doocus first ver

doocus/extractors/image.py

@@ -0,0 +1,79 @@
+"""
+Image-family extractor: jpg/jpeg, png.
+
+Reads EXIF (including capture date) into metadata, writes a downscaled JPEG
+thumbnail, and — when options['ocr'] is set and pytesseract is available — OCRs
+the image into content. Without OCR, content is a short placeholder note.
+"""
+import io
+from pathlib import Path
+
+from .base import ExtractionResult
+
+EXTRACTOR_ID = "image/pillow@1"
+THUMB_MAX = 640  # longest edge, px
+
+
+def _exif(img) -> dict:
+    try:
+        from PIL.ExifTags import TAGS
+    except ImportError:
+        return {}
+    raw = getattr(img, "_getexif", lambda: None)()
+    if not raw:
+        return {}
+    out = {}
+    for tag_id, value in raw.items():
+        name = TAGS.get(tag_id, str(tag_id))
+        if name in ("DateTime", "DateTimeOriginal", "Make", "Model", "Orientation"):
+            out[name] = str(value)
+    return out
+
+
+def _thumbnail(img) -> bytes:
+    from PIL import Image  # noqa: F401 (ensure Pillow present)
+    im = img.copy()
+    im.thumbnail((THUMB_MAX, THUMB_MAX))
+    if im.mode not in ("RGB", "L"):
+        im = im.convert("RGB")
+    buf = io.BytesIO()
+    im.save(buf, format="JPEG", quality=80)
+    return buf.getvalue()
+
+
+def extract(source: Path, options: dict) -> ExtractionResult:
+    from PIL import Image  # lazy
+
+    warnings = []
+    metadata = {}
+    content = ""
+    thumb = None
+
+    with Image.open(source) as img:
+        metadata["width"], metadata["height"] = img.size
+        metadata["mode"] = img.mode
+        exif = _exif(img)
+        if exif:
+            metadata["exif"] = exif
+            if "DateTimeOriginal" in exif or "DateTime" in exif:
+                metadata["captured"] = exif.get("DateTimeOriginal", exif.get("DateTime"))
+        try:
+            thumb = _thumbnail(img)
+        except Exception as e:
+            warnings.append(f"thumbnail failed: {e}")
+
+        if options.get("ocr"):
+            try:
+                import pytesseract
+                content = pytesseract.image_to_string(img).strip()
+                metadata["ocr"] = True
+            except ImportError:
+                warnings.append("--ocr requested but pytesseract not installed")
+            except Exception as e:
+                warnings.append(f"ocr failed: {e}")
+
+    if not content:
+        content = f"![{source.name}](thumb.jpg)\n\n_Image {metadata.get('width')}×{metadata.get('height')}. Run with --ocr to extract any text._\n"
+
+    return ExtractionResult(content=content, metadata=metadata, thumb=thumb,
+                            warnings=warnings, extractor=EXTRACTOR_ID)