dcmspec 0.4.0__tar.gz → 0.4.2__tar.gz

{dcmspec-0.4.0 → dcmspec-0.4.2}/CHANGELOG.md

@@ -1,113 +1,190 @@
-# Release Notes
-
-These release notes summarize key changes, improvements, and breaking updates for each version of **dcmspec**.
-
-## [0.4.0] - 2026-09-11
-
-### Added
-
-- `DOMSectionSpecParser` parses a DICOM standard section, given its anchor id.
-- `ModuleSpecBuilder` builds a module attribute model and resolves the sections it
-  directly references, caching each one separately and downloading their images.
-- `IODSpecBuilder` accepts an optional `module_builder` so each module built as part
-  of a full IOD has its referenced sections resolved.
-
-### Changed
-
-- **Breaking change:** `XHTMLDocHandler.download`/`PDFDocHandler.download` renamed to
-  `download_to_cache`. Callers using `load_document` are unaffected.
-
-### Fixed
-
-- `DOMTableSpecParser.get_version` no longer raises `UnboundLocalError` when the document has no
-  `<div class="titlepage">`; it will return `""` with a warning.
-- `SpecModel.exclude_titles` no longer raises `AttributeError` when a model's metadata has no
-  `column_to_attr` such as a section model.
-
-## [0.3.1] - 2026-08-27
-
-### Added
-
-- Reorganized `tests/` into `tests/unit`, `tests/integration`, and `tests/e2e` tiers, and added an opt-in `tests/e2e`
-  canary suite that downloads and parses real, current DICOM standard tables (Part 3 IOD+modules, Part 6 data
-  dictionary, Part 4 UPS DIMSE attributes, and a Part 3/Part 6 merge) to catch upstream markup changes between
-  standard releases. Excluded from the default test run; also runs on a monthly schedule via
-  `.github/workflows/e2e.yml`.
-
-### Fixed
-
-- `PDFDocHandler.select_tables` now raises `ValueError` when a header cell contains a DICOM tag pattern (e.g. `(300A,0230)`), indicating that a data row was absorbed into the header — typically a `table_header_rowspan` over-count. This fails loudly instead of silently dropping an attribute row. Opt out with `strict_header_check=False` to restore the prior warn-and-continue behavior.
-- `PDFDocHandler.concat_tables` now reconstructs page-straddling description cells. When an attribute's description wraps across a PDF page boundary, pdfplumber emits the continuation as an untagged row (empty name and tag, description only) atop the next page's table; this is now merged into the preceding tagged row's description instead of becoming a floating node. Recovers enum values and conditional requirements that were silently lost at page breaks (e.g. HDSS Contour Geometric Type `(3006,0042)` losing its `POINT` / `CLOSED_PLANAR` / `CLOSEDPLANAR_XOR` enumeration, where `CLOSEDPLANAR_XOR` reflects the original DICOM spelling).
-- `PDFDocHandler.load_document` / `extract_tables_pdfplumber` now accept a `snap_tolerance_overrides` dict (keyed by 1-indexed page number) to lower pdfplumber's `snap_tolerance` for individual pages. The default (8) snaps two near-coincident header rules together on a few pages, fusing a repeated continuation sub-header (e.g. `Presence | Specific Rules`) into the first data row of the next page; that row then trips the tag-in-header guard (rowspan over-count) and the page cannot be extracted. Lowering the tolerance for just that page (e.g. `{34: 6}`) keeps the two rules distinct so the sub-header is recognized as a header row, recovering the first attribute of the continuation table (e.g. TPPC-Brachy HDR/PDR Source `(0008,1040)`) without disturbing extraction of any other page. This also un-fuses line-snapped "frankenrows" at the source on the affected page; the `concat_tables` split below remains as a downstream safety net for pages that still fuse at the default tolerance.
