omdev 0.0.63__tar.gz → 0.0.64__tar.gz

{omdev-0.0.63/omdev.egg-info → omdev-0.0.64}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: omdev
-Version: 0.0.63
+Version: 0.0.64
 Summary: omdev
 Author: wrmsr
 License-Expression: BSD-3-Clause
@@ -14,21 +14,21 @@ Classifier: Programming Language :: Python :: 3.15
 Requires-Python: >=3.14
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: omcore==0.0.63
+Requires-Dist: omcore==0.0.64
 Provides-Extra: ominfra
-Requires-Dist: ominfra==0.0.63; extra == "ominfra"
+Requires-Dist: ominfra==0.0.64; extra == "ominfra"
 Provides-Extra: omllm
-Requires-Dist: omllm==0.0.63; extra == "omllm"
+Requires-Dist: omllm==0.0.64; extra == "omllm"
 Provides-Extra: black
 Requires-Dist: black~=26.10; extra == "black"
 Provides-Extra: c
-Requires-Dist: pycparser~=3.0; extra == "c"
+Requires-Dist: pycparser~=3.1; extra == "c"
 Requires-Dist: pcpp~=1.30; extra == "c"
 Provides-Extra: doc
 Requires-Dist: docutils~=0.23; extra == "doc"
 Provides-Extra: ocr
-Requires-Dist: pytesseract~=0.3; extra == "ocr"
-Requires-Dist: rapidocr-onnxruntime~=1.4; extra == "ocr"
+Requires-Dist: rapidocr~=3.10; extra == "ocr"
+Requires-Dist: onnxruntime~=1.30; extra == "ocr"
 Provides-Extra: pillow
 Requires-Dist: pillow~=12.3; extra == "pillow"
 Provides-Extra: prof
@@ -57,20 +57,20 @@ Requires-Dist: tree-sitter-toml~=0.7; extra == "tui-syntax"
 Requires-Dist: tree-sitter-xml~=0.7; extra == "tui-syntax"
 Requires-Dist: tree-sitter-yaml~=0.7; extra == "tui-syntax"
 Provides-Extra: cext
-Requires-Dist: omdev-cext==0.0.63; extra == "cext"
+Requires-Dist: omdev-cext==0.0.64; extra == "cext"
 Provides-Extra: rs
-Requires-Dist: omdev-rs==0.0.63; extra == "rs"
+Requires-Dist: omdev-rs==0.0.64; extra == "rs"
 Provides-Extra: cli
-Requires-Dist: omdev-cli==0.0.63; extra == "cli"
+Requires-Dist: omdev-cli==0.0.64; extra == "cli"
 Provides-Extra: all
-Requires-Dist: ominfra==0.0.63; extra == "all"
-Requires-Dist: omllm==0.0.63; extra == "all"
+Requires-Dist: ominfra==0.0.64; extra == "all"
+Requires-Dist: omllm==0.0.64; extra == "all"
 Requires-Dist: black~=26.10; extra == "all"
-Requires-Dist: pycparser~=3.0; extra == "all"
+Requires-Dist: pycparser~=3.1; extra == "all"
 Requires-Dist: pcpp~=1.30; extra == "all"
 Requires-Dist: docutils~=0.23; extra == "all"
-Requires-Dist: pytesseract~=0.3; extra == "all"
-Requires-Dist: rapidocr-onnxruntime~=1.4; extra == "all"
+Requires-Dist: rapidocr~=3.10; extra == "all"
+Requires-Dist: onnxruntime~=1.30; extra == "all"
 Requires-Dist: pillow~=12.3; extra == "all"
 Requires-Dist: gprof2dot~=2025.4; extra == "all"
 Requires-Dist: basedpyright~=1.40; extra == "all"
@@ -92,9 +92,9 @@ Requires-Dist: tree-sitter-sql~=0.3; extra == "all"
 Requires-Dist: tree-sitter-toml~=0.7; extra == "all"
 Requires-Dist: tree-sitter-xml~=0.7; extra == "all"
 Requires-Dist: tree-sitter-yaml~=0.7; extra == "all"
-Requires-Dist: omdev-cext==0.0.63; extra == "all"
-Requires-Dist: omdev-rs==0.0.63; extra == "all"
-Requires-Dist: omdev-cli==0.0.63; extra == "all"
+Requires-Dist: omdev-cext==0.0.64; extra == "all"
+Requires-Dist: omdev-rs==0.0.64; extra == "all"
+Requires-Dist: omdev-cli==0.0.64; extra == "all"
 Dynamic: license-file
 
 # Overview

{omdev-0.0.63 → omdev-0.0.64}/omdev/.om-manifests.json

@@ -195,6 +195,18 @@
       }
     }
   },
+  {
+    "module": ".ocr.__main__",
+    "attr": "_CLI_MODULE",
+    "file": "omdev/ocr/__main__.py",
+    "line": 4,
+    "value": {
+      "!.cli.types.CliModule": {
+        "name": "ocr",
+        "module": "omdev.ocr.__main__"
+      }
+    }
+  },
   {
     "module": ".pipdeptree.__main__",
     "attr": "_CLI_MODULE",
@@ -486,18 +498,6 @@
       }
     }
   },
-  {
-    "module": ".tools.ocr",
-    "attr": "_CLI_MODULE",
-    "file": "omdev/tools/ocr.py",
-    "line": 166,
-    "value": {
-      "!.cli.types.CliModule": {
-        "name": "ocr",
-        "module": "omdev.tools.ocr"
-      }
-    }
-  },
   {
     "module": ".tools.pawk.pawk",
     "attr": "_CLI_MODULE",

