Q>8->+p8MVZP$a@uz{_|W(hyBYgE)f
zJapn1Z-bo3yvq2egh&0>Z^-z8l3EV6UJDRZ$V0mOK!()}bQR>kILn9wnQy!^rv!8^
zELW)KPpCc|Zapjn17;^SR><mJbD|4OrzXd@YRU5RF3n|YuWyO%C*=BNIx`v~xYlfv
z>}g{@l=FtuKy@hP<iS6;av?2X9g}A8Vaj#-WoTw@E+Zy?B!}2MZ4W9f>tfLpTxcw|
z!N7R*s3G?+GH1gBeh)SHXIW!0c6J3>I+Z=3#I8}{n~!ArdqCV`JJlDghH6iG?GT)D
z5ccx1c9CfasS>N+5~%0*fYcUJRC`})ow>0BVlbGWIjowl#e{6yIJ-EX49gPbTW~q3
z<Xw7Ifit;lAzwY{lD9z7y9wgsG>M?Fl1_B}$`x-jn_@)V_k7bu{kWw@Bbn+NHa9R1
z?PJoFhyiUO4OaZpT68Ez>_bHDLgk%|m@?e{_V~H{U}Vlxow03ujK-D;6{I>H+#G)l
zRF>0mvgSQoq%R^eQeczkh%3H{VQa!!PF!j|&|f5yAZc6cv<<U=y}KwnUGeMqFcdf@
zpiEVa@KCQk`wuzlpajT>U7$2%gk-w+n9C1hUmR%eFHh}LEy{b=Uzs8GDkJlAf3M5v
zeuhDgOL*P`_9TnmA%(><Kb!l7FTL{dC;BmUf`_@gh_aqK(Y#1u|L4SgJR#Z&&F8I4
zJ#j|+eS(UGw0m16D5-u-#o#uoND85-C5OdB1~PciZ0>HWT}D{7P-E{dsha0NG+r}3
zplJpAM7ki|Y@iIed<NqmsE<<sUP#cL)jZzSJ&9uDOQTaEX4<1%gH2{m3wxu~YO^><
zl#-8ASCGj9X&7oW)M$lgXoDI<VO|!#@T1f}+xE4DvR$fC9ws(@72&)$L8igPnT6^S
zUTLhr0-3=<6a$&kz9YJ)*vq~WlR$L?P^^zGTs;O%5)tp#6mj{(jF$_GS8`5W!mn%o
zWW7n?75vpXP|FN!dr+x3`_jA5Nkr#AyYifCEgY62pF4QqUV1KUCYc5zwM^e9ge<)^
zw`dU8v=7B_SgCH%#$~do(|l=m*yQG;6Au6NSGKBzD;$D1Y`Dc0Vsw;5>0#k#V+#{P
zmMcP`^^C|X9V4N-OpTT}N}n92YKm4a-Cs0!@Fu6%r6$Aa@+8sE*QQn<SjX3IFq?49
z$h(YP6I9o>4;ddjq0);4MqAV;tOO}};<LFamhsmViR{f&8LvA+30X+*vIk_96Wj(2
z1S%N`Sj-Q2mMctTd~e$H3>9jW>14FB2?^vw&ZH6<nX<uanFt|}1O|N*8@P{ULt+k!
zAV_4|0Rk6=<#Y2>{k^yIZoK4U6d-(x21iF+uUbap6@0aC=Ey)#AliW9c@k?cH3v(E
z`6JrWGC~%ggF4?nz!>(9_GIfgL++iI66SoRlKO(p-UQU%Ai(`qk}#6q!S9aSJsU-+
zm+5c0oRYo+{g}xwuj|ztc78L%_dd#!L@;lXF6TAvPfIL)Ui}1A8yDpQ+fffDnpxEc
zMa5?&H0r&R%6@S^LL0c^Ra(#RukkFVPKl`L6fsF?=qzV+C0Urs;M+V?@HA~Lu!t!O
z=MI_<r}3l|&ovfChfYd7eJ<MlP5whc<nk0ZorA*l{M*{dK$u}&y@O-&2QJf*P63dD
zD2Wv35f*Iv@5hflr|1W4nP$2~X{|U^OwFXZ{70u)vrB)alp4GY`mM&gw#gt>kiz2h
zbeehTX0kG$MjUH|b;O!bILt$MF-{^P_Rj!UoA~-wx+cWh87pWj?*;c~rmx(H_;C8u
zNSX_{HJ|-b6_5vHwab-TQxy{sX~eW3C&c?emc&b_=@s$>XMizlC=`C`g8g_yI=6L2
z&xqIb3WyRQrUrwS#<t;f$CkzU8Hk<t5nb?3E;{maWst77D_+e^)9L~T3srd#Eisy0
z7D176s{W*B^dl!z+-Lw<uRyx*g!pqCH9&MAs<8OkkY&8Wrt=rqMk7v6RB?^nD>f}l
zpvY1|8oZy9=3P19-qlv14eDmk9;ib+5AH4k_aSs^%kc9iEoV0vz?RXiK*4Ki_vhFw
zt&l(5lg05y;p}1f0rKGWN@!->f4o^)*^`zsc-3DM@uX|U#ccXYhLTOUVe8Q&j|**-
zc@$@^bpfYc?S*Ns#^!IIvcq`Ldlv|9U25ZcsdtbcAR3w?z_SmT5bxi-kFl<2N9ZJ7
z9t1Bt_#cc08iwEYxyV8;q{wZIa?$_Y8;n+#7illMJ4%`B#JCAMD!hU*$@~%{-M%&0
zn<FlR+zC0p{AR>S=O6Jk`7ze7NeEf?8ia3@juq^YUzKUgJ=lolwtk9p+r-TmXwPf?
z{{rW-vc;DDLK!(Z&#z&buND%<#!ca=JjPEMIuj;#RfWAw^X;8B;ah#Ep{=c5qa<?~
z?SX~PVJ&_AGvCuzNHhw_un!;ZYyH4yW@grV<EchFYja<pIihbBE8?QYMlg@0XL!w>
zBxD?+4qLhZib9AGJtrLd0_+HGY}YB32yKlmM%V9CE|!J9YpcAOX0V!N^?NseL|m=7
zE@XKI>@)@77_L?q!#PXH)ByLySl_&XqaiN8$1&E9gHyXfy4`AxS6#ZgC~01_rtz>|
zsQxO;4x@C#X;&b1@&rF2Jcu(2b)K!h{`};T=hwDqVa%l%X|wlar`97ugsXc@^xmt9
zNu!&2-#9KwtqUruX=*Vt-`lOH!}UHPZ4_zT%e=9TA?i9N`%k%AYm$|gJA!@QoEKts
zoNq_ylRS>f84eSRqnyH}yLdBfzzfLA1~uBxQI@C~U>cNNH~sXo+%cVadUI}>y)cLL
zwHT$Ym^Ba*{e{fKrnrad>o338VV-u72!3&Ns*s0;$8nJ}=!ELN+MCcv*D*D)G|(^n
zNchS5VuPc{$xqV+l)F57ED4n4zJ6if`y`Uce@VQ<5JPbEK=9nuz<V7;EYzYa_m%VF
zltqltgBFGJf-X8+wM$Gm>L`pr*<+X9+vW}DUcCp>)9$+UBE0NFKY-}a!f_mh?b=m`
zg(uDl$T0}ClDsh5qQp{A8z1fE>7cCOwL0GLmdt6cZ1`1IC`F$m^@jV{Sp9pjJP4Iw
z2@s=Kh~mQDeFv=tK6XPmHVF-M#nn*Q1Wg83<P4D7+j%^OPw##D*X+L~l}5_O*bWrr
zhc8L^*usiqUzHmZ6ZZS3Fjz~PH5gC4Fy*|uYwit6$B@DXA<5h8app5ABN`vv7Q@Me
znT;uuk+}+>fX4x1g$#tQajPQ&)siX`aPCANC?S+AKtb~@!W@NjprmjAKka>GR8(!)
zt|Frd1}dQ-f<c!^Dbk8cNas+Z(xu4It%M+;G)N0bBOqM{p|mh`i*!jhoO`1n@3+2n
z{-3qZ`@`k(otgdYy`Q+_iVM3hDEbvj&se4Mrf=-Ow(s8!66xJUfH{UfC-27Zuj;SO
z5GEyxKF_GbZ&5jT)~dS)6Q0S|Y;k;d{z$?0GN9sjX5aRrz#x6*P6-rv^ni!M*5k^2
zBZ)k}30R?iv?H4~&{VW2KXtbq6zj69Z*@`LFng?6vxn_`K^Ofgo;zI|B57uedPCKy
z<&yG;y<a(`zE5zazn#mkY7j7aI-^zCm49Je_FuMvJkaD*c=6ObbcLdyo{WfU$P%!I
zJ2_~eILU8_<Q(MhvVen2ahL+iWe8$TWhmQf_8~dQX+iP59wf(uT6mat8}BHh=p5~X
zPrsA>n?=rZ2oA>@@zQ~$W)iA8nRQVDeVEKvijoF!dPj^ygk{12c1j?tVfL~O=UqhH
zEb9#Pde`TpwkInqjC7ys?9NttTr^H_w`eiZiqT%H1MWf(;S<OQBBhY!MniAZxuu?B
zC>l-8DQrzqdkb{{2%#-DyF@$m&k=b5=hP_}=WPqzia+P!&LoXqL3+fTTfZv3$+H=r
z1{A6>I(0TSzE&ux(T<TEZ=EK#0Kv!Bl^=J9WGA!ebf5Wi*BjJk@we8*$|^{I3k#EW
z>tap}s23?b)ee{v{K5#cZ^gFa7c;+^nvPvJ0Av7cbVI>H*tX6kf1RSSr>(7irJJuX
zHp@}dnD@)WruJ!>%$saXC2TB#xs%q;wkDm{0$p>N_Ssu8&1vP7w9$rqe6bs49UVm^
