ocr-projects-archive / ocr-tech /tests /test_apply_limbu_book_split_worksheet_review.py
prosperious's picture
Add 1000 files (OCR_INDRA/.gitignore ...)
f9bc0c1 verified
Raw History Blame Contribute Delete
11.2 kB
from __future__ import annotations
import csv
import importlib.util
import json
import sys
from argparse import Namespace
from pathlib import Path
import pytest
from PIL import Image, ImageDraw
from ocrtech.manifest import sha256_file
KNOWN_PACKET_DIR_BUG = pytest.mark.xfail(
strict=True,
raises=ValueError,
reason=(
"known bug: the wrapper writes its reviewed TSV into a new directory but does not pass the "
"original unresolved packet dir (packet_dir) to apply_unresolved_review_packet, whose audit "
"needs book-unresolved-review-packet.json there; nothing in the chain records that dir"
),
)
def _load_tool(name: str):
path = Path(__file__).resolve().parents[1] / "tools" / f"{name}.py"
spec = importlib.util.spec_from_file_location(name, path)
assert spec is not None
assert spec.loader is not None
module = importlib.util.module_from_spec(spec)
sys.modules[spec.name] = module
spec.loader.exec_module(module)
return module
def _write_book(tmp_path: Path) -> Path:
image = tmp_path / "page.png"
page = Image.new("RGB", (260, 100), "white")
draw = ImageDraw.Draw(page)
draw.rectangle([20, 24, 70, 42], fill="black")
draw.rectangle([105, 24, 150, 42], fill="black")
draw.rectangle([180, 24, 230, 42], fill="black")
page.save(image)
book = tmp_path / "book"
doc_dir = book / "book-documents" / "p001"
doc_dir.mkdir(parents=True)
document = {
"schema_version": "1.0",
"source_path": str(image),
"pages": [
{
"page_index": 0,
"width": 260,
"height": 100,
"text_lines": [
{
"line_id": "line-1",
"page_index": 0,
"text": "old merged text",
"bbox": [10, 18, 230, 32],
"confidence": 0.6,
"metadata": {"structural_role": "text"},
}
],
"blocks": [
{
"block_id": "block-1",
"block_type": "text",
"page_index": 0,
"bbox": [10, 18, 230, 32],
"order": 1,
"text": "old merged text",
"confidence": 0.6,
"line_ids": ["line-1"],
"metadata": {"structural_role": "text"},
}
],
}
],
"tables": [],
"figures": [],
"metadata": {},
}
document_json = doc_dir / "document.json"
document_json.write_text(json.dumps(document, ensure_ascii=False) + "\n", encoding="utf-8")
(doc_dir / "document.md").write_text("old merged text\n", encoding="utf-8")
(doc_dir / "document.body.md").write_text("old merged text\n", encoding="utf-8")
(book / "book-run.json").write_text(
json.dumps(
{
"tool": "test",
"output_dir": str(book),
"page_count": 1,
"line_count": 1,
"review_line_count": 0,
"script_counts": {"devanagari_limbu": 1},
"book_md": str(book / "book.md"),
"book_body_md": str(book / "book.body.md"),
"pages": [
{
"page_number": 1,
"page_id": "p001",
"book_document_json": str(document_json),
"book_document_dir": str(doc_dir),
"score_document_sample_id": "p001",
"line_count": 1,
"review_line_count": 0,
"script_counts": {"devanagari_limbu": 1},
}
],
},
ensure_ascii=False,
)
+ "\n",
encoding="utf-8",
)
(book / "book.md").write_text("# Limbu Book OCR\n\n## Page 1: p001\n\nold merged text\n", encoding="utf-8")
(book / "book.body.md").write_text("# Limbu Book OCR\n\n## Page 1: p001\n\nold merged text\n", encoding="utf-8")
return book
def _write_prefilled_packet(tmp_path: Path, book: Path) -> Path:
crop = tmp_path / "crop.png"
Image.open(tmp_path / "page.png").crop((10, 18, 240, 50)).save(crop)
overlay = tmp_path / "overlay.png"
Image.open(tmp_path / "page.png").crop((10, 18, 240, 50)).save(overlay)
document_json = book / "book-documents" / "p001" / "document.json"
packet = tmp_path / "prefilled.tsv"
with packet.open("w", encoding="utf-8", newline="") as handle:
fieldnames = [
"packet_row_id",
"review_item_type",
"book_page_number",
"book_page_id",
"book_document",
"book_line_id",
"candidate_row_id",
"candidate_line_id",
"candidate_structural_role",
"review_category",
"suggested_review_action",
"confidence",
"bbox",
"text",
"source_texts",
"review_reasons",
"crop_path",
"crop_sha256",
"review_status",
"corrected_text",
"split_corrected_texts",
"split_bboxes",
"reviewer",
"notes",
"split_suggestion_overlay_path",
"split_suggestion_overlay_sha256",
"split_suggestion_segment_count",
]
writer = csv.DictWriter(handle, fieldnames=fieldnames, delimiter="\t", lineterminator="\n")
writer.writeheader()
writer.writerow(
{
