from __future__ import annotations import json from pathlib import Path from ocrtech.manifest import ManifestEntry, sha256_file, sha256_text, write_manifest from ocrtech.schemas import BBox, Document, Page, Table, TableCell from ocrtech.table_analysis import analyze_table_cells def test_analyze_table_cells_reads_benchmark_documents(tmp_path: Path) -> None: image = tmp_path / "sample.png" image.write_bytes(b"image") reference = tmp_path / "reference.json" reference.write_text( json.dumps({"text": "ignored", "tables": [[["नाम", "रकम"], ["राम", "१२"]]]}, ensure_ascii=False) + "\n", encoding="utf-8", ) text_hash = sha256_text("ignored") manifest = tmp_path / "eval.jsonl" write_manifest( [ ManifestEntry( sample_id="sample-1", dataset="fixture", split="eval", image_path=str(image), text="ignored", sha256=sha256_file(image), metadata={"reference_path": str(reference), "text_sha256": text_hash}, ) ], manifest, ) benchmark_dir = tmp_path / "benchmark" rows = [] for system, grid in { "candidate": [["नाम", "रकम"], ["राम", ""]], "surya": [["नाम", "रकम"], ["श्याम", "१२"]], "tesseract": [["नाम", "रकम"], ["राम", "१२"]], }.items(): output_dir = benchmark_dir / system / image.stem output_dir.mkdir(parents=True) document = _document(image, grid, duplicate_table=system == "surya") (output_dir / "document.json").write_text(json.dumps(document.to_dict(), ensure_ascii=False, indent=2) + "\n", encoding="utf-8") rows.append( { "baseline": system, "input_path": str(image), "status": "ok", "latency_seconds": 0.1, "sample_id": "sample-1", "slices": ["table"], "metrics": {}, } ) report = benchmark_dir / "report.json" report.write_text(json.dumps(rows, ensure_ascii=False, indent=2) + "\n", encoding="utf-8") analysis = analyze_table_cells(report, manifest, tmp_path / "analysis", candidate="candidate", baselines=["surya", "tesseract"]) assert analysis.systems["candidate"].reference_cells == 4 assert analysis.systems["candidate"].reference_tables == 1 assert analysis.systems["candidate"].predicted_tables == 1 assert analysis.systems["candidate"].exact_cells == 3 assert analysis.systems["candidate"].missing_cells == 1 assert analysis.systems["surya"].predicted_tables == 2 assert analysis.systems["surya"].duplicate_overlapping_tables == 1 assert analysis.systems["tesseract"].exact_cells == 4 assert analysis.paired["surya"].candidate_better == 1 assert analysis.paired["surya"].baseline_better == 1 assert analysis.paired["tesseract"].baseline_exact_candidate_not == 1 candidate_sample = next(item for item in analysis.samples if item.system == "candidate") assert candidate_sample.reference_shapes == [[2, 2]] assert candidate_sample.predicted_shapes == [[2, 2]] assert candidate_sample.exact_cells == 3 surya_sample = next(item for item in analysis.samples if item.system == "surya") assert surya_sample.predicted_tables == 2 assert surya_sample.duplicate_overlapping_tables == 1 assert (tmp_path / "analysis" / "table-cell-analysis.md").exists() def _document(image: Path, grid: list[list[str]], *, duplicate_table: bool = False) -> Document: cells = [ TableCell(row=row_index, col=col_index, text=value, bbox=BBox(col_index * 10, row_index * 10, 10, 10)) for row_index, row in enumerate(grid) for col_index, value in enumerate(row) ] tables = [Table(table_id="p0-t0", page_index=0, bbox=BBox(0, 0, 20, 20), cells=cells)] if duplicate_table: tables.append( Table( table_id="p0-t1", page_index=0, bbox=BBox(0, 0, 20, 20), cells=[TableCell(row=0, col=0, text="".join(value for row in grid for value in row), bbox=BBox(0, 0, 20, 20))], ) ) return Document( source_path=str(image), pages=[Page(page_index=0)], tables=tables, )