v0.5.2

v0.5.2

Bug fixes

• Move linkml to runtime dependencies: linkml was only in dev dependencies despite being imported at runtime (data_loaders.py uses linkml.validator.loaders). Without the constraint, downstream resolvers could pair linkml-runtime 1.10.0 with an older incompatible linkml, causing Format.JSON import errors. (#192)

Full Changelog: v0.5.1...v0.5.2

v0.5.1

v0.5.1

Bug fixes

• Fix requires-python upper bound: <=3.13 was interpreted as <=3.13.0 by uv, excluding 3.13.1+ patch versions and breaking downstream dependency resolution. Changed to <3.14. (#191)

Full Changelog: v0.5.0...v0.5.1

v0.5.0

v0.5.0

Behavioral changes

• Null-safe expressions: {x} and bare x in expressions now propagate None through arithmetic, ordering comparisons, membership tests, and function calls instead of aborting the entire expression. Equality (==, !=) uses Python's native handling (None == "x" → False). This means case() fallback branches now fire correctly when variables are null — the previous behavior silently skipped them, which caused up to 84-93% row loss in pipelines with sparse companion variables. (#180, #181, #182)

Features

• Cross-table joins: populated_from now supports dot notation (table.field) for looking up values from joined tables via joins on ClassDerivation. (#159)
• --continue-on-error flag: map-data CLI can now collect row-level errors and report a summary instead of aborting on the first failure. Errors include structured context (class, slot, row index). (#162, #163)
• --target-schema flag: map-data CLI now accepts an explicit target schema. (#156)
• is_numeric() expression function: Curators can guard expressions against bad data: case((is_numeric(x), x * 2.54), (True, None)) (#186)
• Arithmetic coercion: Operators now coerce numeric strings from tabular input — "100" / "50" → 2.0. Non-coercible operands return None with a warning. (#186)
• none_if_non_numeric option on unit_conversion: Opt-in flag to return None for non-numeric values instead of raising. For columns with coded values (e.g., 'A', 'M') mixed with numeric data. (#175, #176)
• Multi-enum any_of support: Enum mapping now works for slots with multiple enum ranges via any_of. (#172)
• Enhanced mapping metadata: TransformationSpecification now supports FAIR mapping metadata and agent hierarchy. (#81)

Bug fixes

• Expression errors no longer masked: Real runtime errors (division by zero, type errors, non-numeric data) now propagate with actual messages instead of the misleading "Expression not in safe subset." (#185, #186)
• Source context in error messages: TransformationError now includes populated_from for both class and slot derivations, so curators see which source table/column caused the error. (#183, #184)
• Numeric string coercion in unit_conversion: String values like "120" from TSV/CSV columns are now coerced to float before conversion. (#175, #176)
• Key-based class handling: dynamic_object.py now handles classes with key: true attributes, not just identifier: true. (#169, #177)
• Expr with spaced class names: Expression slot derivations no longer resolve to None when the source class name contains a space. (#171)
• Scalars with no range: map_object no longer recurses into multivalued scalars with no declared range. (#173)

Infrastructure

• Bump linkml to 1.10.0 and linkml-runtime to 1.10.0. (#189)
• Require Python >=3.10 (3.9 is EOL). Ruff target bumped to py310 with pyupgrade modernization. (#189)
• Ruff lint and format enforced in CI via pre-commit. (#179)
• Biolink subsetting tests use local fixtures instead of live API fetch. (#168)

Full Changelog: v0.4.0...v0.5.0

v0.5.0-rc3

v0.5.0-rc3

All changes from v0.5.0-rc2, plus:

Bug fixes

• Expression errors no longer masked: Real runtime errors (division by zero, type errors, non-numeric data) now propagate with actual messages instead of the misleading "Expression not in safe subset." (#185, #186)

Features

• is_numeric() expression function: Curators can guard expressions against bad data: case((is_numeric(x), x * 2.54), (True, None)) (#186)
• Arithmetic coercion: Operators now coerce numeric strings from tabular input — "100" / "50" → 2.0. Non-coercible operands return None with a warning, enabling case() guards to work correctly. (#186)

v0.5.0-rc2

v0.5.0-rc2

All changes from v0.5.0-rc1, plus:

