PDFminer empty output

Question

While processing a file with pdfminer (pdf2txt.py) I received empty output:

dan@work:~/project$ pdf2txt.py  docs/homericaeast.pdf 

dan@work:~/project$ 

Can anybody say what wrong with this file and what I can do to get data from it?

Here's dumppdf.py docs/homericaeast.pdf output:

<trailer>
<dict size="4">
<key>Info</key>
<value><ref id="2" /></value>
<key>Root</key>
<value><ref id="1" /></value>
<key>ID</key>
<value><list size="2">
<string size="16">on&#10;&#164;&#181;F&#164;5&#193;&#62;&#243;_&#253;v&#172;`</string>
<string size="16">on&#10;&#164;&#181;F&#164;5&#193;&#62;&#243;_&#253;v&#172;`</string>
</list></value>
<key>Size</key>
<value><number>27</number></value>
</dict>
</trailer>

<trailer>
<dict size="4">
<key>Info</key>
<value><ref id="2" /></value>
<key>Root</key>
<value><ref id="1" /></value>
<key>ID</key>
<value><list size="2">
<string size="16">on&#10;&#164;&#181;F&#164;5&#193;&#62;&#243;_&#253;v&#172;`</string>
<string size="16">on&#10;&#164;&#181;F&#164;5&#193;&#62;&#243;_&#253;v&#172;`</string>
</list></value>
<key>Size</key>
<value><number>27</number></value>
</dict>
</trailer>

Answer 1

Now I have fixed the problem with /OneByteIdentityH similarly to the code for two byte unicode mapping /Identity-H. The patch is in PR #179

Answer 2

The problem is that pdfminer doesn't understand the CMap that you are using in this PDF.

If you do a custom build of pdfminer switching STRICT=1 on in psparser.py you'll get an error a bit like this:

pdfminer.psparser.PSTypeError: Literal required: <PDFStream(21): raw=267, {u'Filter': /'FlateDecode', u'CMapName': /u'OneByteIdentityH', u'Type': /u'CMap', u'CIDSystemInfo': <PDFObjRef:20>, u'Length': 266}>

I'm not hugely familiar with the code, but even allowing this through doesn't help, because it doesn't recognize the mapping (even if I hard code the name to OneByteIdentityH and ask it to look that up). The net result is that the CMap contains no mappings and so it translates every character in your PDF to an empty string (well None if I'm being picky).

The fix is probably to create a mapping for this CMap that simply returns the character that was passed in similar to the other Identity maps already implemented in cmapdb.py