fix: make paddleocr install optional (#75)

* Pin protobuf to avoid error
* Add pin note
* Bump version and release
* Stop importing paddleocr unless needed
* make paddleocr optional
* update reqs

Author: qued

CHANGELOG.md

@@ -1,3 +1,8 @@
+## 0.3.1
+
+* Pin protobuf version to avoid errors
+* Make paddleocr an extra again
+
 ## 0.3.0
 
 * Fix for text block detection

requirements/dev.txt

@@ -67,7 +67,7 @@ ipykernel==6.22.0
     #   nbclassic
     #   notebook
     #   qtconsole
-ipython==8.11.0
+ipython==8.12.0
     # via
     #   -r requirements/dev.in
     #   ipykernel
@@ -143,7 +143,7 @@ matplotlib-inline==0.1.6
     #   ipython
 mistune==2.0.5
     # via nbconvert
-nbclassic==0.5.3
+nbclassic==0.5.4
     # via notebook
 nbclient==0.7.2
     # via nbconvert
@@ -304,6 +304,8 @@ traitlets==5.9.0
     #   nbformat
     #   notebook
     #   qtconsole
+typing-extensions==4.5.0
+    # via ipython
 uri-template==1.2.0
     # via jsonschema
 wcwidth==0.2.6

setup.py

@@ -59,6 +59,22 @@
         "opencv-python==4.6.0.66",
         "onnxruntime",
         "transformers",
-        'unstructured.PaddleOCR ; platform_machine=="x86_64"',
     ],
+    extras_require={
+        "tables": [
+            'unstructured.PaddleOCR ; platform_machine=="x86_64"',
+            # NOTE(crag): workaround issue for error output below
+            # ERROR test_unstructured/partition/test_common.py - TypeError: Descriptors cannot not
+            # be created directly.
+            # If this call came from a _pb2.py file, your generated code is out of date and must be
+            # regenerated with protoc >= 3.19.0.
+            # If you cannot immediately regenerate your protos, some other possible workarounds are:
+            #  1. Downgrade the protobuf package to 3.20.x or lower.
+            #  2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python
+            #     parsing and will be much slower).
+            "protobuf<3.21",
+            # NOTE(alan): Pin to get around error: undefined symbol: _dl_sym, version GLIBC_PRIVATE
+            "paddlepaddle>=2.4",
+        ]
+    },
 )

unstructured_inference/__version__.py

@@ -1 +1 @@
-__version__ = "0.3.0"  # pragma: no cover
+__version__ = "0.3.1"  # pragma: no cover

unstructured_inference/models/paddle_ocr.py

@@ -1,11 +1,11 @@
-from unstructured_paddleocr import PaddleOCR
-
-paddle_ocr: PaddleOCR = None
+paddle_ocr = None  # type: ignore
 
 
 def load_agent():
     """Loads the PaddleOCR agent as a global variable to ensure that we only load it once."""
 
+    from unstructured_paddleocr import PaddleOCR
+
     global paddle_ocr
     paddle_ocr = PaddleOCR(use_angle_cls=True, lang="en", mkl_dnn=True, show_log=False)