"packet_row_id": "existing-1",
"review_item_type": "existing_book_line",
"book_page_number": "1",
"book_page_id": "p001",
"book_document": str(document_json),
"book_line_id": "line-1",
"review_category": "existing_low_confidence_line",
"suggested_review_action": "correct_existing_line_or_accept",
"confidence": "0.6",
"bbox": "[10,18,230,32]",
"text": "old merged text",
"source_texts": "",
"review_reasons": "low_confidence",
"crop_path": str(crop),
"crop_sha256": sha256_file(crop),
"review_status": "pending",
"corrected_text": "",
"split_corrected_texts": "",
"split_bboxes": json.dumps([[20, 24, 50, 18], [105, 24, 45, 18], [180, 24, 50, 18]], ensure_ascii=False),
"reviewer": "",
"notes": "",
"split_suggestion_overlay_path": str(overlay),
"split_suggestion_overlay_sha256": sha256_file(overlay),
"split_suggestion_segment_count": "3",
}
)
return packet
def _write_worksheet(tmp_path: Path, packet: Path, *, reviewed: bool) -> Path:
prepare = _load_tool("prepare_limbu_book_split_transcription_worksheet")
report = prepare.prepare_worksheet(
Namespace(
prefilled_packet=packet,
out=tmp_path / "worksheet",
segment_padding=2,
include_reviewed=False,
)
)
report_path = Path(report["report_json"])
if reviewed:
worksheet_tsv = Path(report["worksheet_tsv"])
with worksheet_tsv.open("r", encoding="utf-8", newline="") as handle:
reader = csv.DictReader(handle, delimiter="\t")
fieldnames = reader.fieldnames
rows = list(reader)
assert fieldnames is not None
edits = {"1": ("text", "पहिलो"), "2": ("non_text", ""), "3": ("verified", "दोस्रो")}
for row in rows:
row["segment_status"], row["segment_text"] = edits[row["segment_index"]]
with worksheet_tsv.open("w", encoding="utf-8", newline="") as handle:
writer = csv.DictWriter(handle, fieldnames=fieldnames, delimiter="\t", lineterminator="\n")
writer.writeheader()
writer.writerows(rows)
payload = json.loads(report_path.read_text(encoding="utf-8"))
payload["worksheet_tsv_sha256"] = sha256_file(worksheet_tsv)
report_path.write_text(json.dumps(payload, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
return report_path
@KNOWN_PACKET_DIR_BUG
def test_apply_limbu_book_split_worksheet_review_applies_reviewed_split(tmp_path: Path) -> None:
tool = _load_tool("apply_limbu_book_split_worksheet_review")
book = _write_book(tmp_path)
packet = _write_prefilled_packet(tmp_path, book)
worksheet_report = _write_worksheet(tmp_path, packet, reviewed=True)
report = tool.apply_split_worksheet_review(
Namespace(
book_output_dir=book,
prefilled_packet=packet,
worksheet_report=worksheet_report,
out=tmp_path / "reviewed-application",
mark_status="corrected",
accepted_statuses=["corrected"],
text_statuses=["text", "verified"],
drop_statuses=["non_text"],
)
)
assert report["compiled_count"] == 1
assert report["accepted_counts"] == {"existing_book_line": 1}
assert report["book_line_applied_count"] == 1
assert report["candidate_line_applied_count"] == 0
assert report["application_audit_passed"] is True
final_book = Path(report["application_dir"]) / "final-book"
book_md = (final_book / "book.md").read_text(encoding="utf-8")
assert "पहिलो" in book_md
assert "दोस्रो" in book_md
assert "old merged text" not in book_md
final_run = json.loads((final_book / "book-run.json").read_text(encoding="utf-8"))
assert final_run["line_count"] == 2
assert final_run["pages"][0]["book_line_review_split_count"] == 1
@KNOWN_PACKET_DIR_BUG
def test_apply_limbu_book_split_worksheet_review_keeps_pending_noop(tmp_path: Path) -> None:
tool = _load_tool("apply_limbu_book_split_worksheet_review")
book = _write_book(tmp_path)
packet = _write_prefilled_packet(tmp_path, book)
worksheet_report = _write_worksheet(tmp_path, packet, reviewed=False)
report = tool.apply_split_worksheet_review(
Namespace(
book_output_dir=book,
prefilled_packet=packet,
worksheet_report=worksheet_report,
out=tmp_path / "pending-application",
mark_status="corrected",
accepted_statuses=["corrected"],
text_statuses=["text", "verified"],
drop_statuses=["non_text"],
)
)
assert report["compiled_count"] == 0
assert report["accepted_counts"] == {}
assert report["book_line_applied_count"] == 0
assert report["candidate_line_applied_count"] == 0
assert report["application_audit_passed"] is True
final_book = Path(report["application_dir"]) / "final-book"
assert "old merged text" in (final_book / "book.md").read_text(encoding="utf-8")