{omdev-0.0.63 → omdev-0.0.64}/omdev/__about__.py

@@ -27,7 +27,7 @@ class Project(ProjectBase):
         ],
 
         'c': [
-            'pycparser ~= 3.0',
+            'pycparser ~= 3.1',
 
             'pcpp ~= 1.30',
         ],
@@ -37,9 +37,8 @@ class Project(ProjectBase):
         ],
 
         'ocr': [
-            'pytesseract ~= 0.3',
-
-            'rapidocr-onnxruntime ~= 1.4',
+            'rapidocr ~= 3.10',
+            'onnxruntime ~= 1.30',
         ],
 
         'pillow': [

omdev-0.0.64/omdev/capi/darwin/vn.py

@@ -0,0 +1,280 @@
+"""
+Only the Objective-C calls needed for synchronous Vision text recognition of encoded image bytes.
+
+No Python callbacks, Objective-C blocks, structure-return messages, or general-purpose object bridge are used. Objects
+returned by alloc/init/new are released explicitly; other temporaries live in a per-call autorelease pool. NSError is
+translated to VisionError. Objective-C exceptions cannot be unwound safely through ctypes.
+"""
+import contextlib
+import ctypes as ct
+import os
+import sys
+import typing as ta
+
+from omcore import lang
+
+
+##
+
+
+class VisionError(Exception):
+    pass
+
+
+class _Vision:
+    def __init__(self) -> None:
+        super().__init__()
+
+        if getattr(sys, 'platform') != 'darwin':
+            raise OSError(sys.platform)
+
+        # Retain the library handles for at least as long as the bound callables.
+        self._foundation = ct.CDLL('/System/Library/Frameworks/Foundation.framework/Foundation')
+        self._vision = ct.CDLL('/System/Library/Frameworks/Vision.framework/Vision')
+        self._objc = ct.CDLL('/usr/lib/libobjc.A.dylib')
+
+        self._get_class = self._objc.objc_getClass
+        self._get_class.argtypes = [ct.c_char_p]
+        self._get_class.restype = ct.c_void_p
+
+        self._get_selector = self._objc.sel_registerName
+        self._get_selector.argtypes = [ct.c_char_p]
+        self._get_selector.restype = ct.c_void_p
+
+        # Each prototype is a distinct immutable callable. Never mutate objc_msgSend.argtypes/restype between calls:
+        # besides being the wrong ABI for some signatures, that would race on free-threaded Python.
+        # Apple's BOOL is C bool on arm64 and signed char on Intel macOS.
+        objc_bool = ct.c_bool if os.uname().machine == 'arm64' else ct.c_byte
+        symbol = ('objc_msgSend', self._objc)
+
+        self._id: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+        )(symbol)
+
+        self._id_id: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+        )(symbol)
+
+        self._id_id_id: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+        )(symbol)
+
+        self._id_uint: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_ulong,
+        )(symbol)
+
+        self._id_bytes_size: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_ulong,
+        )(symbol)
+
+        self._id_cstr: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_char_p,
+        )(symbol)
+
+        self._uint: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_ulong,
+            ct.c_void_p,
+            ct.c_void_p,
+        )(symbol)
+
+        self._cstr: ta.Any = ct.CFUNCTYPE(  # noqa
+            ct.c_char_p,
+            ct.c_void_p,
+            ct.c_void_p,
+        )(symbol)
+
+        self._void: ta.Any = ct.CFUNCTYPE(  # noqa
+            None,
+            ct.c_void_p,
+            ct.c_void_p,
+        )(symbol)
+
+        self._void_id: ta.Any = ct.CFUNCTYPE(  # noqa
+            None,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+        )(symbol)
+
+        self._void_int: ta.Any = ct.CFUNCTYPE(  # noqa
+            None,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_long,
+        )(symbol)
+
+        self._void_bool: ta.Any = ct.CFUNCTYPE(  # noqa
+            None,
+            ct.c_void_p,
+            ct.c_void_p,
+            objc_bool,
+        )(symbol)
+
+        self._perform: ta.Any = ct.CFUNCTYPE(  # noqa
+            objc_bool,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.c_void_p,
+            ct.POINTER(ct.c_void_p),
+        )(symbol)
+
+    def _cls(self, name: bytes) -> int:
+        if not (obj := self._get_class(name)):
+            raise VisionError(f'Objective-C class not found: {name!r}')
+        return obj
+
+    def _sel(self, name: bytes) -> int:
+        return self._get_selector(name)
+
+    def _require(self, obj: int | None, description: str) -> int:
+        if not obj:
+            raise VisionError(description)
+        return obj
+
+    def _release(self, obj: int) -> None:
+        self._void(obj, self._sel(b'release'))
+
+    def _string(self, obj: int | None) -> str:
+        if not obj:
+            return ''
+        value = self._cstr(obj, self._sel(b'UTF8String'))
+        if value is None:
+            raise VisionError('NSString could not be converted to UTF-8')
+        return value.decode('utf-8')
+
+    def _make_languages(self, languages: ta.Sequence[str]) -> int:
+        array = self._require(
+            self._id(self._cls(b'NSMutableArray'), self._sel(b'array')),