zQ$GVO08N?eI<K!nQNUxVub1$FoNEOy>Cw=7vo#ClL3Kv^a?7_TW<HqoTHdxFYuoM$
z!k<sg$Y5>15%={s0#di$K^Z_SZUkmMD#qrZq?<6^Cv&fmuoEx_@1z5MS2yOQ)QB_R
zV@x+0pdSkKFI~Wv8>EUyl;XSrV7Q=WBp~%oyNt5$TZIN$ODg}$JA^quw$T?a9vJ@K
z&91fRa@4Y(XrZ=dErD9t^6x0llxA^(r{bf+?<F&AC?7nXi6QD+zqDK%m5U;}-8Awb
zV4M39bCtcidIV;`u5;;xcY8OK&<ZmLYX;maJboyMJl$}kw#yq^nf9@6!xK|qjL7^_
zdW3ZRu8&WO{~>^a?=Tbq9*yCyIokQ5^eR%Hnr<Vj^X;s-Yiw*Rompeq8RH(3riLyZ
z@_q7s@iS@4OO*-=_ETIL39(}<KilV)u8(A#pjhJZS-7FvQYBlarDL-qn#Mq!zu;%6
z)iKwLVk^<lBsF<D-yKo8DE3)=+h@?^gyWTHO1Vppk%d3a*!k{h(|#&QnY2)mA(+g2
zkv86t`=NZMh&DC+8`7|(%SXyDJ)hTC27OHx1Tox^oQ6S2NAg)uJrJO;?wFI;eYhDG
zkk=P@zrm&txJ|)RNKS{ysQ#q6;lM5CXw9s;sM*g`cu4sVb;H^0)B~d$vktj{{Y&Bn
zT&eo@iC%)uCQg8#SX!ETV=@s16767h9Rx+Hk$NlR?yIu{mEE#5x{=li6RkVOKfJrh
z(@Wc+<WdK0`0(;|i29nA`Wzh*b%E2(d#9zsl}{1nDSEnc`{0;nL>`<b?PnVj`u}3C
z5}4N<x0f9)0MF8e0NFuI;jOyE>QL@Tq-_Fo6)q_G2;L%1Bu*#)8h@TqF!VBGG<n46
z-H5B`u)Dc_M5%N*WNr8P2smgclb+)+_k1~{Uj965-X(&q%y5OD^a!&ITQl}dJOa9s
zubBLyr*NA)YZAJqn3D>p($r5UH6@;^9dAiV(`|@+!$9h~leB-6+>K%;NhHw>z;FP;
z1hX{iVDQf=v02Qb+~_~TWu)9>B`r!mK)t(36IvfctKM-Sl@Z)KA)Tv<k$U6f@2BOQ
zvH%i*xQQWBc-@yh#sTOSM~x;t*8KN4r7FZ>77{1EI-D1FOwagSlt3DWA~`Wfd%|=6
z#J=5f11WDMucs91(QOW~I_hWR!gFE~s^)SH`NNStfTMj8LfNeL`$YdJRnZ%`NHr44
zN}!1wtm#e9wm0k`B!>38Liumh3`mt!WynN+kNbm1qL-2r-^bZK!8L8t`0hM90Vz-z
zI+UEeD))z~@^^^I_aITFo}9FnQk{SKh{@vjRWIPrALWca0pYjE&gtCQnyQ8t$w^37
zpeDq2&aIC5P_0V-d}SQEE^}(Hj%(K7%!!;o)g&JOO}4a^f;ypB-T^5r)4KY-UHEJC
zBre*d+~4Nbg_g^wQsZ*8JEj!zcUi!IH7z?J^q>AW-d*H~z3xY_`xxLS5oSu4`rXfe
zBkbr{<Us#X6kPo;diw}9fsW4S{~8q>@R@UUo5D$d-y)n8!oW;9hkrSAFEgMfgu~=$
zW5<<$*Sd(5*o6amcY}FnS-i&1ETZv!rxq11)E6#N`R(Bj2vwYURaIN<*2op}w7oh$
z4d~!I>}%u@(nZ5P;{5dc2fDc7%miQ9n5b&4V6@u^j50}c$<wGkshEq)e_4qefA>&^
zH<SCf%@N@6ug!4|m*Eb=uv7%`98fh)3*nJ@qld@A$%*)gK-H0-3>8bH&G^A$3JpTY
z*}x_aG^U`l!bSD2xpSt4=ps0rLR6kE(*vVE@!a|edzfeJ3~ZP0?Ze>+Wrr#b;ndaD
z<^Ew?>IRBS+rvbR=KTl5`Taj&IQXQ$(GpyaMw(sNc0F}>^@5zN1}V#ir>su;Hdgff
zj$yWv8!<WBE-sLuBGEtcN5N6TxY|0e_`T;nPW0`WkvLjaab!R^ZSP;zP<XP#43%^L
zzNlJcnfFTO+<aAHZF~;3dif2{e=ph(ULuXw?|rN65KcjEE;hlh$m2}@<N^MaznoA&
z#4tnsqj;Gz{x4w?;6RMgK|xbFxeJwf=7VZ9ZC_5Ro&G4HII#Ejq<=vDh0$hNb}u8P
zdWfTymn%>DtgOhoJX!4Nzma4C>pirDSbXn#crTb6om+#_>6k6aeZ4m@?bMeA{;lsz
z$V|V)_4Z%g|C=yX4P-3gSnZ7w%D%PM8HVYGxSD*{&wF_kp*~E<H@$o+Bql&#vio_~
zi-NZn+OIkVr@fH%@q<tgU@gRPIzvqeE(_9&1>9?p{7Z!z%$AvCw!i76o;sDIT0_B6
zscq(s_mIZ$Hw1wUYE=AdkG%vq8w{q4s*Ku&h&5D`fp#^-0?(^^nQCb7c_qsTrZL+J
zjJ0ggL)&CBye5|+aecX~;mYmq`_>$FQpPAY;$w@lsY4wlG1Li<`cLjIHfcQ=xGM>b
z<G=g=f|vgwmNZU6GAxMb@hj>+`KptOI`|KBYo&_1@}nHZOL5uk33s&`4mkgY_w1%C
zx(^d4L-PB1@K&}`tIJ`w7JNB+eOp!9-=urHG?fBuf_v#45<(`+eL6|A)(|PIDpY&h
zOT5fgC!Lw?=tISkr+0Q|AC)wGiHcw!BofN>a2e)?G@KAL7FMk}Ev-lB&nng38;L#+
zcV^swbze&6oQm=!ESDfZDd#$$_ZB0h@+WDAk$FVI?l3xaMr_}ze~II8H8U+JB#C?R
z&GSCtW?RXR^~dIB=cql-;eO@#dl|YTaFwAC!E~2*pQw}I5CNUl*pX``orf)GPu}zh
z{a8|cBw?5L3s^5iYhu<2KmRY6o69Tf7!L8}zp&}dboC-diHO8|9JtWFKp0qxnwpf}
zI5M7D#LBSG>}28p6U|pI_oJKdbG)oq4VEgQ51-x2yH)Nv`XLFOM@M7h0}yh}>+I}o
z)$XUj*<b5_c?20VB)Uv$`HhE9kcB_{t}hzOevWZ(V}PoS&_NWDzl*0zPPxR1hr`Vg
zy!y7Mh++WYy0dvm-$6>Udb%<Q1M^R^dJuPxgzhn6Hnnmn7Y&V2ZLuXO<KNnyF?=}1
z*jG+---lgrJMkZ`F|)8^XiY<-W-IyWg=$&Ixs;7Q#prChp#op%?_SCk`Sz3da{F1i
zW~YCv-d=0ytxhZSA30lY{%({dzvuMx!S}dqY0C#yy@EfMgjd&m+5HI%WV9y1G7t6*
z;1W&$ku#gR8CI}HxLIwRc*YWLIMmdhSABT8{8jtb`|E$c`~qsU*KluV?cF8ZuQM0!
zM!$B=v%z%T+|nXTG4a~lc~2dAglf5-gu4DICu-vn#-+a6yobQmcUJ(NrgH$5PKshk
zav)`@k-l}P6e0Um0)KT|U%H0>-_J$I!BEAk8!C5qU*yZcY7q!&pF4lHe4)t4HWY<o
zJJamKH9RCv$d2ds_c;+s!9tl$CCH2aZ;+8c^%~i&qz^a)KSV7ZY5*iumRYwCIE#^Z
zz%FD{Ua2{JAy7?eyw%6aCD?(gS@k2GP;J~)jvSB9W$A0qfN^Ht!t$mR%v{LUIue+A
z-smh0>DwK+8kZQ;KX^Axiy_9>QiRPj57W`M8P7~!iV|kq`I3t5ASI<w!*Ze4(tY>$
zjqFS!!gB{*Lf&)`6XEXMqz&p4X`B%xe9BBZoSd9f(}4)`AG%Ic03HM<zF)~rCaWrY
zPZN2r)q>Hlg=&LHg7qIgY&xBN?8LD@loG-O!MIDd02@FgG<R8yW^We}0l`T`NfgBy
z2-tW-kXcQ(-9!L<;6>Q4LAz}Yn0Ukm=)1qwZj<QGh|MLU(_o1i^Db8MZ!H9kyz2RW
z^r3}O-v_RpwbX%ko913pbgYY4#}yPiaWpTCc``M-{u+uBDbpbMKm;;R@_)thAxjHI
z{AXfUqI=q5|8*lSpS1tfI)24CBYN@jU$qwD(;gD9=UfO_o-osa97L@}$`XK%3P=JF
zYn+MUO;=XC9qgjqL^T4+H?c<{iW5yH8caKneeaI|iJk3NLkLzP4O(sObDbJ8`I$P>
zb-wQ0wJ`zQ6|_jCl$wkxV-a)TG-#_r-PSL$2Z&S;!=~RO?J1C_(=y;I08CLA)TXNP
