88e91a5900
- OcrService 提取 PDF 文本层后若有效字符达到阈值，直接构建文档并写入结果缓存，不再触发 OCR worker，仅无文本层时才解析 python_bin/worker_path 调用 worker - _build_text_layer_document 复用 AggregatedOcrDocument 聚合文本层片段，_has_usable_pdf_text_layer 基于 meaningful_char_count 判定 - docker-compose 与 paddleocr bootstrap 脚本补装 poppler-data，保证 PDF 文本层抽取的中文编码正确 - 新增文本层直取与运行时依赖两项 ocr_service 单测
421 lines
15 KiB
Python
421 lines
15 KiB
Python
from __future__ import annotations
|
||
|
||
import stat
|
||
import subprocess
|
||
from pathlib import Path
|
||
|
||
from app.core.config import get_settings
|
||
from app.services.ocr import OcrService
|
||
|
||
|
||
def test_ocr_runtime_installers_include_poppler_cjk_data() -> None:
|
||
repo_root = Path(__file__).resolve().parents[2]
|
||
dependency_sources = [
|
||
repo_root / "docker-compose.yml",
|
||
repo_root / "server" / "scripts" / "bootstrap_paddleocr_mobile.sh",
|
||
repo_root / "server" / "scripts" / "bootstrap_paddleocr_gpu.sh",
|
||
]
|
||
|
||
for path in dependency_sources:
|
||
assert "poppler-data" in path.read_text(encoding="utf-8")
|
||
|
||
|
||
def test_ocr_service_uses_worker_runtime_and_keeps_unsupported_files_as_warnings(
|
||
monkeypatch,
|
||
tmp_path: Path,
|
||
) -> None:
|
||
fake_python = tmp_path / "fake-ocr-python.py"
|
||
fake_python.write_text(
|
||
"""#!/usr/bin/env python3
|
||
import json
|
||
import sys
|
||
|
||
inputs = []
|
||
for index, arg in enumerate(sys.argv):
|
||
if arg == "--input" and index + 1 < len(sys.argv):
|
||
input_path = sys.argv[index + 1]
|
||
inputs.append(
|
||
{
|
||
"input_path": input_path,
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"text": "增值税电子发票 发票号码12345678 金额 100 元 2026-05-13",
|
||
"summary": "增值税电子发票，金额 100 元。",
|
||
"avg_score": 0.98,
|
||
"line_count": 1,
|
||
"page_count": 1,
|
||
"warnings": [],
|
||
"lines": [
|
||
{
|
||
"text": "增值税电子发票 发票号码12345678 金额 100 元 2026-05-13",
|
||
"score": 0.98,
|
||
"box": [[1, 2], [10, 2], [10, 8], [1, 8]],
|
||
"page_index": 0,
|
||
}
|
||
],
|
||
}
|
||
)
|
||
|
||
payload = {
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"documents": inputs,
|
||
}
|
||
print("__OCR_JSON__=" + json.dumps(payload, ensure_ascii=False))
|
||
""",
|
||
encoding="utf-8",
|
||
)
|
||
fake_python.chmod(fake_python.stat().st_mode | stat.S_IEXEC)
|
||
|
||
monkeypatch.setenv("OCR_PYTHON_BIN", str(fake_python))
|
||
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
|
||
get_settings.cache_clear()
|
||
try:
|
||
result = OcrService().recognize_files(
|
||
[
|
||
("invoice.png", b"fake-image", "image/png"),
|
||
("notes.txt", b"plain-text", "text/plain"),
|
||
]
|
||
)
|
||
finally:
|
||
get_settings.cache_clear()
|
||
|
||
assert result.engine == "paddleocr_mobile"
|
||
assert result.model == "PP-OCRv5_mobile"
|
||
assert result.total_file_count == 2
|
||
assert result.success_count == 1
|
||
assert len(result.documents) == 2
|
||
|
||
recognized = next(item for item in result.documents if item.filename == "invoice.png")
|
||
assert recognized.summary == "增值税电子发票，金额 100 元。"
|
||
assert recognized.line_count == 1
|
||
assert recognized.document_type == "vat_invoice"
|
||
assert recognized.document_type_label == "增值税发票"
|
||
assert any(field.label == "金额" and field.value == "100元" for field in recognized.document_fields)
|
||
assert any(field.label == "票据号码" and field.value == "12345678" for field in recognized.document_fields)
|
||
