Version 1.7.8

What's Changed

• Handle None Relationships when parsing LAYOUT_FIGURE

Full Changelog: v1.7.7...v1.7.8

Version 1.7.7

What's Changed

• Handle None bounding box when parsing Queries by @Belval in #340

Full Changelog: v1.7.6...v1.7.7

Version 1.7.6

What's Changed

• Add CITATION.cff by @Belval in #332
• Add missing entities in docs by @Belval in #334
• Handle null EntityTypes by @Belval in #339

Full Changelog: v1.7.5...v1.7.6

Version 1.7.5

What's Changed

• Make KeyValue.key an EntityList by @Belval in #320
• Remove numpy from explicit dependencies by @Belval in #324
• Hide key value layouts by @Belval in #325
• Return query and query answer with get_text() by @Belval in #329
• Convert image to RGB in EntityList for Jupyter compatibility by @Belval in #330
• Support for Python 3.12 by @tb102122 in #311

Full Changelog: v1.7.4...v1.7.5

Version 1.7.4

What's Changed

• Fix table title .get_text() by @Belval in #314
• Fix .to_pandas() raising an exception by @Belval in #315

Full Changelog: v1.7.3...v1.7.4

Version 1.7.3

What's Changed

• Table linearization improvements by @Belval in #313

  • Add .get_text(), .to_html() and .to_markdown() functions to Linearizable which is now implemented by Document, Page, DocumentEntity and EntityList
  • Add HTMLLinearizationConfig and MarkdownLinearizationConfig as pre-configured TextLinearizationConfig
  • Add the follow parameters to TextLinearizationConfig
    • duplicate_text_in_merged_cells duplicates the text in merge cells to preserve row-level alignment
    • table_flatten_headers combines multi-row headers into a single row, duplicating the merged cells horizontally as needed
    • table_tabulate_remove_extra_hyphens removes extra hyphens '-' in markdown tables to reduce context length
    • max_number_of_consecutive_spaces defines the maximum number of contiguous whitespace characters, similar to max_number_of_consecutive_new_lines

• Fixes:

  • Fix trailing whitespace in cell text
  • Fix table_column_separator being hardcoded as '\t'
  • Fix table_row_separator being hardcoded as '\n'
  • Resets BytesIO buffer to 0 position by @abest0 in #310

New Contributors

• @abest0 made their first contribution in #310

Full Changelog: v1.7.2...v1.7.3

Version 1.7.2

What's Changed

• Fix for page objects not always having an image attached, causing an exception on .visualize()

Full Changelog: v1.7.1...v1.7.2

Version 1.7.1

What's Changed

• Fix issue where a table within a container layout could be duplicated in the .get_text() output.

Full Changelog: v1.7.0...v1.7.1

Version 1.7.0

What's Changed

• Loosen XlsxWriter version constraints by @mdscruggs in #292
• Rework the linearization heuristic to ensure that no words are missing or duplicated
• Fix KeyValues being assigned twice on overlapping table cells, going forward KVs inside a tables are ignored (table structure takes precedence)
• Hardens parser code against missing children in layouts or KeyValues with missing keys
• Fix markdown tables not having header rows when one of the cell is empty
• Add support for Python 3.11 and 3.12 in the GitHub action workflows
• Add textractor.__version__ to allow easier identification of the installed Textractor version in code
• Added hide_table_layout
• Remove amazon-textract-response-parser as a dependency as its use for validating the input schema could add +200 ms of latency in some cases. Textractor-only parsing takes <30ms.

Breaking changes

• Remove linearize_table and linearize_key_value from TextLinearizationConfig as both were not used
• Remove the s3_output_path parameter from analyze_expense as the API does not support outputting to S3

New Contributors

• @mdscruggs made their first contribution in #292

Full Changelog: v1.6.1...v1.7.0

Version 1.6.1

What's new

• Fix bug in table to markdown

Full Changelog: v1.6.0...v1.6.1