Improvements

• Source context in error messages: TransformationError now includes populated_from for both class and slot derivations, so curators see which source table/column caused the error without opening the YAML spec. (#183, #184)

  Before: class_derivation=Quantity; slot_derivation=value_decimal
  After: class_derivation=Quantity (from pht004041); slot_derivation=value_decimal (from phv00191041)

v0.5.0-rc1

v0.5.0-rc1

Behavioral changes

• Null-safe expressions: {x} and bare x in expressions now propagate None through arithmetic, ordering comparisons, membership tests, and function calls instead of aborting the entire expression. Equality (==, !=) uses Python's native handling (None == "x" → False). This means case() fallback branches now fire correctly when variables are null — the previous behavior silently skipped them, which was not the intended semantics and caused up to 84-93% row loss in pipelines with sparse companion variables. (#180, #181, #182)

Features

• Cross-table joins: populated_from now supports dot notation (table.field) for looking up values from joined tables via joins on ClassDerivation. (#159)
• --continue-on-error flag: map-data CLI can now collect row-level errors and report a summary instead of aborting on the first failure. Errors include structured context (class, slot, row index). (#162, #163)
• --target-schema flag: map-data CLI now accepts an explicit target schema. (#156)
• none_if_non_numeric option on unit_conversion: Opt-in flag to return None for non-numeric values instead of raising. For columns with coded values (e.g., 'A', 'M') mixed with numeric data. (#175, #176)
• Multi-enum any_of support: Enum mapping now works for slots with multiple enum ranges via any_of. (#172)
• Enhanced mapping metadata: TransformationSpecification now supports FAIR mapping metadata and agent hierarchy. (#81)

Bug fixes

• Numeric string coercion in unit_conversion: String values like "120" from TSV/CSV columns are now coerced to float before conversion. (#175, #176)
• Key-based class handling: dynamic_object.py now handles classes with key: true attributes, not just identifier: true. (#169, #177)
• Expr with spaced class names: Expression slot derivations no longer resolve to None when the source class name contains a space. (#171)
• Scalars with no range: map_object no longer recurses into multivalued scalars with no declared range. (#173)

Infrastructure

• Ruff lint and format enforced in CI via pre-commit. (#179)
• Biolink subsetting tests use local fixtures instead of live API fetch. (#168)
• Removed unused Dragon AI workflow.
• Added Copilot code review ruleset.

v0.4.0

What's Changed

• Implement pivot operation for EAV/measurement data transformation by @cmungall in #108
• Relax dependency version constraints for library compatibility by @amc-corey-cox in #109
• Add markers and instructions to extract branches from long functions by @amc-corey-cox in #105
• offset implementation by @madanucd in #110
• Fix known dependency vulnerabilities by @jackhiggs in #119
• Allow multiple class_derivations for the same target class by @amc-corey-cox in #118
• Add uuid5() function to transformation expressions by @amc-corey-cox in #117
• Replace hand-rolled expression evaluator with simpleeval by @amc-corey-cox in #122
• Add multi-output stream writers and -O CLI option by @amc-corey-cox in #115
• Coerce numeric strings in comparison operators by @amc-corey-cox in #135
• NMDC: Document inlined-object flattening to lakehouse-style columns (tests only) by @turbomam in #130
• NMDC: Document string/scalar-to-structured measurement transforms (tests only) by @turbomam in #131
• Add cross-table lookup support for join-based transformations by @amc-corey-cox in #136
• Add NMDC transformation pattern tests (flattening + uuid5) by @turbomam in #145
• Add NMDC Biosample-Study cross-table join tests (#136) by @turbomam in #148
• Fix JSON output: omit null values and add trailing newline by @amc-corey-cox in #149
• Clarify architectural vision: TransformationSpecification as IR by @amc-corey-cox in #153
• Document expression language, functions, and multi-output for RC by @amc-corey-cox in #154

New Contributors

• @madanucd made their first contribution in #110
• @jackhiggs made their first contribution in #119

