[
https://issues.apache.org/jira/browse/PDFBOX-3710?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15897067#comment-15897067
]
Roman commented on PDFBOX-3710:
-------------------------------
OK, I found an ugly solution - I've overided whole method *showGlyph()* from
*LegacyPDFStreamEngine* class. (I had to override 4 more
class-member-properties and one function for calculating them, and also a
constructor). So, this solution has performance overhead and very lot of
copy-pasting. At the same time, it is intended to do very little, just to avoid
returning in this piece of code:
{code}
// when there is no Unicode mapping available, Acrobat simply coerces
the character code
// into Unicode, so we do the same. Subclasses of PDFStreamEngine don't
necessarily want
// this, which is why we leave it until this point in
PDFTextStreamEngine.
if (unicode == null)
{
if (font instanceof PDSimpleFont)
{
char c = (char) code;
unicode = new String(new char[] { c });
}
else
{
// Acrobat doesn't seem to coerce composite font's character
codes, instead it
// skips them. See the "allah2.pdf" TestTextStripper file.
return;
}
}
{code}
now changed to:
{code}
// when there is no Unicode mapping available, Acrobat simply coerces
the character code
// into Unicode, so we do the same. Subclasses of PDFStreamEngine don't
necessarily want
// this, which is why we leave it until this point in
PDFTextStreamEngine.
if (unicode == null)
{
// if (font instanceof PDSimpleFont)
// {
char c = (char) code;
unicode = new String(new char[] { c });
// }
// else
// {
// // Acrobat doesn't seem to coerce composite font's character
codes, instead it
// // skips them. See the "allah2.pdf" TestTextStripper file.
//
// return;
// }
}
{code}
My only left question: can you tweak LegacyPDFStreamEngine class to be more
flexible. For example, we may add new boolean method *deepLegacy* as here:
{code}
// when there is no Unicode mapping available, Acrobat simply coerces
the character code
// into Unicode, so we do the same. Subclasses of PDFStreamEngine don't
necessarily want
// this, which is why we leave it until this point in
PDFTextStreamEngine.
if (unicode == null)
{
if (deepLegacy() || font instanceof PDSimpleFont)
{
char c = (char) code;
unicode = new String(new char[] { c });
}
else
{
// Acrobat doesn't seem to coerce composite font's character
codes, instead it
// skips them. See the "allah2.pdf" TestTextStripper file.
return;
}
}
{code}
> Text Stripper in 2.0 lost some texts - regression
> -------------------------------------------------
>
> Key: PDFBOX-3710
> URL: https://issues.apache.org/jira/browse/PDFBOX-3710
> Project: PDFBox
> Issue Type: Bug
> Components: Text extraction
> Reporter: Roman
> Attachments: highlight19.pdf_page1-marked-1.png,
> highlight19.pdf_page1.pdf, regression_in_blue.png
>
>
> After migration of our App from pdfbox 1.8 to 2.0, we noticed a regression: 4
> lines of texts are disappeared. Those are the texts followed by black bullet
> (3 lines) and also "OVERALL" word which is placed above in table.
> Problematic PDF attached - [^highlight19.pdf_page1.pdf]
> Also, attached the result of
> [DrawPrintTextLocations|https://apache.googlesource.com/pdfbox/+/trunk/examples/src/main/java/org/apache/pdfbox/examples/util/DrawPrintTextLocations.java]
> example -
> [highlight19.pdf_page1-marked-1.png|https://issues.apache.org/jira/secure/attachment/12856229/highlight19.pdf_page1-marked-1.png]
> Notice, that unicodes, red and blue boxes missing for problematic text. The
> main problem that these glyphs are absent in *textPositions* parameter which
> is passed to *writeString* function, line #275. In the 1.8 version these
> characters ARE present, so their positions along with their char codes could
> be extracted fine in our App.
> Also, attached picture of regression in our App - [^regression_in_blue.png].
> Here, blue boxes drawn where text WAS present and disappeared afterwards.
> (The purple boxes are OK and should be ignored.)
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
