Updated README.md [skip ci]

CHANGELOG.md

@@ -4,6 +4,20 @@ All notable changes to this project will be documented in this file. Dates are d
 
 Generated by [`auto-changelog`](https://github.com/CookPete/auto-changelog).
 
+#### [v0.42.72](https://github.com/RhetTbull/osxphotos/compare/v0.42.71...v0.42.72)
+
+> 2 August 2021
+
+- Improved caching of detected_text results [`fa2027d`](https://github.com/RhetTbull/osxphotos/commit/fa2027d45308738d2335d4b5a72c3ef5c478491a)
+
+#### [v0.42.71](https://github.com/RhetTbull/osxphotos/compare/v0.42.70...v0.42.71)
+
+> 29 July 2021
+
+- Updated text_detection to detect macOS version [`7376223`](https://github.com/RhetTbull/osxphotos/commit/7376223eb87a4919fd54cc685a3f263e83626879)
+- Updated detected_text docs to make it clear this only works on Catalina+ [`ecd0b8e`](https://github.com/RhetTbull/osxphotos/commit/ecd0b8e22f8bf1f8d1e98d64834bebf0394dd903)
+- Fix for #500, check for macOS version before loading Vision [`673243c`](https://github.com/RhetTbull/osxphotos/commit/673243c6cd1c267b6b741b5429cdb63c062648d1)
+
 #### [v0.42.70](https://github.com/RhetTbull/osxphotos/compare/v0.42.69...v0.42.70)
 
 > 29 July 2021

README.md

@@ -35,6 +35,7 @@ OSXPhotos provides the ability to interact with and query Apple's Photos.app lib
   + [Raw Photos](#raw-photos)
   + [Template System](#template-system)
   + [ExifTool](#exiftoolExifTool)
+  + [Text Detection](#textdetection)
   + [Utility Functions](#utility-functions)
 * [Examples](#examples)
 * [Related Projects](#related-projects)
@@ -1700,7 +1701,7 @@ Substitution                    Description
 {lf}                            A line feed: '\n', alias for {newline}
 {cr}                            A carriage return: '\r'
 {crlf}                          a carriage return + line feed: '\r\n'
-{osxphotos_version}             The osxphotos version, e.g. '0.42.71'
+{osxphotos_version}             The osxphotos version, e.g. '0.42.73'
 {osxphotos_cmd_line}            The full command line used to run osxphotos
 
 The following substitutions may result in multiple values. Thus if specified for
@@ -2788,7 +2789,8 @@ Some substitutions, notably `album`, `keyword`, and `person` could return multip
 
 See [Template System](#template-system) for additional details.
 
-#### `detected_text(confidence_threshold=TEXT_DETECTION_CONFIDENCE_THRESHOLD)`
+
+#### <a name="detected_text_method">`detected_text(confidence_threshold=TEXT_DETECTION_CONFIDENCE_THRESHOLD)`</a>
 
 Detects text in photo and returns lists of results as (detected text, confidence)
 
@@ -2800,6 +2802,8 @@ Returns: list of (detected text, confidence) tuples.
 
 Note: This is *not* the same as Live Text in macOS Monterey.  When using `detected_text()`, osxphotos will use Apple's [Vision framework](https://developer.apple.com/documentation/vision/recognizing_text_in_images?language=objc) to perform text detection on the image.  On my circa 2013 MacBook Pro, this takes about 2 seconds per image.  `detected_text()` does memoize the results for a given `confidence_threshold` so repeated calls will not re-process the photo.  This works only on macOS Catalina (10.15) or later.
 
+See also [Text Detection](#textdetection).
+
 ### ExifInfo
 [PhotosInfo.exif_info](#exif-info) returns an `ExifInfo` object with some EXIF data about the photo (Photos 5 only).  `ExifInfo` contains the following properties:
 
@@ -3554,7 +3558,7 @@ The following template field substitutions are availabe for use the templating s
 |{lf}|A line feed: '\n', alias for {newline}|
 |{cr}|A carriage return: '\r'|
 |{crlf}|a carriage return + line feed: '\r\n'|
-|{osxphotos_version}|The osxphotos version, e.g. '0.42.71'|
+|{osxphotos_version}|The osxphotos version, e.g. '0.42.73'|
 |{osxphotos_cmd_line}|The full command line used to run osxphotos|
 |{album}|Album(s) photo is contained in|
 |{folder_album}|Folder path + album photo is contained in. e.g. 'Folder/Subfolder/Album' or just 'Album' if no enclosing folder|
@@ -3634,6 +3638,14 @@ osxphotos.exiftool also provides an `ExifToolCaching` class which caches all met
 