-- `PDFDocHandler.concat_tables` now splits fused "frankenrow" rows back into their constituent attribute rows. pdfplumber's line-snapping can merge two adjacent table rows whose separating rule falls within `snap_tolerance` into one row, newline-joining every column (e.g. tag `(300A,0214)\n(300A,0216)`, type `1\n3`). A legitimate attribute row carries exactly one DICOM tag, so a tag cell holding N≥2 DICOM-tag patterns is unambiguously a fusion; the row is split into N rows **only** when every structured column yields exactly N aligned newline-parts (the description must be blank or also N parts), otherwise it is left intact and logged. Mirrors the continuation-merge discriminator and recovers attributes (e.g. TPPC-Brachy `(300A,0214)` Source Type / `(300A,0216)` Source Manufacturer) that were fused by line-snapping into a single malformed node.
-- `IODSpecBuilder.build_from_url` no longer raises a spurious `DeprecationWarning` when callers use only `progress_observer`. An internal call was incorrectly re-passing it as the legacy `progress_callback` argument as well.
-
-## [0.3.0] - 2025-11-27
-
-### Added
-
-- CSV output mode in `SpecPrinter` via `print_csv` method
-- Excel OOXML output mode in `SpecPrinter` via `print_xlsx` method
-- Optional `output` parameter in `SpecPrinter` for writing to files
-- Optional `column_width` parameter in `print_table` and `print_xlsx`
-- `--print-mode` option in `modattributes` CLI for CSV and OOXML output
-- `--output` option in `modattributes` CLI for directing output to files
-
-### Changed
-
-- Switch PyPI version badge in `README` from `badge.fury.io` to `shields.io` for faster updates
-- Update modattributes CLI example for CSV and XLSX output
-- Improved nesting level color schemes for consistency and better contrast and readability.
-
-### Fixed
-
-- Correct handling of `include_table` names in `DOMTableSpecParser`
-- Correct handling of empty node attributes in `SpecPrinter`
-
-## [0.2.3] - 2025-09-29
-
-### Fixed
-
-- Hotfix: Force UTF-8 decoding for DICOM standard XHTML downloads to prevent mojibake when server omits charset ([#85](https://github.com/dwikler/dcmspec/issues/85)).
-- Hotfix: Add missing `progress_observer` argument to `CSVTableSpecParser.parse` for interface compatibility and to prevent `TypeError` when used with `SpecFactory` ([#86](https://github.com/dwikler/dcmspec/issues/86)).
-
-## [0.2.2] - 2025-09-25
-
-### Fixed
-
-- Fix CONTRIBUTING.md link in README for PyPI compatibility
-- Remove focus border and misleading text cursor in iod-explorer details panel
-
-### Changed
-
-- Update README: add PyPI and Python version badges
-- Replace Unicode ▶ with ASCII > in status bar for compatibility
-- Improve DICOM Modules usage condition parsing using regex for robustness to missing spaces
-- Add PR template to remind contributors to check the target branch and check tests and docs were updated
-- Move detailed table parsing logs to DEBUG level for less verbose INFO output
-
-## [0.2.1] - 2025-09-19
-
-### Fixed
-
-- Sanitize node and attribute names to remove "/" in DOMTableSpecParser ([#56](https://github.com/dwikler/dcmspec/issues/56))
-
-### Changed
-
-- Major project restructure: move CLI and UI apps to new `apps/cli` and `apps/ui` folders
-- Improve installation instructions and documentation
-- Prepare and publish the package to [PyPI](https://pypi.org/project/dcmspec/)
-
-## [0.2.0] - 2025-09-13
-
-### Changed
-
-- **Breaking change:** `IODSpecBuilder.build_from_url` now returns a tuple `(iod_model, module_models)` instead of just the IOD model. All callers must be updated to unpack the tuple
-- Update CLI and UI applications to support new return value
-- Add registry mode to `IODSpecBuilder` for efficient module model sharing
-
-## [0.1.0] - 2025-05-25
-
-### Added
-
-- Initial release
+# Release Notes
+
+These release notes summarize key changes, improvements, and breaking updates for each version of **dcmspec**.
+
+## [0.4.2] - 2026-11-14
+
+### Fixed
+
+- `PDFDocHandler.load_document` now accepts `strict_header_check`, so the check for a DICOM tag in the header
+  can be turned off through the `handler_kwargs` of `SpecFactory`
+  ([#117](https://github.com/dwikler/dcmspec/issues/117)).
