Added support for the command line program pdftotext from the poppler-utils packages to extract text from PDF documents without doing OCR

apps/main/views.py

@@ -94,6 +94,7 @@ def check_settings(request):
         {'name': 'OCR_TESSERACT_LANGUAGE', 'value': ocr_settings.TESSERACT_LANGUAGE},
         {'name': 'OCR_NODE_CONCURRENT_EXECUTION', 'value': ocr_settings.NODE_CONCURRENT_EXECUTION},
         {'name': 'OCR_REPLICATION_DELAY', 'value': ocr_settings.REPLICATION_DELAY},
+        {'name': 'OCR_PDFTOTEXT_PATH', 'value': ocr_settings.PDFTOTEXT_PATH, 'exists': True},
 
         # Search
         {'name': 'SEARCH_LIMIT', 'value': search_settings.LIMIT},

apps/ocr/api.py

@@ -14,6 +14,8 @@ from documents.models import DocumentPage
 
 from ocr.conf.settings import TESSERACT_PATH
 from ocr.conf.settings import TESSERACT_LANGUAGE
+from ocr.conf.settings import PDFTOTEXT_PATH
+from exceptions import TesseractError, PdftotextError
 
 
 def get_language_backend():
@@ -27,10 +29,6 @@ def get_language_backend():
 backend = get_language_backend()
 
 
-class TesseractError(Exception):
-    pass
-
-
 def cleanup(filename):
     ''' tries to remove the given filename. Ignores non-existent files '''
     try:
@@ -51,26 +49,53 @@ def run_tesseract(input_filename, output_filename_base, lang=None):
         raise TesseractError(error_text)
 
 
+def run_pdftotext(input_filename, output_filename, page_number=None):
+    command = [unicode(PDFTOTEXT_PATH)]
+    if page_number:
+        command.extend(['-nopgbrk', '-f', unicode(page_number), '-l', unicode(page_number)])
+    command.extend([unicode(input_filename), unicode(output_filename)])
+    proc = subprocess.Popen(command, close_fds=True, stderr=subprocess.PIPE, stdout=subprocess.PIPE)
+    return_code = proc.wait()
+    if return_code != 0:
+        error_text = proc.stderr.read()
+        raise PdftotextError(error_text)
+  
+
 def do_document_ocr(document):
     for page_index, document_page in enumerate(document.documentpage_set.all()):
-        imagefile = convert_document_for_ocr(document, page=page_index)
         desc, filepath = tempfile.mkstemp()
+        imagefile = None
+        source = u''
         try:
+            if document.file_mimetype == u'application/pdf':
+                pdf_filename = os.extsep.join([filepath, u'pdf'])
+                document.save_to_file(pdf_filename)
+                run_pdftotext(pdf_filename, filepath, document_page.page_number)
+                cleanup(pdf_filename)
+                if os.stat(filepath).st_size == 0:
+                    #PDF page had no text, run tesseract on the page
+                    imagefile = convert_document_for_ocr(document, page=page_index)
                     run_tesseract(imagefile, filepath, TESSERACT_LANGUAGE)
                     ocr_output = os.extsep.join([filepath, 'txt'])
+                    source = _(u'Text from OCR')
+                else:
+                    ocr_output = filepath
+                    source = _(u'Text extracted from PDF')
+            else:
+                imagefile = convert_document_for_ocr(document, page=page_index)
+                run_tesseract(imagefile, filepath, TESSERACT_LANGUAGE)
+                ocr_output = os.extsep.join([filepath, 'txt'])
+                source = _(u'Text from OCR')
             f = codecs.open(ocr_output, 'r', 'utf-8')
             document_page = document.documentpage_set.get(page_number=page_index + 1)
             document_page.content = ocr_cleanup(f.read().strip())
-            document_page.page_label = _(u'Text from OCR')
+            document_page.page_label = source
             document_page.save()
             f.close()
             cleanup(ocr_output)
-        except TesseractError, e:
-            cleanup(filepath)
-            cleanup(imagefile)
-            raise TesseractError(e)
         finally:
             cleanup(filepath)
+            if imagefile:
                 cleanup(imagefile)
 
 

apps/ocr/conf/settings.py

@@ -6,3 +6,4 @@ TESSERACT_LANGUAGE = getattr(settings, 'OCR_TESSERACT_LANGUAGE', u'eng')
 REPLICATION_DELAY = getattr(settings, 'OCR_REPLICATION_DELAY', 10)  # In seconds
 NODE_CONCURRENT_EXECUTION = getattr(settings, 'OCR_NODE_CONCURRENT_EXECUTION', 1)
 AUTOMATIC_OCR = getattr(settings, 'OCR_AUTOMATIC_OCR', False)
+PDFTOTEXT_PATH = getattr(settings, 'OCR_PDFTOTEXT_PATH', u'/usr/bin/pdftotext')

apps/ocr/exceptions.py

@@ -1,2 +1,10 @@
 class AlreadyQueued(Exception):
     pass
+
+
+class TesseractError(Exception):
+    pass
+    
+    
+class PdftotextError(Exception):
+    pass

settings.py

@@ -224,6 +224,7 @@ TEMPLATE_CONTEXT_PROCESSORS = (
 #OCR_TESSERACT_LANGUAGE = u'eng'
 #OCR_REPLICATION_DELAY = 10
 #OCR_AUTOMATIC_OCR = False
+#OCR_PDFTOTEXT_PATH = u'/usr/bin/pdftotext'
 
 # Permissions
 #ROLES_DEFAULT_ROLES = []