 `ExifTool()` runs `exiftool` as a subprocess using the `-stay_open True` flag to keep the process running in the background.  The subprocess will be cleaned up when your main script terminates.  `ExifTool()` uses a singleton pattern to ensure that only one instance of `exiftool` is created.  Multiple instances of `ExifTool()` will all use the same `exiftool` subprocess.
 
+### <a name="textdetection">Text Detection</a>
+
+The [PhotoInfo.detected_text()](#detected_text_method) and the `{detected_text}` template will perform text detection on the photos in your library. Text detection is a slow process so to avoid unnecessary re-processing of photos, osxphotos will cache the results of the text detection process as an extended attribute on the photo image file.  Extended attributes do not modify the actual file.  The extended attribute is named `osxphotos.metadata:detected_text` and can be viewed using the built-in [xattr](https://ss64.com/osx/xattr.html) command or my [osxmetadata](https://github.com/RhetTbull/osxmetadata) tool.  If you want to remove the cached attribute, you can do so with osxmetadata as follows:
+
+`osxmetadata --clear osxphotos.metadata:detected_text --walk ~/Pictures/Photos\ Library.photoslibrary/`
+
+
+
 ### Utility Functions
 
 The following functions are located in osxphotos.utils

osxphotos/_version.py

@@ -1,3 +1,3 @@
 """ version info """
 
-__version__ = "0.42.71"
+__version__ = "0.42.73"

osxphotos/cli.py

@@ -4074,6 +4074,10 @@ def _get_selected(photosdb):
 @click.pass_context
 def repl(ctx, cli_obj, db):
     """Run interactive osxphotos shell"""
+
+    from osxphotos import PhotosDB, PhotoInfo, ExifTool
+    from rich import inspect as _inspect
+
     pretty.install()
     print(f"python version: {sys.version}")
     print(f"osxphotos version: {osxphotos._version.__version__}")
@@ -4090,6 +4094,10 @@ def repl(ctx, cli_obj, db):
     show = _show_photo
     get_selected = _get_selected(photosdb)
 
+    def inspect(obj):
+        """inspect object"""
+        return _inspect(obj, methods=True)
+
     print(f"Found {len(photos)} photos in {tictoc:0.2f} seconds")
     print("The following variables are defined:")
     print(f"- photosdb: PhotosDB() instance for {photosdb.library_path}")
@@ -4105,5 +4113,8 @@ def repl(ctx, cli_obj, db):
     print(
         f"- help(object): print help text including list of methods for object; for example, help(PhotosDB)"
     )
+    print(
+        f"- inspect(object): print information about an object; for example inspect(photosdb)"
+    )
     print(f"- quit(): exit this interactive shell\n")
     code.interact(banner="", local=locals())

osxphotos/photoinfo/photoinfo.py

@@ -14,6 +14,7 @@ from datetime import timedelta, timezone
 from typing import Optional
 
 import yaml
+from osxmetadata import OSXMetaData
 
 from .._constants import (
     _MOVIE_TYPE,
@@ -1118,6 +1119,28 @@ class PhotoInfo:
 
         Returns: list of (detected text, confidence) tuples
         """
+
+        try:
+            return self._detected_text_cache[confidence_threshold]
+        except (AttributeError, KeyError) as e:
+            if isinstance(e, AttributeError):
+                self._detected_text_cache = {}
+
+            try:
+                detected_text = self._detected_text()
+            except Exception as e:
+                logging.warning(f"Error detecting text in photo {self.uuid}: {e}")
+                detected_text = []
+
+            self._detected_text_cache[confidence_threshold] = [
+                (text, confidence)
+                for text, confidence in detected_text
+                if confidence >= confidence_threshold
+            ]
+            return self._detected_text_cache[confidence_threshold]
+
+    def _detected_text(self):
+        """detect text in photo, either from cached extended attribute or by attempting text detection"""
         path = (
             self.path_edited if self.hasadjustments and self.path_edited else self.path
         )
@@ -1125,24 +1148,12 @@ class PhotoInfo:
         if not path:
             return []
 