z1Nl!h?0?VF7jPVYXAKo79FXJu{XCEv+2ez%OyUD}J~W2&1{zeP%)tHtV<Qx$f3Wmb
zgfVX-3jn;ykVih57x-*;M70nZO2;j`BG4r0*#LJa+6?qkh9Q^HRaZk!{*eUy6xAfl
zdf<}uc!!(&)11c9!q-#A-)vCqT&1JD>=^feY3>KeU1)oe50_KEyHTd5OyF#O(0)l|
zlegvjdU$I^Dt|`alj8xUXD;)KRJ9(SAyee8y@HW`u+EpK(=Tw6@4bQ>XG&yFm}R86
zM2fL>L`bWoc<pe+a`v#sI&G`zTrZ=(O^jc<|3ZIf^xzlu;%&|zKgDg+I%12+Zm`(p
zIdL1peL|3hKPHP<@8$xmUh%a}Mt@urFwCYdY~ir$)|>ZyUKEk+`IN1nHkP4E(f~NH
z`=(n}#5tR`vB4tVm*noLU)QzGTm4292IsdE<w|<~yv&8Q@Ul5UakT5a{OQB2x4@u|
zYCOuhZjdGOe(x>7LP}dY%3SkWn^hos@zAK;(79tWuW}(bc&*vifRKF{)&fTlpRG{c
zZNcWZuh&qr2Wzcl37b!9&>Z1XzjI?_fQSA_ss%}p*>c7iarsW`rtNSs`PUFeN-yCg
zMX_ni#6aSx3XR<LmSV(hsurPqgPI@RhYd&z&6<hhCb3CpELTp*KYaMwiRCrX?Vstt
zg?J<;Je<(4N5jiATTY3z20Hm%wkI)rcHPJQ0}9I{Gr~|Wr&Z$@h0DR=U$1H)D9Dic
z31WMj$6^Nhp<@aBy*w--9bh&|qHl{OcwOn|5|_Zn=HGY3NU5#gV6|$M_JF%6OfNwc
zGKG`89&D4c+YdCkW>PvKp7r-5c{?0dzET#x9XLkLRXgQ}_>tEe6cjB(5I+fRSF>iv
z)f&yt3$~M5(2xv9n&j}xEXaASu2Py2+fm8wC*2d$IXU5YiV9VHVq@#VduUKwTYf!M
zsA7IjlHegtPVF(Y>z_S3ke)D<Y6z6{Cc$F-Ni|SD*!hTFU?~a#J!uHlP?K1%yvoRx
zY0EN0sK-R8-eCmpsx{)|l~-yKJr8xg?vhqY%%LbSmHWxJTVdVZ5QPcp1PSFX2Tu3%
z{bbOPCS<4AOBVTaUI$)8*j+y&*1J;b{VKeSMjK<Mm=&?B2G#|@p_E%TEo+I_Jb*c=
z44W;&Vju3i(H&LAj?;m!_ravlRyU=>tN<@@F(2*J*48$XMOJg^HBB_-95k*gzq}yC
z#islzB;W--e!~(S^A^|yl9n^(VdHO`BrWH-grKpekF5FJ@0AkBY1W7XZgn4>pC2A6
z4*dpcOu?^SeEbXGV<zxK!CUj%lm1VitR-&m-q)Xl(c{k}7QxUeJt8{Z)$(l6P?aDg
z>8ABfOPb|*2+f=~B~BqnGZSt#&%IEQP34_K<=>A)pGR%<$8S%z#I%HjK*RnO_tJMx
zkEFM2bkxSC6S#lcj6}EKSYr$5*p(Ie8xRTRLrD0;ammb-(IYh9*_X2WipRO>Eh=v0
zZl1$|eG~87`2HOaw%AidyHP4bI-K@nl9Gh%F6*x^L6)unX;<ot9=`v;y%=NeV4)9`
zd=HULQR)ILS4v{5JWx@;5dJ%-wSI|S*r^zF;Ihe8eI<gC)W%ZlbXtX=`yhxQy$<RL
z`|9?IpHwd+a!ApSkTDKdb^ebr2GT`8cql8#_JLaGCPd&9wN#X>GqOk6o_kgxac;Cc
z5{pnbSnQ8QkW*KQDhdfV<#LuZM$VT@;D-gVo~K4~5*Nr6|3jet^ZcDRP<RiAO{>CS
z?1Au`DQ0<c`f!1NZ|8usA3o7VkFh-LtA;2Yw#qoLxzQQKqiQUL_<YG*8D}?!`uiQ{
zpZG5zq9;=Bsf~&<$!(~3DO!yDuUHz0)(mW4%<T(hch0-UNI_L02^4k7+qH5>Hf=)|
zbkG59+t}K>poc|(4X|1_+9Uq#4eJY$-chkN%aguSwxS^g{)niMG$F(>C@pHa|Bill
zbl()6rh7ZYM67R^9{2^X@SU1)HA0_$Ow_*-_{QxV+>no;(S~*c!pnt*vN9EyVM2wA
zE*nBQ_6R?~On*5~oNiBiV(+9z?xuwrlE1<w?wtK9J-Zi1R8VmLAO7+jT_dOn!OZva
z<-g$@sDgqUqf2wD*WAH0JFB67{@;bsaOvNby3<i2vDn|*#=pN2BU$qo%d0IrAg{f*
z`*->Q5bOWZFC)4HgB?U>|EW!hQC;v=Wg<yQ<dF&0!+T;R@<>2lnErYYgBGjo5!_^P
zr+G!pl&Qf^Pgz=hZzERIN6QXNRN}YL2s&kTE?Lcv*JLX2PKfoLu9A(P(HeIb96I`>
zH@Vne%=^!C8%>0T>4Tre|Gz$RJaeSr>As?T10M(JquPVC2#Q_|Ks3ynrHC}lq4~B2
zgd_#LEk7>J=D&(F_ufKb+s+w@1ABpRxozAz5BU9Vwgi)>Nl7Jx>=^4_rF22tDx8tz
z>YioSR2AE(N5&LNVarG_9m>^ROg|CYn{Ip%ZD7WyiHtH6Xg$>=;ZWQ#eQGy!TRoRR
zH#-Uw`6H7kptGp5i3h30fskBidfuONGM;o)e(e(Aspigv@Y&|}0GsrsmnY)CurjOE
zJ|f^~UUW_aSnNxS8IC;`efG<*Y7Djj#@oGtxS)>kFW)EvJ(1Hj0?-q%04rA&IXVBg
zTo#kc2qU>?0>n0}mWui|Su!WVv{f6in3)0<ag$CcKpVnM>pd*hSb)}Q3s0KPTml^q
zRGjiAfgzc-BQ=))6F=|1djAEGoLB-_cje(@X9>|a%w_CbRdiJ#`l`u>{{2!;hJBqR
z7*Q^32z{sA9KDN-vS69#Fs)Ri5un>LZ2Nj=Zx~<`KlVG#wPs2&6gxRGt*SO{QQR<#
z>#Bl+<|3{8Tz7Bgrve6|;cwb=9Ssg#v#UF?2mkC#sTb(VCH|^To^()ZKDrK(%3Hdq
z`iQ=bVeTpDb(_b&LT@L$H4zIx1Q3-U!@6b<MKp-y9{`#6rD5o=<74b1>sRI05a}9V
zX2l&R=VEHUl$K!mtGHse#WO3>bYt)xbWMR7mRg>x>9DF@)bB!I^zHRU3-FwzfzOH9
zu3Li<9YP(YuD=KngAYZmh#>L0b6lDsQ{Y4$ci+buyOEnLf=SB=#W})ZZOrBmqm?e+
z$t2vrZV+sD7E9Uv2|e6CX>QF0uf(`tAX_fQmX}b;b{i(0Dnl6I9IW54CKq!|Y-{d^
zUZdUufuEbxn>^&pIG!vs@Uy)`B@UZw08ylz&YjuBA$J6?tL_a@9nN3X5M<=HobJvK
zlDlyCg7xE#_X3;xv1$POzP|w?2Hs_}mE<q0Kp^<cytNG3!5M;(E+rm*Q0k{p!lO{N
znVISOAU2#6`3l_D_|iAiAGNP%P5aPOu!|v-#I9*XJt=>A!eIks;vMXu>-?0JZhANm
zV`p%X^#fTHmofV1N>k?2*7{5lW>wDWAcZ)ohLu+Lx-jW-X>kQYzOyv4)~p@m(R~rx
z$wP8o$(5cWtr!HG2e-;F4s(ZTPL@0|w)e@gR2!RbAb7wR6?yl-feXx{w}s^M8+0ND
z7pyIzvvs9<)G-0E@=QO$!@W9ES}t5n_tJ@f!dzl=bH>qhadlu3t)I<XSa}0_)SYTH
z);f3VRnKdF_p7j<qNL=x<CpQS%A0(|=dGHL^O9I^ufvkd>*;5TBPTMbGm6~i9NIA*
zxt8AX^M2z`fDT!OiHPUTzhhwPdll)se7i5mR&MmrJE;c1m;t9A=0@Z>p{+*SY-YTC
z0if3fe00;kl}c_2W|ERmQkVLmM^}kRFNLAQ?Z(74vqIT*J3;fQ+m2OQ4_yDaSgl?w
z^RHG61#3>DF^MhyXqn{r1k2CANMsCe=0q>BR3fU^OGri?c#HW~z{k2<CvKFdy=x7F
zWLWc|dY;p=F0|&Jm0h1u(&Yn9xxt2?1ut+k1Us<8{j>tBq7Ue?SwLIOFt5Ub!`c*j
zeq4D4>J2@!3!quRm>EerYTmqH=xETyg~btESFidF`eLI#Nx+`}&b$UT#`%LkK@{L>
z4*qq3TB%#olwnL&tKOHTL1s|U!iIQKIg&Jp5F)r&#0Yq6#Br+%5v_nhb=9tako?F$
