[PDFBOX-347] Spaces removed after text extraction

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.8.0-incubator
• Component/s: Text extraction
• Assignee: Unassigned
• Reporter: Jukka Zitting
• Created: 04/Aug/08 17:48
• Updated: 21/Oct/09 10:28
• Resolved: 01/Apr/09 14:36

Description

[Issue from SourceForge]
http://sourceforge.net/tracker/index.php?func=detail&aid=1912364&group_id=78314&atid=552832

The spaces between words from the attached PDF file are removed upon text
extraction.

I traced the code and found that the cause seems to be a "division by 0"
bug in PDCIDFont.java

In PDCIDFont.getAverageFontWidth(), widths is returned as null from

COSArray widths = (COSArray)font.getDictionaryObject( COSName.getPDFName(
"W" ) );

,causing characterCount to be 0.

The result is that the following line
float average = totalWidths / characterCount;

returns a NaN, which gets propagated up the method calls to result in the
spaces being removed.

I suggest the following fix, instead of:
float average = totalWidths / characterCount;

Have:
float average = defaultWidth;

if (characterCount > 0) {
average = totalWidths / characterCount;
}

[Comment on SourceForge]
Date: 2008-03-12 03:01
Sender: choongyong
Logged In: YES
user_id=2033885
Originator: NO

Realised that I was considered not login when I raised the request.
Sending this comment so that the developer can contact me.

[Comment on SourceForge]
Date: 2008-03-17 21:50
Sender: nobody
Logged In: NO

I have noticed that there is no spaces between 2 words, if they are
separated by a new line (or the 2nd word is on the next line because it
reaches the right margin).

Could you correct please ?

Issue Links

• duplicates: PDFBOX-349 Spaces between words ignored in scanned pdf files