-        try:
+        md = OSXMetaData(path)
-            return self._detected_text[(path, confidence_threshold)]
+        detected_text = md.get_attribute("osxphotos_detected_text")
-        except (AttributeError, KeyError) as e:
+        if detected_text is None:
-            if isinstance(e, AttributeError):
+            detected_text = detect_text(path)
-                self._detected_text = {}
+            md.set_attribute("osxphotos_detected_text", detected_text)
-
+        return detected_text
-            try:
-                detected_text = detect_text(path)
-            except Exception as e:
-                logging.warning(f"Error detecting text in photo {self.uuid} at {path}: {e}")
-                detected_text = []
-
-            self._detected_text[(path, confidence_threshold)] = [
-                (text, confidence)
-                for text, confidence in detected_text
-                if confidence >= confidence_threshold
-            ]
-            return self._detected_text[(path, confidence_threshold)]
 
     @property
     def _longitude(self):

osxphotos/phototemplate.py

@@ -1445,25 +1445,8 @@ def _get_detected_text(photo, exportdb, confidence=TEXT_DETECTION_CONFIDENCE_THR
         else TEXT_DETECTION_CONFIDENCE_THRESHOLD
     )
 
-    detected_text = exportdb.get_detected_text_for_uuid(photo.uuid)
+    # _detected_text caches the text detection results in an extended attribute
-    if detected_text is not None:
+    # so the first time this gets called is slow but repeated accesses are fast
-        detected_text = json.loads(detected_text)
+    detected_text = photo._detected_text()
-    else:
+    exportdb.set_detected_text_for_uuid(photo.uuid, json.dumps(detected_text))
-        path = (
-            photo.path_edited
-            if photo.hasadjustments and photo.path_edited
-            else photo.path
-        )
-        path = path or photo.path_derivatives[0] if photo.path_derivatives else None
-        if not path:
-            detected_text = []
-        else:
-            try:
-                detected_text = detect_text(path)
-            except Exception as e:
-                logging.warning(
-                    f"Error detecting text in image {photo.uuid} at {path}: {e}"
-                )
-                return []
-        exportdb.set_detected_text_for_uuid(photo.uuid, json.dumps(detected_text))
     return [text for text, conf in detected_text if conf >= confidence]

osxphotos/text_detection.py

@@ -52,6 +52,9 @@ def detect_text(img_path: str) -> List:
         vision_request.dealloc()
         vision_handler.dealloc()
 
+        for result in results:
+            result[0] = str(result[0])
+
         return results
 
 

requirements.txt

@@ -16,7 +16,7 @@ dataclasses==0.7;python_version<'3.7'
 wurlitzer==2.1.0
 photoscript==0.1.4
 toml==0.10.2
-osxmetadata==0.99.26
+osxmetadata==0.99.31
 textx==2.3.0
 rich==10.6.0
 bitmath==1.3.3.1

setup.py

@@ -91,7 +91,7 @@ setup(
         "wurlitzer==2.1.0",
         "photoscript==0.1.4",
         "toml==0.10.2",
-        "osxmetadata==0.99.26",
+        "osxmetadata==0.99.31",
         "textx==2.3.0",
         "rich==10.6.0",
         "bitmath==1.3.3.1",

tests/test_template.py

@@ -1179,13 +1179,3 @@ def test_detected_text(photosdb):
     for template, value in TEMPLATE_VALUES_DETECTED_TEXT.items():
         rendered, _ = photo.render_template(template)
         assert value in "".join(rendered)
-
-
-def test_detected_text_caching(photosdb):
-    """Test {detected_text} template caches values"""
-    exportdb = ExportDBInMemory(None)
-    exportdb.set_detected_text_for_uuid(UUID_DETECTED_TEXT, json.dumps([["foo", 0.9]]))
-    photo = photosdb.get_photo(UUID_DETECTED_TEXT)
-    options = RenderOptions(exportdb=exportdb)
-    rendered, _ = photo.render_template("{detected_text}", options=options)
-    assert rendered[0] == "foo"