z%Rk(+I@12gHwS_GM7-s&d*L_S1H{Ye;`sUCdiw2+5rISEJ*~)+f1P@X1^{5W)6cFH
zTr-eKns0)_MGqS%W?bHqNk6m?IFJ_Kse|)Le74x)2QSMROgu|185E0h<as}qtum7T
z$md3-x%=G+MvRfI07EBC#2NJP*u2yTjeVtG?B(dzjIG6&k+6f^T|O5VNyVd?q&)-}
zs2|C1jps(*Ev>>v7nv!%vt4UHX6`(u&WLcLOl-#)rZzu}*Wdh(l5RLo!N@~7W(6-&
z@?oye<(zzZ-%}BW7l86z&buw5x~}r|o}S0SMZq5hk06UeM80ab;1-Kox?syMkmZv%
z_rYR<=K9IL2ag~Gak?}($(q)(7kP0xEM65>k9#Q_G;ydijfua;xiwhjYboSHx%Jz9
z`qDSUqhqBDhI}YGPEAtIDQe8oA6DF;JLIk^{|JxaLma_*fvKz-7UxF<$;F-NW$9(b
zg_Viovrt)j(3_FJ^*B<T32|J2SDQ59lffn;*Sx(ovt20YUx;VT{CpA~RvzM9oX%K=
z(}2j3Jq&~mu93^6TkrXiladlO3tuQ}#7R5F@D-1rzkW3{pCvyK9N+>&hRg2p@o$ce
zGo(*j7vgDxZK6Qv$myeY8-tXV2gL>26Ir#_V&hDF*J2aw<1R7<z|9D>gNjrb9tVYa
zqfo-LJGS(T%Kd6Bk;>Y2N`lf+jN?3;b*2tAD4LezN}_|@6WR9!ipQSa6{54}((#uS
zs4}~e9#B?za0cpfimIv;m9Ll3Oi25y+m{tzdh~q7N?pg~%b(1EfLhL<_pg`P@+~0y
zIb1=i$$<*t*zhh~-vYFZ7?IezLX^Vzc!uF63f+!4pYXf@fhr}g>H(H_lw-tKu(Qz*
z-nBJ<-Snk36*+#yyms_#lkchlzy{0%vBn#~+6=9^b+Ez6AyKNgVeOPMz4h4(>g;3d
zLegR-l}(sDw+C}**}|Da2so(+qtsX^u@nE%r5|`COVai>&6O(J>jCv<a#YxcrVwX+
zROv59k3_FnWu{>?yU#gxnoLhLw>G<l-{HkJ)-8im7yuIGFSekg)6-er1Z#|zv~8xh
zaFqP0C_Fr<D1KeQeq1gp%Q&~A(q3iZc_7(W$w!oaVyw}co0x$VlA=ylho)cyR0t8S
z^aPqZ=Y;GB(X!!*mAOLgBBxx3z_tZ{NuKYTe1qBOwxprq;IfUO@~3H!=oe$vHGj+b
zW=0RnkvEBy;iTiY+x9G}sbI&c-#Y16x?Fqr6=&^qJV~z@S)O9rcdVGG=m;srkwTT1
zIJjZ&S*68xbdXIZq#9v$Co^YOzkUByY8wg(A?U{wHHG|#{8h~+l-C`%XSNGGf%n-}
zJnhG#*Y6t~E_!A08B_3}81wZg@PM{7k>7GK)UtQ8?P0!eZLRLAZjITxB66!ADP*+K
zGkBw2C*0M&PbE^0V1*UO*rQfoCAE|Z1ZDZKI`ZnG5;@cbN{^9od}F3uc>ivM*(Zal
z<%`sMdlmn1I{U&c_j)$6ye`UHU43hzx{>-B#dDtPPs__b$GP}#ttON_mE5#j2rrD<
zG>rLW_;mBYyFq$Rl=`D2>V$N6Y?IuGs5)hmWsrlF;tcU0cK8@{rh@7tnMdgRkI=R-
zT;W_jD4{NxA`;KyM;u+towOL|f;%3YFs+ZjMn~_M^IBZeCW=g6c7kwoE`?>(LHv5!
zpl0Lx1S4KD!Bv-l(tBf3Ia0-JL1@&uO~qG;^jfTw%~YOzQ$;#=W!QvCoR_9Lt|QR1
zC78$;gsYMYA28s>PpyikU6WHK8M}@(7trS6=M-2ye1i;MNZoT$TQ+U!Ui_KHttHhO
z29LN$x4B#9{BiCY=*I>bki}l=3K)aM7?}1zn`xvkU<%O*S&tDBFd8jB6@XP&qhI9p
zFEnZ#QO>OjEL=CRkFZ=Alroc*I=nrv7L|tIIB_@7$4-ZKq+!D6DBXkb?l8?Xia_kG
zeu!2iA5YJGMW{#)H7clDb`<?ho8F%*VhXI&wwo*IF%k>U(PmzE0);;d_jpv2um4c*
zTWPB*RLzSb+Nh^NP3dl<=~jKZd4g9*I0gD7+LXgG6uZfS1VxXN63Ht}G-4yAOb%}C
zBw$1cF#ktlhqN-znyVaX8JXps9&tZ^mhjD>hiie$fL@f0>>0<*xyq)TV>ZPWel41H
zCAgE!=+!4CmrZ<6+VDQcdEW8q;^8ygnD6qky=Q18&DGsnwApVY4w2kjeYZgrKqYcl
zkoh)8WTNGK=<S#d&Fw|)I@8Sr65K7<TnFzqR0d8x&j}3gXnK99ADVZ}B&6_@Qr>j|
zy!0WAP{yjmobO42_MX<+v5Y~1iL))gao2?JF+@0(Zj2HXxipF4FVI<~28kw#R^6Z5
zl8T7hC{26h{7fNBeyC)rESm9b66JI;c~d^6n0c@&^^BW}wNhbPL*|kXqw#{T<HkEi
zFNaa_>56ELqhba!Lr3x#RK|UkL|-?}-k$a?%ixH#v~A;HC`Q*BM2(}F=rF&E<2SCC
zy2yLoyf0*gy0kfef3B6&tqA8ur_@@iTo`Da&0SI{ss9+oIh^ZPe|b$ZI!>`@%do_B
zf=s+-tguJQLA(Kb7;nq&Pj}Xn_ESheMYfLkMf^nVkV^j_(LIHY3Wiru{@AqJ&LY&_
zoJnJ3;@^u5qrgSh&_wNgo!kM@=XMibZv^;#64pg)Cfc;pGG@)ObxNeFTW4JzxSyg}
zF<S4~LO;GjH__h@x%xSWBY$l?Mg6td#Q1<FZW2KuOXNrU9=(d9n^ubI4V!bbYUdAn
z-%tAF7ey2Y8f8(NOL=9ODi<6{gNQ^HF9+FW6BJ8JVOCmtZ?Afv`V%qDC6N3#JVbFB
zQTM13J7{IBs{R}kt|vi@xu`^twB#eDo@gk+V=sy=BwG_q6WknZWo2w|CH~z&PS!L4
zy3sj?_jBRUq($FL7BBUo#-zv#-qQ@(Ah)G(SQ(g3IK$&(-Yd;r^DDD&!!OU0X=Oe9
zSs<(8Rs)YBk)6CawY|mY-P4<TKerzsA`IM8+-CEtWkYfkxsdGKOy^LHK?kPXEC^Z@
zIJ77b<K$YY+oGG>JCUzMDjq@L@0j@geQm1T^@CH~>Nm%ZD612!wy}~aHA;=Uq!?by
zIo)>h<Jy2EA6i!KZUY3uNwM*^j$o%mYjfmE@R#jb#Wt?O`c_@(^rW=SRLVK?-gQ7Q
zjX5~A6i7z-Z&@u)sXx-caXcmVnSc6Xps!CGC=d>#C4zP^0C*}+K>-(^`>8+0B|3%`
zNjJGRaB?)=i@^KsP@ax_+qL79X@UukN4_8#B(bzQFO5{><YD`^>rUgHZ^Jh}y-yfZ
zdMKbzzV`F|?3Em5)YVObjQRSTtUr204!va55B(LVjgg5zpg*U6LO~}@ilIc%jPZgu
zMRK&C%RL8K((5QzgyF(1n#N4bvFu+smmL4v=VH$5+JKGz_m%nbg56T+5y%J49@&ps
zDFol6&IxF0ms&dwIcU-l<`891W0ll+r8$6H*G#QhT%76;@wDmnB~dSZr;VDdWLZ4K
z#*;@!H|2XX>C&qD@Q#>OTcM1<-(2ITEhq>eD>7MmwX99IPetRmlxSL)xlg0DxX={}
zQ!<U)C+%Ohoi*ONalc_J^F3yB%!VSMbN#&G1bd@^?zmUk1&6}XSg-DFXNOPCzeyaF
z#8G8fVP7KSgQ&=yzRd=Kz}B+#rrEgpPyr+FdKG-Vwa>}F!Y~VvsRpK=t7|J{&vUT)
zoe{HmC`ES8nsyZ<&D`15ux2O!6<6SDlA*q%N5oQkCRV{7bO)xxlZ+?sK2U!&_D8x_
z8vPdv;Q0Zd01oQahds4DS)w&3F){v0;&0FqqbyGNr|G_p(%x@78-&g%A)t1Q!0Hag
zkf6rbhVveZs?#*vVo2LiC51C)G6ACWh0QIS=z~tVh0ok&Y`in$Ov3HgmyC%Y=&_o5
zPbCbhG6!PUvgK$U{0$mX<>y<dWk={J#FIvFFe3_#M40>HC$0{m73&oWJVk7+BS?u-
zB`&y?vu2`|8Svt^*M;BvoL6Fai?*8R{$)wcVs8Ci;QT=f$1Q96#cYwP`1BfcOH&lX
