Windows - point not allowed in filename

[toninlg]

Hi,

It seems that points are not allowed in file name. File name is truncated at the first point during the process. For 0.7.6, with a file name cut.1.pdf, the error is "The system cannot find the file specified: 'text_cut_ocr.pdf". File name that are created are cut.1_1.png, cut.1_1.html and cut_ocr.pdf. There is ocr text in output but it seems this is the same for all pages.

Thank you.

[rgbyrnes]

This isn't specific to Windows. Here are some fixes for the bugs that we found:

diff --git a/pypdfocr/pypdfocr_filer.py b/pypdfocr/pypdfocr_filer.py
index 6217f90f..705e48fd 100644
--- a/pypdfocr/pypdfocr_filer.py
+++ b/pypdfocr/pypdfocr_filer.py
@@ -66,10 +66,8 @@ class PyFiler(object):

     def _split_filename_dir_filename_ext(self, filename):
         dr, fn = os.path.split(filename) # Get directory and filename
-        fn_no_ext = fn.split('.')[0:-1] # Get the filename without ending extension
-        fn_no_ext = ''.join(fn_no_ext)
-        ext = fn.split('.')[-1]
-        return dr, fn_no_ext, ext
+        fn_no_ext, ext = os.path.splitext(fn) # Get root and extension
+        return dr, fn_no_ext, ext[1:] # Remove leading '.' from extension

     def get_target_folder(self):
         return self._target_folder

diff --git a/pypdfocr/pypdfocr_pdf.py b/pypdfocr/pypdfocr_pdf.py
index 30f88d2f..4d03f061 100644
--- a/pypdfocr/pypdfocr_pdf.py
+++ b/pypdfocr/pypdfocr_pdf.py
@@ -112,7 +112,7 @@ class PyPdf(object):
             writer.addPage(orig_pg)

         pdf_dir, pdf_basename = os.path.split(orig_pdf_filename)
-        basename = pdf_basename.split('.')[0]
+        basename = os.path.splitext(pdf_basename)[0]
         pdf_filename = os.path.join(pdf_dir, "%s_ocr.pdf" % (basename))
         with open(pdf_filename, 'wb') as f:
             writer.write(f)
@@ -144,7 +144,7 @@ class PyPdf(object):
         logging.debug("hocr_filename:%s, hocr_dir:%s, hocr_basename:%s" % (hocr_filename, hocr_dir, hocr_basename))
         assert(img_dir == hocr_dir)

-        basename = hocr_basename.split('.')[0]
+        basename = os.path.splitext(hocr_basename)[0]
         pdf_filename = os.path.join("text_%s_ocr.pdf" % (basename))

         # Switch to the hocr directory to make this easier
@@ -177,7 +177,7 @@ class PyPdf(object):
     def overlay_hocr_old(self, dpi, hocr_filename):
         hocr_dir, hocr_basename = os.path.split(hocr_filename)
         logging.debug("hocr_filename:%s, hocr_dir:%s, hocr_basename:%s" % (hocr_filename, hocr_dir, hocr_basename))
-        basename = hocr_basename.split('.')[0]
+        basename = os.path.splitext(hocr_basename)[0]
         pdf_filename = os.path.join("%s_ocr.pdf" % (basename))
         text_pdf_filename = pdf_filename + ".tmp"

@@ -227,7 +227,7 @@ class PyPdf(object):
                 pg_num = i+1
                 # Do a quick assert to make sure our sorted page number matches
                 # what's embedded in the filename
-                file_pg_num = int(img_file.split(basename+"_")[1].split('.')[0])
+                file_pg_num = int(os.path.splitext(img_file.split(basename+"_")[1])[0])
                 if file_pg_num != pg_num:
                     logging.warn("Page number from file (%d) does not match iteration (%d)... continuing anyway" % (file_pg_num, pg_num))

[virantha]

Thanks! Let me try to roll these patches into the next release.

[virantha]

Fixed in 0.8.0

[toninlg]

Thanks to both of you, it's working with the 0.8.0. I've just a bunch of warnings but file output is OK.

WARNING: Could not run command convert 'file.name_7.jpg' -respect-parenthesis \( -clone 0 -colorspace gray -neg
ate -lat 15x15+5\% -contrast-stretch 0 \) -compose copy_opacity -composite -opaque none +matte -modulate 100,100 -blur 1x1 -adaptive-sharpen 0x2 -negate -define morp
hology:compose=darken -morphology Thinning Rectangle:1x30+0+0 -negate  'file.name_7_preprocess.jpg'
Invalid Parameter - -respect-parenthesis

[virantha]

Hmm, I'm guessing your ImageMagick version is old, or it's not in your path (that convert utility is from there, and used to clean up the image a bit before OCR). I'll add in a check for version in the next release to clean up that warning.