+            'Could not create recognitionLanguages',
+        )
+        for language in languages:
+            string = self._require(
+                self._id_cstr(self._cls(b'NSString'), self._sel(b'stringWithUTF8String:'), language.encode('utf-8')),
+                'Could not create a recognition language string',
+            )
+            self._void_id(array, self._sel(b'addObject:'), string)
+        return array
+
+    def recognize_text(
+            self,
+            data: bytes,
+            *,
+            languages: ta.Sequence[str] | None,
+            fast: bool,
+            no_language_correction: bool,
+    ) -> str:
+        with contextlib.ExitStack() as stack:
+            pool = self._require(
+                self._id(self._cls(b'NSAutoreleasePool'), self._sel(b'new')),
+                'Could not create an autorelease pool',
+            )
+            stack.callback(self._void, pool, self._sel(b'drain'))
+
+            # dataWithBytes:length: copies the bytes; do not replace this with a no-copy initializer.
+            buffer = ct.create_string_buffer(data)
+            ns_data = self._require(
+                self._id_bytes_size(self._cls(b'NSData'), self._sel(b'dataWithBytes:length:'), buffer, len(data)),
+                'Could not create image NSData',
+            )
+            options = self._require(
+                self._id(self._cls(b'NSDictionary'), self._sel(b'dictionary')),
+                'Could not create image request options',
+            )
+
+            request = self._require(
+                self._id(self._cls(b'VNRecognizeTextRequest'), self._sel(b'new')),
+                'Could not initialize VNRecognizeTextRequest',
+            )
+            stack.callback(self._release, request)
+
+            # VNRequestTextRecognitionLevel: accurate = 0, fast = 1 (NSInteger).
+            self._void_int(request, self._sel(b'setRecognitionLevel:'), 1 if fast else 0)
+            self._void_bool(request, self._sel(b'setUsesLanguageCorrection:'), not no_language_correction)
+            if languages is not None:
+                self._void_id(request, self._sel(b'setRecognitionLanguages:'), self._make_languages(languages))
+
+            allocated_handler = self._require(
+                self._id(self._cls(b'VNImageRequestHandler'), self._sel(b'alloc')),
+                'Could not allocate VNImageRequestHandler',
+            )
+            # Only the result of init is owned: init may replace the allocated object or release it and return nil.
+            handler = self._require(
+                self._id_id_id(allocated_handler, self._sel(b'initWithData:options:'), ns_data, options),
+                'Could not initialize VNImageRequestHandler',
+            )
+            stack.callback(self._release, handler)
+
+            requests = self._require(
+                self._id_id(self._cls(b'NSArray'), self._sel(b'arrayWithObject:'), request),
+                'Could not create the request array',
+            )
+            error = ct.c_void_p()
+            if not self._perform(handler, self._sel(b'performRequests:error:'), requests, ct.byref(error)):
+                description = self._string(self._id(error, self._sel(b'localizedDescription'))) if error.value else ''
+                raise VisionError(description or 'Vision performRequests:error: failed without an NSError')
+
+            results = self._id(request, self._sel(b'results'))
+            if not results:
+                return ''
+
+            lines = []
+            for i in range(self._uint(results, self._sel(b'count'))):
+                observation = self._id_uint(results, self._sel(b'objectAtIndex:'), i)
+                candidates = self._id_uint(observation, self._sel(b'topCandidates:'), 1)
+                if not candidates or not self._uint(candidates, self._sel(b'count')):
+                    continue
+                candidate = self._id_uint(candidates, self._sel(b'objectAtIndex:'), 0)
+                lines.append(self._string(self._id(candidate, self._sel(b'string'))))
+
+            # Strings are copied into Python before draining the pool. Preserve Vision's observation order.
+            return '\n'.join(lines)
+
+
+@lang.cached_function
+def _get_vision() -> _Vision:
+    return _Vision()
+
+
+def recognize_text(
+        data: bytes,
+        *,
+        languages: ta.Sequence[str] | None = None,
+        fast: bool = False,
+        no_language_correction: bool = False,
+) -> str:
+    if not isinstance(data, bytes):
+        raise TypeError(data)
+    if not data:
+        raise ValueError('Empty image data')
+    if isinstance(languages, str):
+        raise TypeError(languages)
+    if languages is not None:
+        languages = tuple(languages)
+        if not languages or any(not isinstance(s, str) or not s or '\0' in s for s in languages):
+            raise ValueError(languages)
+
+    return _get_vision().recognize_text(
+        data,
+        languages=languages,
+        fast=fast,
+        no_language_correction=no_language_correction,
+    )