zO~31<Jh7zS7!1>j0G2rU_BdxO8-rZcyILJv9;NZ{B*Rp@GJ?n16g%=L09)Y>p+IEg
z6@W$dU*y5Q0-oS&!BQcy%D7`qkM}D{|10g`?H)A$d&v}l?1XX0WVQbPm-im(|9@cr
zEm(fqHl68xZp^TK-N*6dCC@>PLK(7ZPJN;T`p0fzP1={}>95_6@$ge+4=tBB2or6t
zI2QAkB_VWc^x?g4XT8~vCzvr_c0Yfl;P!@H>2OC`iLbrNdU$r->Sq0V=vZn=n@wwW
zs|~Z`B>y}p0#bz@xTt-Nc<CQ<a(}t{lggPyi`Mk=U3Dy~EI}=;7-Ax#_8?}}RD+ge
zB=v4b(n$b#FL9lq;CT!+viKb1uGaLz@#Jq)$v+VY;)f@BQF1v3#3^#`ebk@TaVkO+
zHnsD+yvp+XTriQVlu+1IiEXWGSN6n5^lUcQPsUw0fJ9OO;EK<Z)aAHp_nHbwf94yV
z;Jwf{9{7J2T{xdJq{Lldfus#Ryag{{c(?gn;4^SPNx8J+4G3CBwVs}ylTgykcE4cr
z?N&9oseZ2Xr8Gr2wNTAXC>cPcIU5AF!=>*EjUp&*kQkL}*9>9>wc91zm#1ZQ<nlf)
z$BnE*0k<i#RPRSDZrjJeXI8%<LtcKc2I?Wz0}6Wkxe}h21@Eam?L}F->=gf<O2Cq4
z3B;<~p|+p}rbqb^5Vfa+=$xc2Xev$tyLNU4Y!EYH5Z&-xhx&j!uGtsKYu%IwZ*~LJ
zk{Tt$uWNy*cNW{D%65d)(PuY*KSAy?<u(Y|7nAA*oXJWek1u(Q;a~lA#!-W!nf)Xs
ze|My<i63UBw=QL5Xh<cHLE*HTf6g<mrLmIR_QAS?4*C1riUlN8akV*Dr~Ys>2xSC?
z;&7@F&nVzg1V0H9t^iAnPUl1K-5Iy%8w7ga$qK20CEX+>(F<!x@$_U@DMCGglG9P+
zE9<e+(?FTZ)hMwE6)yhy^I=Ux!*VhhxGgrCSNTvNUZB<y2)hq(>4VM*!X;1_yx5xR
z(u7d*+${xwf0Ll$M5=E7X6$?*;u1x|r2GJ=Ad9~djqfFJKvxBG(S}6Z$#yomnQ9OK
z>)hH}H!T9KOXJD*Oi=c`X=(_rB3!@?lJmV4I(C8c@g~TszigTHeCmi*mETU1PqA2|
zeeGs7VJ&>a?AN{8MTiULE~MH7AzEE<$x%LO&R-7T{nMH#bJ=z%P|5CIsay9+3Lb1a
zM5&Dd@$+)=@)z)Z7kNPo_&v}n?Rz6_zH5LTgh}+K%{H*LKLX=!tz`5ZwkwxonT!dD
z-I9?0Hf(f)>ec6@1IstKz4-_!_jMh=#T+0Q`9&C<@Tb4|4Z-!|WTap0P7(|8m*txx
zdOX#UU3I7&{4Uq%$<X#7qKF0$Q(4Y8=G!pw@87DqCAd13*M}4oc^}K4tN~Vh$E{lE
z*;h3vT`dH?;w~@<O)UXFP_L~`u4!NIa#EY_^-_p1Pf`$jNJ1PvU9zmv^+D_TCd_SJ
zY>N-5OIZO0&li82Lqb(+B-gmBxb%xWUztla>;X2{^PUdgvE5o7*|gH%E-y+x?iycM
z53;bvoCet6gz^LZ*d)142clv6v+_?8ZBJ19Yx1029V3G}`+{ZCa+09@*R4gL!^g>f
zUCdl^gxV@UpP<S#r*blBMloU#ylojz#r5LnKtcOI#4Oco#)CGp!(u^yo50^>`@uGO
zksaVaS?!_tt3-aTP0>P!W^SA-HB3S2V&6e4rZl@~y08YwJS2|S$40<0w91CwfM0wu
zfgA`}sp-Q-#ZVEQf|Bc!{kUkNpp4fJ2m@yFBYY-OR*gx%Re4dqLm0=|lB2htYn`a2
zjkNkgG?_%#P4E?ai_e&Ng}z)4ZyAJsRxYA+3lrRmrQUKeE2@7+gMPwhQKv#fJ@hdu
z(odx*SDpmU#_Xy<%PSivA23X^EizCp_3TA-nC_QowM=j#PiWCIRkN&A9aOEkp7aZv
z3T#+|I?Kl3=jTT@3qXu$Xolr{8&{X@@eu<SdDjL8_-*gw+e#<J@%No1|8q4{E%;%7
z>$u7@kh!E$T_svHw*^J9uWhoX81I!2QYA#|ixVSLNz2hTEXkxh9N{82H+^^MOjy9B
zoh#sGV>ew3F<b{&o&K8Vt(5dwkB(_Zcx9UJpdQ^CbL&WyrQFwV;z7VgU1v$tx!}o*
zeS5X``?;kQT`fAY6KS#0S8-zolvVO(+hhn+VxL?mTC|au`IOucx{)=D*VY&-hEGpS
z<n9_^E}^Ak_kCD{IMMC&Hp`a}YXVy<m(T0$Bb^FLqVVXSRf|JA9tk6Elj?ydwGs;*
zod-%P?3wI5Q*s>n;c4Qml`|yu9VNDT$Io8<WJ|X+`r|TJs`7pP_!(7`0_qrrI<0e_
z!{YK%{KwZ@pF9nOY1@-Wy{#9r49AXfZ?SbQ$OnUVd%S+&Q&cdirrlFvOkG~$Y09v%
zA~!)s!TSndc>R-ng4?#PW@y{l&a#jNrTNvflKVc102TQb_fLiw>Y-C-?Ry_phkr*d
zP+~c3y!eBjRMxy)#8-!X?fU9Vgg4?+EeuBA4C{MtpV9~-c(&{Aa|J(f84W6q_K6>C
zI_cw@$9uJ0XUs3EDG#X<ls%Ti-juoJpVJSRTvrdIyVU1yBWp3!%NJ8XsG<LU__KTl
zUs`QiOn>1>BRR)FoSfzYx$wo*VKez38*d7SVU*fX$X&7@kUBe^P9;~uqJVuSeXEan
zLt)``1*^Y1PKV!&tn&xA{s}%;zmz`hETtcdbo*l6+r*7CYT6$0^xIP>-V98Jz2Blu
z`sF)x?EMo2k%{!Rvk0a0<3$WOh}J=2a8|1qt#)Ug)S%r`Mn<N(wKLavq6(xvMh`|S
z#`>T?UdyrhrO#4@IL#lbKq}4driLv~P@dzPD*#8Uu)(<Hg0HCI_-fKXuNL+A6DEh0
zz|x!#wGd-S-12CHrja`wkYt*50*2aVo9{)mh8v=0p0OAZKM4de&Gp>a?eD*Wzu&sm
zGnD*5R>y`L-xZjhKR>4r6Gi}i3lLR>-E_awmQxGv<QNpVWgC9JvVTFzpaLiuZZX||
z+3lZY5n<Q-`orUGc9}4OXD9xyx^Q|zH?@Wkwy)1)SIF>_IIv38tgNhBDHlS>PNct<
zWfXTG=+>bBLb(_se<0yv2YwlX_Y6BEb1`OE@R$fedpRld`FAKKjw5j-0ip`%9D|`0
z+lgLz>>;@ro$mv`2JBp|;K-Z5IG`0pvUD^h#C=4n)&J^wf|2;Hpj7hG_npaUepMe?
zyYF64K$$kJYR@&s3W0FslS4TI_G{*dY$R_!_NO3&O_kZnHl_lFh{x+<y(!~Kc98k!
zg4^15YVM+`wAYICz5~gFA$szI!}I3(oV$Fq*qWPmzed0Rrf#;-{GmZ<Mj~+0HUaQ)
zfAJ0RXQaQy2bLIO6W<Wti8s7Y&_zT-vexZ4=H~w1@bXrOo8^SNo0~Br?nY8s(e`pp
z#aLhH9G{Z%FPi28bn5&3{PvMiCf||o<|T9dk@c*e?va}F!owpYRVMehX^1!&?lJks
z_A8gDkusxy%{wF;J4^(+Ga2s%tbyW!?^3Y-StcdTUQRme3vX;?leN?zH%tqc9Tw2Y
zj!*cD4CpvY7f@q?`y3~24nM;@Z~W!^Ct19no%ksa?ix#_*tiGQzl)T6Gaf(h_4CbS
zhd2v|d}PDhNt>Y#{(bVl8;<yN(Y#i_qdAS_*S0M^cxkn%uoVHGCSAF85C{L_WBkIF
z`){ditKx*&%6_fn)INX!N&Z4UoD*@#jN^@@&&6bY$-)gN)KUCeGH5ynMuip+95X5S
zF=wiz>=X8G@ZZnD-voyU8yR}hSME-<2UMvI$Dg2QJi;xtOD+4)M}byD^35jSGTqc;
z>3;yH{*ENt>?onEV)}Aj^S*O-FGHh}9>B$;XY6JiaN8aEzk7hcC92`P2nd|8pWhR+
z{SV?r{`x`TYW!uf{EK@xzW?hC!H|Ls+4S);?>oKzjh|T339vFHr%mwoe`$9d&ehZq
pXczQyOYJ)z{#{+r^~}o240qMf>qjsDIspHPib&p07k=RKe*oMoH0b~U

