feat: validate python-bidi before initializing Arabic OCR (issue #719)

python/pyproject.toml

@@ -42,6 +42,9 @@ Changelog = "https://github.com/RapidAI/RapidOCR/releases"
 [project.scripts]
 rapidocr = "rapidocr.main:main"
 
+[project.optional-dependencies]
+rtl = ["python-bidi>=0.4.2,<1.0"]
+
 [tool.setuptools]
 include-package-data = true
 platforms = ["Any"]

python/rapidocr/ch_ppocr_rec/main.py

@@ -24,7 +24,8 @@ from rapidocr.inference_engine.base import FileInfo, get_engine
 from ..utils.download_file import DownloadFile, DownloadFileInput
 from ..utils.log import logger
 from ..utils.model_resolver import normalize_lang
-from ..utils.utils import reorder_bidi_for_display
+from ..utils.typings import LangRec
+from ..utils.utils import reorder_bidi_for_display, validate_rtl_dependency
 from ..utils.vis_res import VisRes
 from .typings import TextRecInput, TextRecOutput
 from .utils import CTCLabelDecode
@@ -32,10 +33,14 @@ from .utils import CTCLabelDecode
 DEFAULT_DICT_PATH = Path(__file__).parent.parent / "models" / "ppocr_keys_v1.txt"
 DEFAULT_DICT_URL = "https://www.modelscope.cn/models/RapidAI/RapidOCR/resolve/v2.0.7/paddle/PP-OCRv4/rec/ch_PP-OCRv4_rec_infer/ppocr_keys_v1.txt"
 DEFAULT_MODEL_PATH = Path(__file__).parent.parent / "models"
+RTL_LANGS = {LangRec.ARABIC.value}
 
 
 class TextRecognizer:
     def __init__(self, cfg: Dict[str, Any]):
+        if normalize_lang(cfg.lang_type) in RTL_LANGS:
+            validate_rtl_dependency()
+
         self.session = get_engine(cfg.engine_type)(cfg)
 
         # onnx has inner character, other engine get or download character_dict_path
@@ -50,8 +55,6 @@ class TextRecognizer:
 
         self.cfg = cfg
 
-        self.RTL_LANGS = {"arabic", "fa", "ur"}
-
     def get_character_dict(self, cfg):
         character = None
         dict_path = cfg.get("rec_keys_path", None)
@@ -139,7 +142,7 @@ class TextRecognizer:
         all_line_results, all_word_results = list(zip(*rec_res))
         txts, scores = list(zip(*all_line_results))
 
-        if normalize_lang(self.cfg.lang_type) in self.RTL_LANGS:
+        if normalize_lang(self.cfg.lang_type) in RTL_LANGS:
             txts = reorder_bidi_for_display(txts)
 
         elapse = time.perf_counter() - start_time

python/rapidocr/utils/utils.py

@@ -12,16 +12,22 @@ import cv2
 import numpy as np
 
 
+def validate_rtl_dependency():
+    try:
+        from bidi.algorithm import get_display  # noqa: F401
+    except ImportError as exc:
+        raise ModuleNotFoundError(
+            "RTL recognition requires the optional dependency 'python-bidi'. "
+            "Install it with: pip install 'rapidocr[rtl]'"
+        ) from exc
+
+
 def reorder_bidi_for_display(
     texts: Tuple[Union[str, bytes], ...],
 ) -> Tuple[Union[str, bytes], ...]:
-    try:
-        from bidi.algorithm import get_display
-    except ImportError as e:
-        raise ModuleNotFoundError(
-            "Required dependency 'python-bidi' is not installed. "
-            "Install it with: pip install python-bidi"
-        ) from e
+    validate_rtl_dependency()
+
+    from bidi.algorithm import get_display  # noqa: F401
 
     return tuple([get_display(text) for text in texts])
 

python/tests/test_rtl_dependency.py

@@ -0,0 +1,70 @@
+# -*- encoding: utf-8 -*-
+# @Author: SWHL
+# @Contact: liekkaskono@163.com
+import builtins
+import sys
+import types
+from pathlib import Path
+from types import SimpleNamespace
+
+import pytest
+
+root_dir = Path(__file__).resolve().parent.parent
+sys.path.append(str(root_dir))
+
+from rapidocr.ch_ppocr_rec import main as rec_main
+from rapidocr.utils import utils
+from rapidocr.utils.model_resolver import normalize_lang
+from rapidocr.utils.typings import EngineType, LangRec
+
+
+def test_validate_rtl_dependency_reports_extra(monkeypatch):
+    original_import = builtins.__import__
+
+    def missing_bidi(name, *args, **kwargs):
+        if name.startswith("bidi"):
+            raise ModuleNotFoundError("No module named 'bidi'")
+        return original_import(name, *args, **kwargs)
+
+    monkeypatch.setattr(builtins, "__import__", missing_bidi)
+
+    with pytest.raises(ModuleNotFoundError, match=r"rapidocr\[rtl\]"):
+        utils.validate_rtl_dependency()
+
+
+def test_reorder_bidi_for_display_uses_bidi_algorithm(monkeypatch):
+    bidi_module = types.ModuleType("bidi")
+    algorithm_module = types.ModuleType("bidi.algorithm")
+    algorithm_module.get_display = lambda text: text[::-1]
+    bidi_module.algorithm = algorithm_module
+    monkeypatch.setitem(sys.modules, "bidi", bidi_module)
+    monkeypatch.setitem(sys.modules, "bidi.algorithm", algorithm_module)
+
+    assert utils.reorder_bidi_for_display(("abc", "123")) == ("cba", "321")
+
+
+def test_arabic_dependency_is_checked_before_engine_initialization(monkeypatch):
+    engine_called = False
+
+    def fail_get_engine(_engine_type):
+        nonlocal engine_called
+        engine_called = True
+        raise AssertionError(
+            "engine initialization must happen after dependency validation"
+        )
+
+    def missing_dependency():
+        raise ModuleNotFoundError("RTL dependency is missing")
+
+    monkeypatch.setattr(rec_main, "get_engine", fail_get_engine)
+    monkeypatch.setattr(rec_main, "validate_rtl_dependency", missing_dependency)
+
+    cfg = SimpleNamespace(lang_type=LangRec.ARABIC, engine_type=EngineType.ONNXRUNTIME)
+    with pytest.raises(ModuleNotFoundError, match="RTL dependency is missing"):
+        rec_main.TextRecognizer(cfg)
+
+    assert engine_called is False
+
+
+def test_arabic_enum_is_normalized_for_rtl_detection():
+    assert normalize_lang(LangRec.ARABIC) in rec_main.RTL_LANGS