+- `SpecPrinter` and `IODSpecPrinter` no longer write ANSI escape sequences or wrap rows at 80
+  characters when stdout is redirected, and `--no-color` also removes the magenta table header
+  ([#122](https://github.com/dwikler/dcmspec/issues/122)).
+- `SpecPrinter` and `IODSpecPrinter` no longer wrap rows at 80 characters when writing to a file
+  with `--output`
+  ([#112](https://github.com/dwikler/dcmspec/issues/112)).
+- `ServiceAttributeModel.select_dimse` now keeps indices of `metadata.column_to_attr`
+  aligned with indices of `metadata.header` when filtering DIMSE columns
+  ([#26](https://github.com/dwikler/dcmspec/issues/26)).
+- `SectionImageResolver` now stores `image_paths` with forward slashes on every OS, instead of
+  the OS-specific separator. Section models cached on Windows by an earlier version keep their
+  backslash paths: delete them from the cache to rebuild them
+  ([#156](https://github.com/dwikler/dcmspec/issues/156)).
+- `DOMTableSpecParser` no longer puts the title of a skipped column in `metadata.header` and
+  drops the last one when `skip_columns` is used. Models cached by an earlier version keep the
+  wrong header: delete them from the cache to rebuild them
+  ([#154](https://github.com/dwikler/dcmspec/issues/154)).
+- The IOD Explorer install instructions now use the correct syntax: `poetry install --extras gui`.
+- The `iodmodules` and `iodattributes` apps now handle Normalized IOD tables correctly. Models
+  cached by an earlier version keep the old values: delete them from the cache to rebuild them
+  ([#51](https://github.com/dwikler/dcmspec/issues/51)).
+- The CLI apps no longer fail with a `UnicodeEncodeError` on Windows when their output is
+  redirected with `>` ([#166](https://github.com/dwikler/dcmspec/issues/166)).
+
+## [0.4.1] - 2026-09-27
+
+### Added
+
+- `ModuleSpecBuilder.resolve_section` resolves a single section by id, for callers that
+  built a module model without resolving its referenced attribute descriptions
+  ([#131](https://github.com/dwikler/dcmspec/issues/131)).
+- The [API Overview](https://dwikler.github.io/dcmspec/api/) page documents the
+  architecture of the library and the SpecFactory pipeline; the API Reference nav is
+  reorganized into consistent functional categories
+  ([#139](https://github.com/dwikler/dcmspec/issues/139)).
+
+### Changed
+
+- **Breaking change:** `SpecFactory.try_load_cache`'s `ref_columns` parameter is replaced
+  by a generic `parser_kwargs` dict ([#124](https://github.com/dwikler/dcmspec/issues/124)).
+- **Breaking change:** `SpecStore` gains a new abstract `file_extension` property, reporting
+  a store's cache file extension; custom subclasses must implement it
+  ([#137](https://github.com/dwikler/dcmspec/issues/137)).
+
+### Fixed
+
+- `ModuleSpecBuilder` no longer resolves a "See Section X" reference that isn't an
+  attribute description of a module or macro; such references are now skipped
+  ([#132](https://github.com/dwikler/dcmspec/issues/132)).
+- A resolved section's cached model is now updated with its resolved image paths ([#131](https://github.com/dwikler/dcmspec/issues/131)).
+- `SpecFactory.build_model`'s docstring now notes `doc_object` is ignored when a cached
+  model is already found, and `try_load_cache` documents how to check for one without a
+  parsed document ([#133](https://github.com/dwikler/dcmspec/issues/133)).
+- `SpecFactory` now checks all `parser_kwargs` (e.g. `skip_columns`, `unformatted`) against
+  a cache hit, not just `ref_columns`; a cache built with different values is now rebuilt
+  instead of silently reused ([#124](https://github.com/dwikler/dcmspec/issues/124)).
+- `SpecFactory.create_model` and `ModuleSpecBuilder.build_from_url` no longer rely on
+  `input_handler.cache_file_name` being set for cache lookup, fixing a mix-up
+  risk when reusing one factory/builder for multiple calls with different
+  `cache_file_name` values ([#138](https://github.com/dwikler/dcmspec/issues/138)).