literal 0
HcmV?d00001


From fd7a63167e1f2de50b6e13637518834aabc819d8 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Tue, 31 Mar 2026 13:17:31 -0400
Subject: [PATCH 15/22] Wordsmithing

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 22 ++++++++++----------
 1 file changed, 11 insertions(+), 11 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 9fe4aa3a..efe25f41 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -46,7 +46,7 @@ class="img-responsive"
 alt="Performance over time"
 />
 
-**Figure 1**: Average and median normalized query execution times for ClickBench queries for DataFusion 51.0.0 compared to previous releases.
+**Figure 1**: Average and median normalized query execution times for ClickBench queries for DataFusion 53.0.0 compared to previous releases.
 Query times are normalized using the ClickBench definition. See the
 [DataFusion Benchmarking Page](https://alamb.github.io/datafusion-benchmarking/)
 for more details.
@@ -65,15 +65,15 @@ matching row groups are skipped entirely.
 
 <figure>
 <img src="/blog/images/limit-pruning/pruning-pipeline.svg" width="80%" alt="Pruning pipeline with limit pruning highlighted"/>
-<figcaption><b>Figure 1</b>: Limit pruning is inserted between row group and page index pruning.</figcaption>
+<figcaption><b>Figure 2</b>: Limit pruning is inserted between row group and page index pruning.</figcaption>
 </figure>
 
 Thanks to [@xudong963] for implementing this feature. Related PRs: [#18868]
 
 ### Improved Filter Pushdown
 
-DataFusion 53 pushes filters down through more join types, and through `UnionExec`,
-as well as has expanded support for pushing down [dynamic filters]. More
+DataFusion 53 pushes filters down through more join types and through `UnionExec`,
+and expands support for pushing down [dynamic filters]. More
 pushdown means fewer rows flow into joins, repartitions, and later operators,
 which reduces CPU, memory, and I/O.
 
@@ -99,11 +99,11 @@ driving this work. Related PRs: [#19918], [#20145], [#20192]
 
 ### Faster Query Planning
 
-DataFusion 53 improves query planning performance, making immutable pieces of
-execution plans cheaper to clone. This helps applications that need extreme low
-latency, plan many or complex queries, and those using prepared statements or
-parameterized queries. In some benchmarks
-overall execution time goes from ~4-5ms to spends ~100us.
+DataFusion 53 improves query planning performance by making immutable pieces of
+execution plans cheaper to clone. This helps applications that need extremely
+low latency, plan many or complex queries, or use prepared statements or
+parameterized queries. In some benchmarks, overall execution time drops from
+roughly 4-5 ms to about 100 us.
 
 Thanks to [@askalt] for leading this work. Related PRs: [#19792], [#19893]
 
@@ -128,7 +128,7 @@ Parquet files. Instead of reading an entire struct column and then extracting
 the field of interest, DataFusion 53 pushes the field extraction into the scan.
 
 For example, the following query reads a struct column `s` and extracts the
-`label` field, for rows where the `value` field is greater than 150:
+`label` field for rows where the `value` field is greater than 150:
 
 ```sql
 SELECT id, s['label']
@@ -370,5 +370,5 @@ grateful for the time and effort that everyone put in to make it happen.
 [datafusion-variant]: https://github.com/datafusion-contrib/datafusion-variant
 [Parquet Variant]: https://parquet.apache.org/blog/2026/02/27/variant-type-in-apache-parquet-for-semi-structured-data/
 [#21034]: https://github.com/apache/datafusion/issues/21034
-[upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading.html
+[upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading/index.html
 [datafusion-spark crate]: https://docs.rs/datafusion-spark/latest/datafusion_spark/index.html

From c79147d3f53ec3a51e7b7eb6ed27060621ca9954 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Tue, 31 Mar 2026 13:27:10 -0400
Subject: [PATCH 16/22] Add known issues

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index efe25f41..cd9b9f8d 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -189,6 +189,13 @@ DataFusion 53 includes some breaking changes, including updates to the SQL
 parser, optimizer behavior, and some physical-plan APIs. Please see the [upgrade
 guide] and [changelog] for the full details before upgrading.
 
+## Known Issues
+
+A small number of issues have been discovered after the 53.0.0 release
+and we are planning a 53.1.0 release soon. See the [53.1.0 release tracking
+issue] and the related [53.1.0 planning comment] for the latest details.
+
+
 ## Thank You
 
 Thank you to everyone in the DataFusion community who contributed code, reviews,
@@ -370,5 +377,7 @@ grateful for the time and effort that everyone put in to make it happen.
 [datafusion-variant]: https://github.com/datafusion-contrib/datafusion-variant
 [Parquet Variant]: https://parquet.apache.org/blog/2026/02/27/variant-type-in-apache-parquet-for-semi-structured-data/
 [#21034]: https://github.com/apache/datafusion/issues/21034
+[53.1.0 release tracking issue]: https://github.com/apache/datafusion/issues/21079
+[53.1.0 planning comment]: https://github.com/apache/datafusion/issues/21079#issuecomment-4120611441
 [upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading/index.html
 [datafusion-spark crate]: https://docs.rs/datafusion-spark/latest/datafusion_spark/index.html

From e404b057a3eecb0d91a378f6e755ee963ec11091 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Tue, 31 Mar 2026 13:30:04 -0400
Subject: [PATCH 17/22] wordsmit

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index cd9b9f8d..962bb68c 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -58,7 +58,7 @@ into earlier and cheaper stages of execution.
 ### `LIMIT`-Aware Parquet Row Group Pruning
 
 DataFusion 53 includes a new optimization that makes Parquet pruning aware of
-SQL `LIMIT`, which is described in full in [limit pruning blog post]. If
+SQL `LIMIT`. This optimization is described in full in [limit pruning blog post]. If
 DataFusion can prove that an entire row group matches the predicate, and those
 fully matching row groups contain enough rows to satisfy the `LIMIT`, partially
 matching row groups are skipped entirely. 
@@ -91,7 +91,7 @@ to both sides before the join. This is especially important with [dynamic filter
 
 <figure>
 <img src="/blog/images/datafusion-53.0.0/join-filter-pushdown.svg" width="80%" alt="Before and after diagram of filter pushdown around a hash join"/>
-<figcaption><b>Figure 2</b>: DataFusion 53 infers additional filters from join conditions and pushes them down in the plan.</figcaption>
+<figcaption><b>Figure 3</b>: DataFusion 53 infers additional filters from join conditions and pushes them down in the plan.</figcaption>
 </figure>
 
 Thanks to [@nuno-faria], [@haohuaijin], and [@jackkleeman] for
@@ -109,7 +109,7 @@ Thanks to [@askalt] for leading this work. Related PRs: [#19792], [#19893]
 
 ### Faster Functions
 
-DataFusion, like all real-world SQL engines, includes [235 built-in functions],
+DataFusion includes [235 built-in functions], 
 and improving the performance of these functions benefits a wide range of
 workloads. This release includes performance improvements for 42 of those
 functions, such as [strpos], [replace], [concat], [translate], [array_has],
@@ -138,7 +138,7 @@ WHERE s['value'] > 150;
 
 <figure>
 <img src="/blog/images/datafusion-53.0.0/field-access-pushdown.svg" width="80%" alt="Before and after diagram of field access pushdown into a data source"/>
-<figcaption><b>Figure 3</b>: DataFusion 53 pushes field-access expressions closer to the scan.</figcaption>
+<figcaption><b>Figure 4</b>: DataFusion 53 pushes field-access expressions closer to the scan.</figcaption>
 </figure>
 
 Special thanks to [@adriangb] for designing and implementing this optimizer
@@ -191,7 +191,7 @@ guide] and [changelog] for the full details before upgrading.
 
 ## Known Issues
 
-A small number of issues have been discovered after the 53.0.0 release
+A small number of issues were discovered after the 53.0.0 release,
 and we are planning a 53.1.0 release soon. See the [53.1.0 release tracking
 issue] and the related [53.1.0 planning comment] for the latest details.
 

From 68e9050b6e94d33b71e9c425eb19421f9332f218 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Tue, 31 Mar 2026 13:30:58 -0400
Subject: [PATCH 18/22] wordsmit

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 962bb68c..cd3c8de7 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -46,7 +46,7 @@ class="img-responsive"
 alt="Performance over time"
 />
 
-**Figure 1**: Average and median normalized query execution times for ClickBench queries for DataFusion 53.0.0 compared to previous releases.
+**Figure 1**: Average and median normalized execution times for DataFusion 53.0.0 on ClickBench queries, compared to previous releases.
 Query times are normalized using the ClickBench definition. See the
 [DataFusion Benchmarking Page](https://alamb.github.io/datafusion-benchmarking/)
 for more details.
@@ -58,7 +58,7 @@ into earlier and cheaper stages of execution.
 ### `LIMIT`-Aware Parquet Row Group Pruning
 
 DataFusion 53 includes a new optimization that makes Parquet pruning aware of
-SQL `LIMIT`. This optimization is described in full in [limit pruning blog post]. If
+`LIMIT`. This optimization is described in full in [limit pruning blog post]. If
 DataFusion can prove that an entire row group matches the predicate, and those
 fully matching row groups contain enough rows to satisfy the `LIMIT`, partially
 matching row groups are skipped entirely. 
@@ -109,9 +109,8 @@ Thanks to [@askalt] for leading this work. Related PRs: [#19792], [#19893]
 
 ### Faster Functions
 
-DataFusion includes [235 built-in functions], 
-and improving the performance of these functions benefits a wide range of
-workloads. This release includes performance improvements for 42 of those
+DataFusion includes [235 built-in functions]. Improving the performance of these
+functions benefits a wide range of workloads. This release includes performance improvements for 42 of those
 functions, such as [strpos], [replace], [concat], [translate], [array_has],
 [array_agg], [left], [right], and [case_when].
 
@@ -163,7 +162,7 @@ work. Related PRs: [#20065], [#20117], [#20239]
 
 * **Spark-Compatible Functions**: This release includes almost 20 new or improved
   Spark-compatible functions and behaviors in the [datafusion-spark crate].
-  This release contains functions such as [collect_list], [date_diff],
+  It includes functions such as [collect_list], [date_diff],
   [from_utc_timestamp], [json_tuple], [arrays_zip], [bin], and [array_contains].
   Thanks to the contributors who drove this work, especially [@cht42],
   [@CuteChuanChuan], [@SubhamSinghal], [@kazantsev-maksim], [@unknowntpo],
@@ -192,7 +191,7 @@ guide] and [changelog] for the full details before upgrading.
 ## Known Issues
 
 A small number of issues were discovered after the 53.0.0 release,
-and we are planning a 53.1.0 release soon. See the [53.1.0 release tracking
+and we expect to publish DataFusion 53.1.0 soon. See the [53.1.0 release tracking
 issue] and the related [53.1.0 planning comment] for the latest details.
 
 

From 7fac3c4dd37411682945039068fba065fb5aa8a7 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Tue, 31 Mar 2026 13:48:45 -0400
Subject: [PATCH 19/22] Add image styles

---
 content/blog/2026-03-25-datafusion-53.0.0.md | 23 ++++++++++++++++----
 1 file changed, 19 insertions(+), 4 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index cd3c8de7..32e2a7f6 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -42,7 +42,7 @@ making this release possible.
 <img
 src="/blog/images/datafusion-53.0.0/performance_over_time_clickbench.png"
 width="100%"
-class="img-responsive"
+class="img-fluid"
 alt="Performance over time"
 />
 
@@ -64,7 +64,12 @@ fully matching row groups contain enough rows to satisfy the `LIMIT`, partially
 matching row groups are skipped entirely. 
 
 <figure>
-<img src="/blog/images/limit-pruning/pruning-pipeline.svg" width="80%" alt="Pruning pipeline with limit pruning highlighted"/>
+<img 
+src="/blog/images/limit-pruning/pruning-pipeline.svg" 
+width="80%" 
+class="img-fluid"
+alt="Pruning pipeline with limit pruning highlighted"
+/>
 <figcaption><b>Figure 2</b>: Limit pruning is inserted between row group and page index pruning.</figcaption>
 </figure>
 
@@ -90,7 +95,12 @@ Now DataFusion can often transform the physical plan so the filter is applied
 to both sides before the join. This is especially important with [dynamic filters].
 
 <figure>
-<img src="/blog/images/datafusion-53.0.0/join-filter-pushdown.svg" width="80%" alt="Before and after diagram of filter pushdown around a hash join"/>
+<img 
+src="/blog/images/datafusion-53.0.0/join-filter-pushdown.svg" 
+width="80%" 
+alt="Before and after diagram of filter pushdown around a hash join"
+class="img-fluid"
+/>
 <figcaption><b>Figure 3</b>: DataFusion 53 infers additional filters from join conditions and pushes them down in the plan.</figcaption>
 </figure>
 
@@ -136,7 +146,12 @@ WHERE s['value'] > 150;
 ```
 
