This release includes improvements to HTML fragment handling, fixes a memory growth issue when processing tables, and adds support for Python 3.15.

Fixes

• Fixed excessive memory growth when processing HTML containing multiple sibling <table> elements. See GHSA-q87h-c5m7-659j for details.
• Fixed HTML fragment handling to prevent synthetic wrapper elements from introducing unwanted spaces or indentation into extracted text.
• Fixed test failures on Windows by explicitly specifying file encodings.

Compatibility

• Added support for Python 3.15.
• Deprecated Python 3.10.

Maintenance

• Updated the GitHub CodeQL Actions dependency.

Credits

Many thanks to @gistrec for identifying and fixing the table-related memory growth issue, and to @OsamaAnsar for improving HTML fragment handling and reporting the Windows-related test issue.

Users of inscriptis 2.0.0–2.7.5 are encouraged to upgrade to benefit from the table-processing fix.

Fixes a denial-of-service issue where excessively large CSS length values in HTML style attributes could cause get_text() or get_annotated_text() to perform an unbounded string allocation (GHSA-4wxf-43cv-cfxx).

Thanks to @NotAFlightRisk for reporting this issue.

• Integrate a bufix contributed by @jimbofreedman that fixes #121 (Annotations in vertically padded table cells are misplaced, or raise IndexError)
• improved performance of Table.get_annotation.
• enhanced TableCell class semantics.
• set default HTTP user-agent in the inscript CLI client and support custom user-agent strings via the --user-agent option.
• improved reporting of benchmark results.

• Integrated the following bugfixes contributed by @santhreal
  • Fix: Single-line table cell annotations ignore center/right align (#117)
  • Fix: HtmlExtractor trailing newlines emit broken (#118)
• Improved typechecking via ty
• Use python:3.14-slim in docker images.
• Updated CI-Pipeline

• Removed upper bound on the Python version in accordance with the corresponding Python Packaging Guidelines (Fixes #112)
• Relaxed lxml upper version to < 6.2.0 (Fixes #111)
• Upgrade docker image to python 3.13-slim-bullseye (#110)
• Updated CI pipeline (#106, #107, #108, #109)
• Removed redundant COM rule in pyproject.toml.

• fixed dependencies to support python-lxml 6.0.x #102
• added Python 3.14 to the CI workflow (i.e., all inscriptis versions are now also tested with Python 3.14)
• minor code style improvement for ruff 0.15.x

• Added support for the value attribute in enumerations (fixes #98)
• Python 3.14 support
• modernized CI/CD pipeline
  • replace tox with ruff for faster static code analysis
  • added additional checks
  • updated development dependencies
• removed Python 3.9 support due to EOL
• optimizations, improved code formatting, and enhanced consistency across the codebase.

• Improved both the HTML and XML Annotation postprocessors.
• Fixed #94 (XML tags are now closed in the correct order).
• Breaking change:
  • the XmlAnnotationProcessor now uses a mandatory root element to ensure that the create XML is valid.
  • Impact:
    1. The created XML will contain a <content> root element, which contains all annotations.
    2. The name of the root element can be overwritten, if the optional root_element parameter is provided to the annotation processor call.
• Documentation fixes.
• Updated dependencies and added pytest as a build dependency.

This is a maintenance release which fixes a regression in the package dependencies introduced in 2.5.1.

• merged #91 to allow for newer requests versions.

• added Python 3.13 to the build pipeline.
• deprecated Python 3.8.
• updated dependencies.
• minor optimizations in the attribute handling.