assert any(field.label == "日期" and field.value == "2026-05-13" for field in recognized.document_fields)
|
||
assert recognized.lines[0].text == "增值税电子发票 发票号码12345678 金额 100 元 2026-05-13"
|
||
|
||
skipped = next(item for item in result.documents if item.filename == "notes.txt")
|
||
assert skipped.line_count == 0
|
||
assert skipped.warnings == ["当前仅支持图片和 PDF 文件进行 OCR。"]
|
||
|
||
|
||
def test_ocr_service_passes_configured_device_to_worker(
|
||
monkeypatch,
|
||
tmp_path: Path,
|
||
) -> None:
|
||
captured_commands: list[list[str]] = []
|
||
|
||
def fake_run(
|
||
command: list[str],
|
||
*,
|
||
capture_output: bool,
|
||
text: bool,
|
||
timeout: int,
|
||
check: bool,
|
||
) -> subprocess.CompletedProcess[str]:
|
||
captured_commands.append(command)
|
||
return subprocess.CompletedProcess(
|
||
args=command,
|
||
returncode=0,
|
||
stdout='__OCR_JSON__={"engine":"paddleocr_mobile","model":"PP-OCRv5_mobile","documents":[]}\n',
|
||
stderr="",
|
||
)
|
||
|
||
monkeypatch.setenv("OCR_DEVICE", "gpu:0")
|
||
get_settings.cache_clear()
|
||
monkeypatch.setattr(subprocess, "run", fake_run)
|
||
try:
|
||
payload = OcrService()._invoke_worker(
|
||
python_bin="python",
|
||
worker_path="worker.py",
|
||
input_paths=[tmp_path / "invoice.png"],
|
||
)
|
||
finally:
|
||
get_settings.cache_clear()
|
||
|
||
assert payload["engine"] == "paddleocr_mobile"
|
||
command = captured_commands[0]
|
||
device_index = command.index("--device")
|
||
assert command[device_index + 1] == "gpu:0"
|
||
|
||
|
||
def test_ocr_service_converts_pdf_to_images_and_returns_image_preview(
|
||
monkeypatch,
|
||
tmp_path: Path,
|
||
) -> None:
|
||
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> list[Path]:
|
||
first = output_dir / "page-1.png"
|
||
second = output_dir / "page-2.png"
|
||
first.write_bytes(b"fake-page-1")
|
||
second.write_bytes(b"fake-page-2")
|
||
return [first, second]
|
||
|
||
def fake_invoke_worker(
|
||
self,
|
||
*,
|
||
python_bin: str,
|
||
worker_path: str,
|
||
input_paths: list[Path],
|
||
) -> dict:
|
||
assert [path.name for path in input_paths] == ["page-1.png", "page-2.png"]
|
||
return {
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"documents": [
|
||
{
|
||
"input_path": str(input_paths[0]),
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"text": "高铁票 深圳北-广州南 车次 G1234 2026-05-13 金额 188 元",
|
||
"summary": "高铁票第一页",
|
||
"avg_score": 0.97,
|
||
"line_count": 1,
|
||
"page_count": 1,
|
||
"warnings": [],
|
||
"lines": [
|
||
{
|
||
"text": "高铁票 深圳北-广州南 车次 G1234 2026-05-13 金额 188 元",
|
||
"score": 0.97,
|
||
"box": [[1, 2], [10, 2], [10, 8], [1, 8]],
|
||
}
|
||
],
|
||
},
|
||
{
|
||
"input_path": str(input_paths[1]),
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"text": "乘车人 张三",
|
||
"summary": "高铁票第二页",
|
||
"avg_score": 0.94,
|
||
"line_count": 1,
|
||
"page_count": 1,
|
||
"warnings": [],
|
||
"lines": [
|
||
{
|
||
"text": "乘车人 张三",
|
||
"score": 0.94,
|
||
"box": [[1, 2], [10, 2], [10, 8], [1, 8]],
|
||
}
|
||
],
|
||
},
|
||
],
|
||
}
|
||
|
||
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
|
||
monkeypatch.setattr(OcrService, "_resolve_python_bin", lambda self: "python")
|
||
monkeypatch.setattr(OcrService, "_resolve_worker_path", lambda self: "worker.py")
|
||
monkeypatch.setattr(OcrService, "_convert_pdf_to_images", fake_convert_pdf_to_images)
|
||