+- `SpecFactory` and `ModuleSpecBuilder` no longer hardcode `.json` when deriving a default
+  cache filename; they now use the configured `model_store`'s file extension
+  ([#137](https://github.com/dwikler/dcmspec/issues/137)).
+- `SpecFactory` no longer reparses when a `parser_kwargs` option was omitted and the default
+  value is used ([#142](https://github.com/dwikler/dcmspec/issues/142)).
+- `SpecFactory` now checks `column_to_attr` and `name_attr` against a cache hit, along with
+  the existing `include_depth` and `parser_kwargs` checks; a cache built with different
+  values is rebuilt instead of silently reused
+  ([#146](https://github.com/dwikler/dcmspec/issues/146)).
+
+## [0.4.0] - 2026-09-11
+
+### Added
+
+- `DOMSectionSpecParser` parses a DICOM standard section, given its anchor id.
+- `ModuleSpecBuilder` builds a module attribute model and resolves the sections it
+  directly references, caching each one separately and downloading their images.
+- `IODSpecBuilder` accepts an optional `module_builder` so each module built as part
+  of a full IOD has its referenced sections resolved.
+
+### Changed
+
+- **Breaking change:** `XHTMLDocHandler.download`/`PDFDocHandler.download` renamed to
+  `download_to_cache`. Callers using `load_document` are unaffected.
+
+### Fixed
+
+- `DOMTableSpecParser.get_version` no longer raises `UnboundLocalError` when the document has no
+  `<div class="titlepage">`; it will return `""` with a warning.
+- `SpecModel.exclude_titles` no longer raises `AttributeError` when a model's metadata has no
+  `column_to_attr` such as a section model.
+
+## [0.3.1] - 2026-08-27
+
+### Added
+
+- Reorganized `tests/` into `tests/unit`, `tests/integration`, and `tests/e2e` tiers, and added an opt-in `tests/e2e`
+  canary suite that downloads and parses real, current DICOM standard tables (Part 3 IOD+modules, Part 6 data
+  dictionary, Part 4 UPS DIMSE attributes, and a Part 3/Part 6 merge) to catch upstream markup changes between
+  standard releases. Excluded from the default test run; also runs on a monthly schedule via
+  `.github/workflows/e2e.yml`.
+
+### Fixed
+
+- `PDFDocHandler.select_tables` now raises `ValueError` when a header cell contains a DICOM tag pattern (e.g. `(300A,0230)`), indicating that a data row was absorbed into the header — typically a `table_header_rowspan` over-count. This fails loudly instead of silently dropping an attribute row. Opt out with `strict_header_check=False` to restore the prior warn-and-continue behavior.
+- `PDFDocHandler.concat_tables` now reconstructs page-straddling description cells. When an attribute's description wraps across a PDF page boundary, pdfplumber emits the continuation as an untagged row (empty name and tag, description only) atop the next page's table; this is now merged into the preceding tagged row's description instead of becoming a floating node. Recovers enum values and conditional requirements that were silently lost at page breaks (e.g. HDSS Contour Geometric Type `(3006,0042)` losing its `POINT` / `CLOSED_PLANAR` / `CLOSEDPLANAR_XOR` enumeration, where `CLOSEDPLANAR_XOR` reflects the original DICOM spelling).
+- `PDFDocHandler.load_document` / `extract_tables_pdfplumber` now accept a `snap_tolerance_overrides` dict (keyed by 1-indexed page number) to lower pdfplumber's `snap_tolerance` for individual pages. The default (8) snaps two near-coincident header rules together on a few pages, fusing a repeated continuation sub-header (e.g. `Presence | Specific Rules`) into the first data row of the next page; that row then trips the tag-in-header guard (rowspan over-count) and the page cannot be extracted. Lowering the tolerance for just that page (e.g. `{34: 6}`) keeps the two rules distinct so the sub-header is recognized as a header row, recovering the first attribute of the continuation table (e.g. TPPC-Brachy HDR/PDR Source `(0008,1040)`) without disturbing extraction of any other page. This also un-fuses line-snapped "frankenrows" at the source on the affected page; the `concat_tables` split below remains as a downstream safety net for pages that still fuse at the default tolerance.