 <figure>
-<img src="/blog/images/datafusion-53.0.0/field-access-pushdown.svg" width="80%" alt="Before and after diagram of field access pushdown into a data source"/>
+<img 
+src="/blog/images/datafusion-53.0.0/field-access-pushdown.svg" 
+width="80%" 
+class="img-fluid"
+alt="Before and after diagram of field access pushdown into a data source"
+/>
 <figcaption><b>Figure 4</b>: DataFusion 53 pushes field-access expressions closer to the scan.</figcaption>
 </figure>
 

From 5bd1d5cc344855441c0a4aa36f06e9479823b247 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Wed, 1 Apr 2026 12:49:33 -0400
Subject: [PATCH 20/22] Update image

---
 content/blog/2026-03-25-datafusion-53.0.0.md  |  24 ++--
 .../join-filter-pushdown.svg                  | 113 +++++++++++-------
 2 files changed, 87 insertions(+), 50 deletions(-)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-03-25-datafusion-53.0.0.md
index 32e2a7f6..51b15f37 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-03-25-datafusion-53.0.0.md
@@ -86,22 +86,28 @@ For example:
 
 ```sql
 SELECT *
-FROM t1
-LEFT JOIN t2 ON t1.k = t2.k
-WHERE t1.k = 1;
+FROM (
+    SELECT *
+    FROM t1
+    LEFT ANTI JOIN t2 ON t1.k = t2.k
+) a
+JOIN t1 b ON a.k = b.k
+WHERE b.v = 1;
 ```
 
-Now DataFusion can often transform the physical plan so the filter is applied
-to both sides before the join. This is especially important with [dynamic filters].
+Now DataFusion can often transform the physical plan so filters and
+[dynamic filters] are pushed deeper into the plan, even through subqueries and
+nested joins. In this example, the filter on `b.v` helps derive dynamic filters
+that can be pushed into both sides of the nested anti join.
 
 <figure>
 <img 
 src="/blog/images/datafusion-53.0.0/join-filter-pushdown.svg" 
-width="80%" 
-alt="Before and after diagram of filter pushdown around a hash join"
+width="100%" 
+alt="Before and after diagram of dynamic filter pushdown through a subquery with nested joins"
 class="img-fluid"
 />
-<figcaption><b>Figure 3</b>: DataFusion 53 infers additional filters from join conditions and pushes them down in the plan.</figcaption>
+<figcaption><b>Figure 3</b>: DataFusion 53 pushes dynamic filters through subqueries and into both sides of nested joins.</figcaption>
 </figure>
 
 Thanks to [@nuno-faria], [@haohuaijin], and [@jackkleeman] for
@@ -207,7 +213,7 @@ guide] and [changelog] for the full details before upgrading.
 
 A small number of issues were discovered after the 53.0.0 release,
 and we expect to publish DataFusion 53.1.0 soon. See the [53.1.0 release tracking
-issue] and the related [53.1.0 planning comment] for the latest details.
+issue] for more details.
 
 
 ## Thank You
diff --git a/content/images/datafusion-53.0.0/join-filter-pushdown.svg b/content/images/datafusion-53.0.0/join-filter-pushdown.svg
index 68bd1101..141dd827 100644
--- a/content/images/datafusion-53.0.0/join-filter-pushdown.svg
+++ b/content/images/datafusion-53.0.0/join-filter-pushdown.svg
@@ -1,4 +1,4 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 900 320">
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1200 430">
   <defs>
     <marker id="arrow" markerWidth="10" markerHeight="7" refX="10" refY="3.5" orient="auto">
       <polygon points="0 0, 10 3.5, 0 7" fill="#555"/>
@@ -6,63 +6,94 @@
   </defs>
   <style>
     text { font-family: 'Segoe UI', Arial, sans-serif; }
-    .title { font-size: 17px; font-weight: 700; fill: #222; }
+    .title { font-size: 20px; font-weight: 700; fill: #222; }
+    .note { font-size: 13px; fill: #555; text-anchor: middle; }
     .stage { fill: #f0f4ff; stroke: #4a7bcc; stroke-width: 1.5; rx: 8; }
     .highlight { fill: #fff3e0; stroke: #e67e22; stroke-width: 2; rx: 8; }
     .label { font-size: 14px; fill: #333; text-anchor: middle; font-weight: 600; }
-    .sublabel { font-size: 13px; fill: #777; text-anchor: middle; }
+    .sublabel { font-size: 12px; fill: #666; text-anchor: middle; }
+    .group { fill: none; stroke: #cfd7e6; stroke-width: 1.5; stroke-dasharray: 5 4; rx: 10; }
   </style>
 
-  <!-- Before -->
-  <text class="title" x="225" y="28" text-anchor="middle">Before</text>
-  <line x1="40" y1="38" x2="410" y2="38" stroke="#ddd" stroke-width="1"/>
+  <text class="title" x="300" y="30" text-anchor="middle">Before</text>
+  <text class="note" x="300" y="50" text-anchor="middle">Dynamic filters stop at the subquery boundary</text>
+  <line x1="40" y1="60" x2="560" y2="60" stroke="#ddd" stroke-width="1"/>
 
-  <rect class="highlight" x="145" y="55" width="160" height="60"/>
-  <text class="label" x="225" y="79">FilterExec</text>
-  <text class="sublabel" x="225" y="100">k = 1</text>
+  <rect class="stage" x="235" y="78" width="130" height="42"/>
+  <text class="label" x="300" y="104">HashJoinExec</text>
+  <text class="sublabel" x="300" y="118">a.k = b.k</text>
 
-  <line x1="225" y1="115" x2="225" y2="128" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="270" y1="120" x2="170" y2="168" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="330" y1="120" x2="430" y2="168" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
 
-  <rect class="stage" x="150" y="132" width="150" height="42"/>
-  <text class="label" x="225" y="158">HashJoinExec</text>
+  <rect class="group" x="35" y="172" width="290" height="190" rx="10"/>
+  <text class="note" x="180" y="190">Subquery a</text>
 