monkeypatch.setattr(OcrService, "_invoke_worker", fake_invoke_worker)
|
||
get_settings.cache_clear()
|
||
try:
|
||
result = OcrService().recognize_files(
|
||
[
|
||
("train-ticket.pdf", b"%PDF-1.4 fake", "application/pdf"),
|
||
]
|
||
)
|
||
finally:
|
||
get_settings.cache_clear()
|
||
|
||
assert result.success_count == 1
|
||
assert len(result.documents) == 1
|
||
recognized = result.documents[0]
|
||
assert recognized.filename == "train-ticket.pdf"
|
||
assert recognized.page_count == 2
|
||
assert recognized.preview_kind == "image"
|
||
assert recognized.preview_data_url.startswith("data:image/png;base64,")
|
||
assert recognized.document_type == "train_ticket"
|
||
assert any(field.label == "金额" and field.value == "188元" for field in recognized.document_fields)
|
||
assert any(field.label == "车次/航班" and field.value == "G1234" for field in recognized.document_fields)
|
||
assert recognized.lines[0].page_index == 0
|
||
assert recognized.lines[1].page_index == 1
|
||
|
||
|
||
def test_ocr_service_uses_pdf_text_layer_without_worker_runtime(
|
||
monkeypatch,
|
||
tmp_path: Path,
|
||
) -> None:
|
||
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> list[Path]:
|
||
page = output_dir / "page-1.png"
|
||
page.write_bytes(b"fake-rendered-page")
|
||
return [page]
|
||
|
||
def fail_resolve_python(self) -> str:
|
||
raise AssertionError("PDF 文本层可用时不应强制解析 OCR worker。")
|
||
|
||
def fail_invoke_worker(self, **kwargs) -> dict:
|
||
raise AssertionError("PDF 文本层可用时不应调用 OCR worker。")
|
||
|
||
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
|
||
monkeypatch.setattr(OcrService, "_resolve_python_bin", fail_resolve_python)
|
||
monkeypatch.setattr(OcrService, "_resolve_worker_path", lambda self: "worker.py")
|
||
monkeypatch.setattr(OcrService, "_convert_pdf_to_images", fake_convert_pdf_to_images)
|
||
monkeypatch.setattr(OcrService, "_invoke_worker", fail_invoke_worker)
|
||
monkeypatch.setattr(
|
||
OcrService,
|
||
"_extract_pdf_text_layer",
|
||
lambda self, pdf_path: (
|
||
"电子发票（铁路电子客票）\n"
|
||
"发票号码:26429165800002785705\n"
|
||
"武汉站\n"
|
||
"上海虹桥站\n"
|
||
"G458\n"
|
||
"票价:￥354.00\n"
|
||
"电子客票号:6580061086021391007342026"
|
||
),
|
||
)
|
||
get_settings.cache_clear()
|
||
try:
|
||
result = OcrService().recognize_files(
|
||
[
|
||
("train-ticket.pdf", b"%PDF-1.7 fake", "application/pdf"),
|
||
]
|
||
)
|
||
finally:
|
||
get_settings.cache_clear()
|
||
|
||
recognized = result.documents[0]
|
||
assert result.success_count == 1
|
||
assert recognized.document_type == "train_ticket"
|
||
assert "电子发票（铁路电子客票）" in recognized.text
|
||
assert "电子客票号:6580061086021391007342026" in recognized.text
|
||
assert any(field.label == "金额" and field.value == "354元" for field in recognized.document_fields)
|
||
assert recognized.preview_kind == "image"
|
||
assert recognized.preview_data_url.startswith("data:image/png;base64,")
|
||
|
||
|
||
def test_ocr_service_reuses_cached_document_for_same_content(
|
||
monkeypatch,
|
||
tmp_path: Path,
|
||
) -> None:
|
||
calls = {"count": 0}
|
||
|
||
def fake_invoke_worker(
|
||
self,
|
||
*,
|
||
python_bin: str,
|
||
worker_path: str,
|
||
input_paths: list[Path],
|
||
) -> dict:
|
||
calls["count"] += 1
|
||
return {
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"documents": [
|
||
{
|
||
"input_path": str(input_paths[0]),
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"text": "增值税电子发票 金额 20 元",
