v0.5.3

Changes

• Fix problems with custom tasks (#21)
• Add tests for custom tasks subclass
• Improve tests organization

v0.5.2

Changes

• @features decorator more flexible, don't requiring pyarrow_type for helper features #19
• @features decorator now supports receiving arguments to be passed to pyarrow_type #20
• number of total tasks when resuming processing is now fixed #18
• Don't use any kind of pools when using num_cpus=1

Known issues

• Problem with memoization when implementing custom tasks class #21

v0.5.1

Changes

• Add PIL.Image.DecompressionBombError to expected exceptions when extracting image #15
• Progress bar showing before first results arrive #14
• Progress bar starts showing skipped files

Rework base structure

Just renamed from v4.0.0 which was wrong!

Changes

• Refactor most of the code (including tests) structure to be scalable on the number of extracted features
• Add support to specify custom features through inheritance
• Add image feature
• Support multiple params to customize text and image extractions (image size, ocr image size, image format, etc)
• Update dependencies
• General small fixes

Known issues

• Saving progress had to be disabled yet for this release #8 . Will be fixed on the next one

v0.5.0

Changes

• Rework "resuming progress" feature
• Add support to receiving a list of files to be processed
• Improves code quality

Fix performance

Changes

• v0.4.0 caused some problems with performance, this fix them

Small improvements

Changes

• Raise exception for invalid input_dir
• Add maximum chunksize default constraint

Bug Fix

Changes

• Raise exceptions for invalid page numbers when specifying tasks

New features

Changes

• Add ability to pass specific tasks to be calculated
• Add ability to return a list instead of pandas DataFrame

Fix High Memory Usage

Changes

• Fix high memory usage caused by last release #3
• Chunksize was not really being calculated