bramp / build-along / 19198730038

89%

Build Type

push

github

Committed by bramp

Commit Message

Refactor FontSizeHints to use element ID-based page classification

This refactoring improves page classification and simplifies the code by using TextHistogram abstractions properly.

Key changes:

1. Replace position-based page split (2/3-1/3) with content-based classification using catalog element IDs (6-7 digit numbers): Pages with >3 element IDs are classified as catalog pages, all other pages are instruction pages.

2. Add TextHistogram.empty() and TextHistogram.update() methods: empty() creates an empty histogram for accumulation, update() merges another histogram's counts (like Counter.update()), simplifies from_pages() to use these new methods.

3. Use three histograms for proper separation of concerns: instruction_histogram for instruction-specific metrics, catalog_histogram for catalog-specific metrics, all_histogram for document-wide metrics (page numbers, remaining fonts).

4. Simplify histogram accumulation: Build histograms page-by-page with direct updates, no intermediate lists or manual counter combining, check page counts instead of None to determine if histograms exist.

5. Fix page_number_size and remaining_font_sizes extraction: Both now use all_histogram (includes all pages) for more accurate statistics from complete dataset.

Benefits: More accurate content-based classification, cleaner code with proper TextHistogram abstractions, better maintainability with single source of truth for element ID detection, correct semantics where document-wide metrics use all pages.

Run Details

66 of 66 new or added lines in 4 files covered. (100.0%)

2 existing lines in 2 files now uncovered.

4519 of 5261 relevant lines covered (85.9%)

0.86 hits per line