Bulk partition UUID

[aasthabharill]

This Pull Request modifies the uniform partitioning (uniformization) logic in the sourcedb-to-spanner template to support tables partitioned on PostgreSQL UUID primary keys.

Changes Made & Rationale

1. Map UUID Columns to a Virtual "UUID" Collation

• Change: In PostgreSQLDialectAdapter.discoverTableIndexes, if the typeName of a column is "uuid", we assign "UUID" as its collation reference.
• Why: CollationMapper.fromDB expects a virtual "UUID" collation tag to trigger the static hexadecimal base-16 mapper (buildStaticUuidMapper). By assigning "UUID" during discovery, the splitter bypasses executing a database query to fetch collation rankings (which would fail or be extremely slow for a native UUID type that has no physical collation).

2. Configure Virtual Type Length to 32 for UUID Columns

• Change: In PostgreSQLDialectAdapter.discoverTableIndexes, if typeLength is null and typeName is "uuid", we set typeLength = 32.
• Why: While a standard canonical UUID is 36 characters long (including hyphens), CollationMapper strips the hyphens out during mapping, leaving exactly 32 hexadecimal characters. Overriding the discovered length to 32 ensures that no additional padding (virtual zero-rank characters) is appended during range partitioning calculations, ensuring a clean 1-to-1 mapping and unmapping.

3. Register State-Based Query and Parameter Cast Wrappers for UUID

• Change: In PostgreSQLDialectAdapter.discoverTableIndexes, if typeName is "uuid", we register explicit SQL cast statements in columnCastWrappers and columnParameterCastWrappers maps.
• Why:
  • columnCastWrappers (CAST(%s AS TEXT)): Used in getBoundaryQuery to query MIN(CAST(col AS TEXT)) and MAX(CAST(col AS TEXT)). This is necessary to retrieve the UUID boundaries safely as standard text strings compatible with JDBC. UUID doesnt have a MIN or MAX.
  • columnParameterCastWrappers (CAST(? AS uuid)): Used in getReadQuery and getCountQuery to bind parameter boundary placeholders as col >= CAST(? AS uuid). This is necessary because PostgreSQL does not support implicit comparison of standard JDBC string parameter bindings against native uuid column types.

4. Verify Changes with Unit & Integration Tests

• Collation Mapper Test: Added testUuidCollationMapper in CollationMapperTest.java to verify that canonical UUID strings are mapped to 128-bit BigIntegers and unmapped back with correct formatting and hyphen insertion.
• Dialect Adapter Test: Added testDiscoverTableIndexesWithUuid in PostgreSQLDialectAdapterTest.java verifying index discovery mappings, boundary query wrapping, and read/count query parameter bindings.
• Integration Test expected data logic: Updated getExpectedData in PostgreSQLWithUniformizationIT.java to support assertions for tables with non-integer primary keys (uuid_pk).

[codecov]

Codecov Report

❌ Patch coverage is 85.33333% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 59.50%. Comparing base (f8472fe) to head (4fd97fa).
⚠️ Report is 9 commits behind head on main.

Files with missing lines	Patch %	Lines
...ctadapter/postgresql/PostgreSQLDialectAdapter.java	76.59%	5 Missing and 6 partials ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #3798      +/-   ##
============================================
+ Coverage     53.41%   59.50%   +6.08%     
+ Complexity     6629     2184    -4445     
============================================
  Files          1082      506     -576     
  Lines         65795    29521   -36274     
  Branches       7328     3240    -4088     
============================================
- Hits          35147    17565   -17582     
+ Misses        28288    10970   -17318     
+ Partials       2360      986    -1374     

Components	Coverage Δ
spanner-templates	74.87% <85.33%> (+2.05%)	⬆️
spanner-import-export	∅ <ø> (∅)
spanner-live-forward-migration	80.86% <ø> (-0.07%)	⬇️
spanner-live-reverse-replication	77.02% <ø> (-0.03%)	⬇️
spanner-bulk-migration	91.02% <85.33%> (-0.08%)	⬇️
gcs-spanner-dv	86.69% <ø> (+0.94%)	⬆️

Files with missing lines	Coverage Δ
.../uniformsplitter/stringmapper/CollationMapper.java	97.82% <100.00%> (+0.55%)	⬆️
...ctadapter/postgresql/PostgreSQLDialectAdapter.java	93.36% <76.59%> (-3.94%)	⬇️

... and 603 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[VardhanThigle]

Please consider moving to binary index (similar to what we use for MySQL varbinary instead of a string for UUID. PG uses binary collation to compare UUIDs and that's more natural)

Do we need to take care of strict UUID version etc?

For PG, mostly not - https://www.db-fiddle.com/f/pVFVr6krWjQ2wHqstc44Hm/0 (please add this fiddle as a comment somewhere in your implementation)