[ 
https://issues.apache.org/jira/browse/PDFBOX-2377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14150565#comment-14150565
 ] 

Andreas Lehmkühler commented on PDFBOX-2377:
--------------------------------------------

{quote}
I don't get why characters < 32 would be handled differently?
{quote}
All codes >= 32 and <= 127 can be mapped using ASCII, Everything < 32 can't be 
mapped that easily, e.g. by using ISO-8859-1 or UTF-8, and we have to use the 
internal font mapping.
But this doesn't work in any case. 
- 357094.pdf some of the "." are reresented as ":" within the pdf, so that 
those will be mapped wrong using ASCII. The font mapping would provide a 
correct value. Adobe provides the same wrong result
- 312888.pdf works fine with the current implementation. It works too if we use 
the internal font mapping for all characters
- 701542.pdf from PDFBOX-2247 works fine with the current implementation, but 
provides rubbish  if we use the internal font mapping for all characters

It looks as if we have to implement a compromise, as it seems that neither of 
the possible implementations will provide a 100% solution

> Apparent regression in character mapping in a few files from govdocs1
> ---------------------------------------------------------------------
>
>                 Key: PDFBOX-2377
>                 URL: https://issues.apache.org/jira/browse/PDFBOX-2377
>             Project: PDFBox
>          Issue Type: Bug
>          Components: Text extraction
>    Affects Versions: 1.8.7
>            Reporter: Tim Allison
>            Assignee: Andreas Lehmkühler
>            Priority: Minor
>              Labels: regression
>         Attachments: 290991-6.txt, 290991-7.txt, 290991-8.txt, 290991.pdf, 
> 312888.pdf, 357094-1.8.6.txt, 357094-1.8.8.txt, 357094.pdf, 764929.pdf, 
> PDFBOX2247-701542.pdf
>
>
> On a small number of test files in a 50k sample of pdfs from govdocs1, it 
> appears that some characters are no longer being extracted correctly in 1.8.7 
> when compared to 1.8.6.  I ran pdfbox's app.jar with ExtractText
> {noformat}
> 764949.pdf
> 1.8.6: Lang, Astrophysical Data: Planets and Stars
> 1.8.7: Lang, AefdaphyeiUSl DSfS: PlSnefe Snd EfSde,
> {noformat}
> and
> {noformat}
> 312888.pdf
> 1.8.6: Self-Assessment \u0026 Capability Description
> 1.8.7: Seff-Ammemmmehn \u0026 Cajabcfcns Demclcjncih
> {noformat}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to