-  <line x1="188" y1="174" x2="125" y2="220" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
-  <line x1="262" y1="174" x2="325" y2="220" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <rect class="stage" x="105" y="205" width="130" height="42"/>
+  <text class="label" x="170" y="231">HashJoinExec</text>
+  <text class="sublabel" x="170" y="245">LeftAnti</text>
 
-  <rect class="stage" x="45" y="224" width="160" height="48"/>
-  <text class="label" x="125" y="249">DataSourceExec</text>
-  <text class="sublabel" x="125" y="267">t1</text>
+  <line x1="145" y1="247" x2="120" y2="290" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="195" y1="247" x2="240" y2="290" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
 
-  <rect class="stage" x="245" y="224" width="160" height="48"/>
-  <text class="label" x="325" y="249">DataSourceExec</text>
-  <text class="sublabel" x="325" y="267">t2</text>
+  <rect class="stage" x="60" y="294" width="120" height="52"/>
+  <text class="label" x="120" y="320">DataSourceExec</text>
+  <text class="sublabel" x="120" y="338">t1</text>
 
-  <!-- After -->
-  <text class="title" x="675" y="28" text-anchor="middle">After</text>
-  <line x1="490" y1="38" x2="860" y2="38" stroke="#ddd" stroke-width="1"/>
+  <rect class="stage" x="185" y="294" width="120" height="52"/>
+  <text class="label" x="245" y="320">DataSourceExec</text>
+  <text class="sublabel" x="245" y="338">t2</text>
 
-  <rect class="stage" x="595" y="55" width="160" height="48"/>
-  <text class="label" x="675" y="82">HashJoinExec</text>
+  <rect class="highlight" x="365" y="205" width="130" height="58"/>
+  <text class="label" x="430" y="230">FilterExec</text>
+  <text class="sublabel" x="430" y="248">b.v = 1</text>
 
-  <line x1="638" y1="103" x2="585" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
-  <line x1="712" y1="103" x2="765" y2="136" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="430" y1="263" x2="430" y2="290" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
 
-  <rect class="highlight" x="505" y="140" width="160" height="60"/>
-  <text class="label" x="585" y="164">FilterExec</text>
-  <text class="sublabel" x="585" y="185">k = 1</text>
+  <rect class="stage" x="365" y="294" width="130" height="52"/>
+  <text class="label" x="430" y="320">DataSourceExec</text>
+  <text class="sublabel" x="430" y="338">t1 as b</text>
 
-  <rect class="highlight" x="685" y="140" width="160" height="60"/>
-  <text class="label" x="765" y="164">FilterExec</text>
-  <text class="sublabel" x="765" y="185">k = 1</text>
+  <text class="note" x="180" y="382">No dynamic filters reach the nested join inputs</text>
 
-  <line x1="585" y1="200" x2="585" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
-  <line x1="765" y1="200" x2="765" y2="217" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <text class="title" x="900" y="30" text-anchor="middle">After</text>
+  <text class="note" x="900" y="50" text-anchor="middle">Dynamic filters pushed through the subquery into both scans</text>
+  <line x1="640" y1="60" x2="1160" y2="60" stroke="#ddd" stroke-width="1"/>
 
-  <rect class="stage" x="505" y="221" width="160" height="48"/>
-  <text class="label" x="585" y="246">DataSourceExec</text>
-  <text class="sublabel" x="585" y="264">t1</text>
+  <rect class="stage" x="835" y="78" width="130" height="42"/>
+  <text class="label" x="900" y="104">HashJoinExec</text>
+  <text class="sublabel" x="900" y="118">a.k = b.k</text>
 
-  <rect class="stage" x="685" y="221" width="160" height="48"/>
-  <text class="label" x="765" y="246">DataSourceExec</text>
-  <text class="sublabel" x="765" y="264">t2</text>
+  <line x1="870" y1="120" x2="790" y2="168" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="930" y1="120" x2="1030" y2="168" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="group" x="620" y="172" width="340" height="190" rx="10"/>
+  <text class="note" x="790" y="190">Subquery a</text>
+
+  <rect class="stage" x="725" y="205" width="130" height="42"/>
+  <text class="label" x="790" y="231">HashJoinExec</text>
+  <text class="sublabel" x="790" y="245">LeftAnti</text>
+
+  <line x1="765" y1="247" x2="720" y2="290" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+  <line x1="815" y1="247" x2="880" y2="290" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="highlight" x="640" y="294" width="150" height="68"/>
+  <text class="label" x="715" y="320">DataSourceExec</text>
+  <text class="sublabel" x="715" y="338">t1 IN &lt;Dynamic Filter&gt;</text>
+
+  <rect class="highlight" x="805" y="294" width="150" height="68"/>
+  <text class="label" x="880" y="320">DataSourceExec</text>
+  <text class="sublabel" x="880" y="338">t2 IN &lt;Dynamic Filter&gt;</text>
+
+  <rect class="highlight" x="965" y="205" width="130" height="58"/>
+  <text class="label" x="1030" y="230">FilterExec</text>
+  <text class="sublabel" x="1030" y="248">b.v = 1</text>
+
+  <line x1="1030" y1="263" x2="1030" y2="290" stroke="#555" stroke-width="1.5" marker-end="url(#arrow)"/>
+
+  <rect class="stage" x="965" y="294" width="130" height="52"/>
+  <text class="label" x="1030" y="320">DataSourceExec</text>
+  <text class="sublabel" x="1030" y="338">t1 as b</text>
+
+  <text class="note" x="790" y="398">Dynamic filters are pushed into both sides of the nested anti join</text>
 </svg>

From 01c177ad2e6ee940943723a611fe4a67d55cb367 Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 2 Apr 2026 08:44:27 -0400
Subject: [PATCH 21/22] Update Date to 2026-04-02

---
 ...-25-datafusion-53.0.0.md => 2026-04-02-datafusion-53.0.0.md} | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
 rename content/blog/{2026-03-25-datafusion-53.0.0.md => 2026-04-02-datafusion-53.0.0.md} (99%)

diff --git a/content/blog/2026-03-25-datafusion-53.0.0.md b/content/blog/2026-04-02-datafusion-53.0.0.md
similarity index 99%
rename from content/blog/2026-03-25-datafusion-53.0.0.md
rename to content/blog/2026-04-02-datafusion-53.0.0.md
index 51b15f37..ed3e41a5 100644
--- a/content/blog/2026-03-25-datafusion-53.0.0.md
+++ b/content/blog/2026-04-02-datafusion-53.0.0.md
@@ -1,7 +1,7 @@
 ---
 layout: post
 title: Apache DataFusion 53.0.0 Released
-date: 2026-03-25
+date: 2026-04-02
 author: pmc
 categories: [release]
 ---

From b4a58fcae4d6c86cab4b342ddcd6cd065031b66c Mon Sep 17 00:00:00 2001
From: Andrew Lamb <andrew@nerdnetworks.org>
Date: Thu, 2 Apr 2026 08:48:11 -0400
Subject: [PATCH 22/22] last minute obsession cleanups

---
 content/blog/2026-04-02-datafusion-53.0.0.md | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/content/blog/2026-04-02-datafusion-53.0.0.md b/content/blog/2026-04-02-datafusion-53.0.0.md
index ed3e41a5..030e4f37 100644
--- a/content/blog/2026-04-02-datafusion-53.0.0.md
+++ b/content/blog/2026-04-02-datafusion-53.0.0.md
@@ -61,7 +61,7 @@ DataFusion 53 includes a new optimization that makes Parquet pruning aware of
 `LIMIT`. This optimization is described in full in [limit pruning blog post]. If
 DataFusion can prove that an entire row group matches the predicate, and those
 fully matching row groups contain enough rows to satisfy the `LIMIT`, partially
-matching row groups are skipped entirely. 
+matching row groups are skipped entirely.
 
 <figure>
 <img 
@@ -97,7 +97,7 @@ WHERE b.v = 1;
 
 Now DataFusion can often transform the physical plan so filters and
 [dynamic filters] are pushed deeper into the plan, even through subqueries and
-nested joins. In this example, the filter on `b.v` helps derive dynamic filters
+nested joins. In this example, the filter on `b.v` helps produce dynamic filters
 that can be pushed into both sides of the nested anti join.
 
 <figure>
@@ -126,7 +126,7 @@ Thanks to [@askalt] for leading this work. Related PRs: [#19792], [#19893]
 ### Faster Functions
 
 DataFusion includes [235 built-in functions]. Improving the performance of these
-functions benefits a wide range of workloads. This release includes performance improvements for 42 of those
+functions benefits a wide range of workloads. This release improves the performance of 42 of those
 functions, such as [strpos], [replace], [concat], [translate], [array_has],
 [array_agg], [left], [right], and [case_when].
 
@@ -213,7 +213,7 @@ guide] and [changelog] for the full details before upgrading.
 
 A small number of issues were discovered after the 53.0.0 release,
 and we expect to publish DataFusion 53.1.0 soon. See the [53.1.0 release tracking
-issue] for more details.
+issue] for the latest status.
 
 
 ## Thank You
@@ -398,6 +398,5 @@ grateful for the time and effort that everyone put in to make it happen.
 [Parquet Variant]: https://parquet.apache.org/blog/2026/02/27/variant-type-in-apache-parquet-for-semi-structured-data/
 [#21034]: https://github.com/apache/datafusion/issues/21034
 [53.1.0 release tracking issue]: https://github.com/apache/datafusion/issues/21079
-[53.1.0 planning comment]: https://github.com/apache/datafusion/issues/21079#issuecomment-4120611441
 [upgrade guide]: https://datafusion.apache.org/library-user-guide/upgrading/index.html
 [datafusion-spark crate]: https://docs.rs/datafusion-spark/latest/datafusion_spark/index.html