|
||
"summary": "增值税电子发票，金额 20 元。",
|
||
"avg_score": 0.97,
|
||
"line_count": 1,
|
||
"page_count": 1,
|
||
"warnings": [],
|
||
"lines": [
|
||
{
|
||
"text": "增值税电子发票 金额 20 元",
|
||
"score": 0.97,
|
||
"box": [[1, 2], [10, 2], [10, 8], [1, 8]],
|
||
}
|
||
],
|
||
}
|
||
],
|
||
}
|
||
|
||
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
|
||
monkeypatch.setattr(OcrService, "_resolve_python_bin", lambda self: "python")
|
||
monkeypatch.setattr(OcrService, "_resolve_worker_path", lambda self: "worker.py")
|
||
monkeypatch.setattr(OcrService, "_invoke_worker", fake_invoke_worker)
|
||
OcrService._result_cache.clear()
|
||
get_settings.cache_clear()
|
||
try:
|
||
first = OcrService().recognize_files([("first.png", b"same-image", "image/png")])
|
||
second = OcrService().recognize_files([("second.png", b"same-image", "image/png")])
|
||
finally:
|
||
OcrService._result_cache.clear()
|
||
get_settings.cache_clear()
|
||
|
||
assert calls["count"] == 1
|
||
assert first.documents[0].filename == "first.png"
|
||
assert second.documents[0].filename == "second.png"
|
||
assert second.documents[0].summary == first.documents[0].summary
|
||
|
||
|
||
def test_ocr_service_prefers_pdf_text_layer_when_rendered_ocr_is_placeholder_heavy(
|
||
monkeypatch,
|
||
tmp_path: Path,
|
||
) -> None:
|
||
def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> list[Path]:
|
||
page = output_dir / "page-1.png"
|
||
page.write_bytes(b"fake-page")
|
||
return [page]
|
||
|
||
def fake_invoke_worker(
|
||
self,
|
||
*,
|
||
python_bin: str,
|
||
worker_path: str,
|
||
input_paths: list[Path],
|
||
) -> dict:
|
||
return {
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"documents": [
|
||
{
|
||
"input_path": str(input_paths[0]),
|
||
"engine": "paddleocr_mobile",
|
||
"model": "PP-OCRv5_mobile",
|
||
"text": "□□□□□□\n□□□□：26319166100006175398\nG456\n□□:□354.00",
|
||
"summary": "□□□□□□；□□□□：26319166100006175398",
|
||
"avg_score": 0.88,
|
||
"line_count": 4,
|
||
"page_count": 1,
|
||
"warnings": [],
|
||
"lines": [
|
||
{
|
||
"text": "□□□□□□",
|
||
"score": 0.88,
|
||
"box": [[1, 2], [10, 2], [10, 8], [1, 8]],
|
||
}
|
||
],
|
||
}
|
||
],
|
||
}
|
||
|
||
monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
|
||
monkeypatch.setattr(OcrService, "_resolve_python_bin", lambda self: "python")
|
||
monkeypatch.setattr(OcrService, "_resolve_worker_path", lambda self: "worker.py")
|
||
monkeypatch.setattr(OcrService, "_convert_pdf_to_images", fake_convert_pdf_to_images)
|
||
monkeypatch.setattr(OcrService, "_invoke_worker", fake_invoke_worker)
|
||
monkeypatch.setattr(
|
||
OcrService,
|
||
"_extract_pdf_text_layer",
|
||
lambda self, pdf_path: (
|
||
"电子发票（铁路电子客票）\n"
|
||
"发票号码:26319166100006175398\n"
|
||
"上海虹桥站\n"
|
||
"武汉站\n"
|
||
"G456\n"
|
||
"票价:￥354.00"
|
||
),
|
||
)
|
||
get_settings.cache_clear()
|
||
try:
|
||
result = OcrService().recognize_files(
|
||
[
|
||
("train-ticket.pdf", b"%PDF-1.4 fake", "application/pdf"),
|
||
]
|
||
)
|
||
finally:
|
||
get_settings.cache_clear()
|
||
|
||
recognized = result.documents[0]
|
||
assert "电子发票（铁路电子客票）" in recognized.text
|
||
assert "上海虹桥站" in recognized.text
|
||
assert "□□□□" not in recognized.summary
|
||
assert recognized.document_type == "train_ticket"
|
||
assert recognized.preview_kind == "image"
|
||
assert recognized.preview_data_url.startswith("data:image/png;base64,")
|