Add corrupt text warning (when using --redo-ocr)

src/ocrmypdf/_pipeline.py

@@ -252,6 +252,13 @@ def is_ocr_required(pageinfo, log, options):
                                 "rasterizing text and running OCR anyway"))
             ocr_required = True
         elif options.redo_ocr:
+            if pageinfo.has_corrupt_text:
+                log.warning(msg.format(
+                    page,
+                    "some text on this page cannot be mapped to characters: "
+                    "consider using --force-ocr instead")
+                )
+            else:
             log.info(msg.format(page,
                                 "redoing OCR"))
             ocr_required = True

src/ocrmypdf/pdfinfo/__init__.py

@@ -108,7 +108,7 @@ InlineSettings = namedtuple('InlineSettings',
 ContentsInfo = namedtuple('ContentsInfo',
     ['xobject_settings', 'inline_images', 'found_text', 'found_vector'])
 
-TextBoxInfo = namedtuple('TextBoxInfo',
+TextboxInfo = namedtuple('TextboxInfo',
     ['bbox', 'is_visible', 'is_corrupt'])
 
 
@@ -603,7 +603,7 @@ def simplify_textboxes(miner):
 
         visible = (first_char.rendermode != 3)
         corrupt = (first_char.get_text() == '\ufffd')
-        yield TextBoxInfo(box.bbox, visible, corrupt)
+        yield TextboxInfo(box.bbox, visible, corrupt)
 
 
 def _pdf_get_pageinfo(pdf, pageno: int, infile, xmltext):
@@ -714,6 +714,10 @@ class PageInfo:
     def has_text(self):
         return self._pageinfo['has_text']
 
+    @property
+    def has_corrupt_text(self):
+        return any(tbox.is_corrupt for tbox in self._pageinfo['textboxes'])
+
     @property
     def has_vector(self):
         return self._pageinfo['has_vector']

src/ocrmypdf/pdfinfo/layout.py

@@ -35,12 +35,27 @@ from pdfminer.utils import matrix2str, bbox2str, fsplit
 from ..exceptions import EncryptedPdfError
 
 
-# Fix pdfminer's regex in name2unicode function
-# Font cids that are mapped to names of the form /g123 seem to be, by convention
-# characters with no corresponding Unicode entry. These can be subsetted fonts
-# or symbolic fonts. There seems to be no way to map /g123 fonts to Unicode,
-# barring a ToUnicode data structure.
-pdfminer.encodingdb.STRIP_NAME = re.compile(r'(?![g])([0-9]+)')
+
+STRIP_NAME = re.compile(r'[0-9]+')
+
+def name2unicode(name):
+    """Fix pdfminer's regex in name2unicode function
+
+    Font cids that are mapped to names of the form /g123 seem to be, by convention
+    characters with no corresponding Unicode entry. These can be subsetted fonts
+    or symbolic fonts. There seems to be no way to map /g123 fonts to Unicode,
+    barring a ToUnicode data structure.
+    """
+    if name in glyphname2unicode:
+        return glyphname2unicode[name]
+    if name.startswith('g'):
+        raise KeyError(name)
+    m = STRIP_NAME.search(name)
+    if not m:
+        raise KeyError(name)
+    return chr(int(m.group(0)))
+
+pdfminer.encodingdb.name2unicode = name2unicode
 
 from math import copysign
 def PDFType3Font__get_height(self):