Full Changelog: v0.3.9...v0.4.0

v0.4.0-rc2

What's Changed

New Features

• Add multi-output stream writers and -O/--additional-output CLI option by @amc-corey-cox in #115
• Add uuid5() function to transformation expressions by @amc-corey-cox in #117
• Allow multiple class_derivations for the same target class by @amc-corey-cox in #118
• Replace hand-rolled expression evaluator with simpleeval by @amc-corey-cox in #122
• Coerce numeric strings in comparison operators by @amc-corey-cox in #135
• Add cross-table lookup support for join-based transformations by @amc-corey-cox in #136
• Implement pivot operation for EAV/measurement data transformation by @cmungall in #108
• Implement offset operation by @madanucd in #110

Bug Fixes

• Fix JSON output: omit null values and add trailing newline by @amc-corey-cox in #149

Security

• Bump asteval>=1.0.5 and duckdb>=1.1.0 to address known security vulnerabilities by @jackhiggs in #119

Documentation

• Clarify architectural vision: TransformationSpecification as IR by @amc-corey-cox in #153
• Document expression language, functions, multi-output, and joins for RC by @amc-corey-cox in #154

Tests

• Add NMDC transformation pattern tests (flattening + uuid5) by @turbomam in #145
• Add NMDC Biosample-Study cross-table join tests by @turbomam in #148
• NMDC: Document inlined-object flattening to lakehouse-style columns (tests only) by @turbomam in #130
• NMDC: Document string/scalar-to-structured measurement transforms (tests only) by @turbomam in #131

Maintenance

• Relax dependency version constraints for library compatibility by @amc-corey-cox in #109

New Contributors

• @jackhiggs made their first contribution in #119
• @madanucd made their first contribution in #110
• @turbomam made their first contribution in #130

Full Changelog: v0.3.9...v0.4.0-rc2

v0.3.9

What's Changed

• Add support for copy directives by @Silvanoc in #35
• add biolink subset test separte from other example tests for ease of … by @sierra-moxon in #39
• add denormalizing transform notebook by @sierra-moxon in #40
• Fix up and re-enable tests by @pkalita-lbl in #44
• Handle overrides on SlotDerivation by @pkalita-lbl in #45
• Clean up dependency specifications by @pkalita-lbl in #46
• Converting all unit tests to pytest by @ialarmedalien in #51
• Ruff format and lint by @ialarmedalien in #52
• Removing poetry and using uv for python and package management by @ialarmedalien in #54
• Ruff linter fixes: automated by @ialarmedalien in #55
• Markdown compiler -- minor template fix by @ialarmedalien in #56
• Automatic python code fixes suggested by ruff by @ialarmedalien in #57
• Automatic ruff lint fixes, plus some type additions and docs. by @ialarmedalien in #58
• add license by @kevinschaper in #66
• Address Page not founds by @hrshdhgd in #77
• Map value when single value provided for multivalued slot. by @martinwellman in #78
• Implement value_mappings by @amc-corey-cox in #65
• Implement transform on in-lined classes in multi-value slot by @amc-corey-cox in #73
• Slot derivation for constant values by @amc-corey-cox in #80
• Fix Model Issues by @amc-corey-cox in #84
• add template for pydantic gen and value to slotderivation by @sierra-moxon in #87
• Adds new testing Framework with example by @amc-corey-cox in #90
• Update deepdiff to fix CVE by @amc-corey-cox in #92
• And read and write TSV for data mapping by @amc-corey-cox in #100
• Allow populated_from to lookup from external class by @amc-corey-cox in #101
• Adding AI integrations from copier template by @cmungall in #102

New Contributors

• @sierra-moxon made their first contribution in #39
• @pkalita-lbl made their first contribution in #44
• @kevinschaper made their first contribution in #66
• @hrshdhgd made their first contribution in #77
• @amc-corey-cox made their first contribution in #65

Full Changelog: v0.3.8...v0.3.9

v0.3.9-rc2

What's Changed

• Fix Model Issues by @amc-corey-cox in #84
• add template for pydantic gen and value to slotderivation by @sierra-moxon in #87
• Adds new testing Framework with example by @amc-corey-cox in #90
• Update deepdiff to fix CVE by @amc-corey-cox in #92

Full Changelog: v0.3.9-rc1...v0.3.9-rc2