bramp / build-along / 19850699447

89%

Build Type

push

github

Committed by bramp

Commit Message

feat: Implement rule-based classifier system and refactor key classifiers

Introduces a flexible, rule-based architecture for PDF element classification,
significantly reducing boilerplate and improving maintainability.

Key changes include:
- **New `RuleBasedClassifier` base class**: Provides a declarative way to define
  classifier logic using a list of `Rule` objects.
- **Generic and specific `Rule` implementations**:
  - `Filter` base class for pass/fail rules (`IsInstanceFilter`, `InBottomBandFilter`).
  - Scoring rules like `RegexMatch`, `FontSizeMatch`, `CornerDistanceScore`,
    `PageNumberValueMatch`.
  - Aggregation rules like `MaxScoreRule` for combining multiple scoring criteria.
  - Domain-specific rules: `PageNumberTextRule`, `PartCountTextRule`,
    `PartNumberTextRule`, `BagNumberTextRule`, `TopLeftPositionScore`,
    `BagNumberFontSizeRule`, and `StepNumberTextRule`.
- **Refactored Classifiers**: Converted `PageNumberClassifier`,
  `PartCountClassifier`, `StepNumberClassifier`, `PartNumberClassifier`, and
  `BagNumberClassifier` to use the new rule-based system.
- **Centralized Configuration**: `min_score` and weights for `BagNumberClassifier`
  are now configurable via `BagNumberConfig`.
- **Improved Test Fixtures**: Updated `conftest.py` to support the new `RuleScore`.
- **Golden File Updates**: Golden test fixtures were regenerated to match the
  new (and in some cases, improved) classification behavior. Notably, some pages
  previously misclassified as "instruction" are now correctly identified as "catalog"
  due to enhanced `StepNumberClassifier` logic.

This refactoring streamlines the addition and modification of classification
heuristics, making the system more robust and easier to debug.

Run Details

341 of 365 new or added lines in 12 files covered. (93.42%)

11 existing lines in 2 files now uncovered.

10298 of 11389 relevant lines covered (90.42%)

0.9 hits per line