+- `PDFDocHandler.concat_tables` now splits fused "frankenrow" rows back into their constituent attribute rows. pdfplumber's line-snapping can merge two adjacent table rows whose separating rule falls within `snap_tolerance` into one row, newline-joining every column (e.g. tag `(300A,0214)\n(300A,0216)`, type `1\n3`). A legitimate attribute row carries exactly one DICOM tag, so a tag cell holding N≥2 DICOM-tag patterns is unambiguously a fusion; the row is split into N rows **only** when every structured column yields exactly N aligned newline-parts (the description must be blank or also N parts), otherwise it is left intact and logged. Mirrors the continuation-merge discriminator and recovers attributes (e.g. TPPC-Brachy `(300A,0214)` Source Type / `(300A,0216)` Source Manufacturer) that were fused by line-snapping into a single malformed node.
+- `IODSpecBuilder.build_from_url` no longer raises a spurious `DeprecationWarning` when callers use only `progress_observer`. An internal call was incorrectly re-passing it as the legacy `progress_callback` argument as well.
+
+## [0.3.0] - 2025-11-27
+
+### Added
+
+- CSV output mode in `SpecPrinter` via `print_csv` method
+- Excel OOXML output mode in `SpecPrinter` via `print_xlsx` method
+- Optional `output` parameter in `SpecPrinter` for writing to files
+- Optional `column_width` parameter in `print_table` and `print_xlsx`
+- `--print-mode` option in `modattributes` CLI for CSV and OOXML output
+- `--output` option in `modattributes` CLI for directing output to files
+
+### Changed
+
+- Switch PyPI version badge in `README` from `badge.fury.io` to `shields.io` for faster updates
+- Update modattributes CLI example for CSV and XLSX output
+- Improved nesting level color schemes for consistency and better contrast and readability.
+
+### Fixed
+
+- Correct handling of `include_table` names in `DOMTableSpecParser`
+- Correct handling of empty node attributes in `SpecPrinter`
+
+## [0.2.3] - 2025-09-29
+
+### Fixed
+
+- Hotfix: Force UTF-8 decoding for DICOM standard XHTML downloads to prevent mojibake when server omits charset ([#85](https://github.com/dwikler/dcmspec/issues/85)).
+- Hotfix: Add missing `progress_observer` argument to `CSVTableSpecParser.parse` for interface compatibility and to prevent `TypeError` when used with `SpecFactory` ([#86](https://github.com/dwikler/dcmspec/issues/86)).
+
+## [0.2.2] - 2025-09-25
+
+### Fixed
+
+- Fix CONTRIBUTING.md link in README for PyPI compatibility
+- Remove focus border and misleading text cursor in iod-explorer details panel
+
+### Changed
+
+- Update README: add PyPI and Python version badges
+- Replace Unicode ▶ with ASCII > in status bar for compatibility
+- Improve DICOM Modules usage condition parsing using regex for robustness to missing spaces
+- Add PR template to remind contributors to check the target branch and check tests and docs were updated
+- Move detailed table parsing logs to DEBUG level for less verbose INFO output
+
+## [0.2.1] - 2025-09-19
+
+### Fixed
+
+- Sanitize node and attribute names to remove "/" in DOMTableSpecParser ([#56](https://github.com/dwikler/dcmspec/issues/56))
+
+### Changed
+
+- Major project restructure: move CLI and UI apps to new `apps/cli` and `apps/ui` folders
+- Improve installation instructions and documentation
+- Prepare and publish the package to [PyPI](https://pypi.org/project/dcmspec/)
+
+## [0.2.0] - 2025-09-13
+
+### Changed
+
+- **Breaking change:** `IODSpecBuilder.build_from_url` now returns a tuple `(iod_model, module_models)` instead of just the IOD model. All callers must be updated to unpack the tuple
+- Update CLI and UI applications to support new return value
+- Add registry mode to `IODSpecBuilder` for efficient module model sharing
+
+## [0.1.0] - 2025-05-25
+
+### Added
+
+- Initial release