Andre created PDFBOX-5528:
-----------------------------
Summary: PDF/UA: Add marked content sections when flattening acro
forms
Key: PDFBOX-5528
URL: https://issues.apache.org/jira/browse/PDFBOX-5528
Project: PDFBox
Issue Type: Improvement
Components: AcroForm
Reporter: Andre
Attachments: correct.png, wrong.png
We need to support PDF/UA compliant documents to some extent. I noticed that
when we take a PDF/UA compliant PDF document and flatten it via
PDAcroForm#flatten, the resulting output is not PDF/UA compliant anymore.
After a little bit of research, the problem is that PDFBox creates /DO
operators with paths representing the appearance of the form fields. According
to the PDF/UA standard, such paths need to be enclosed in marked content
sections (BMC ... EMC, BDC ... EMC, see attached images)
By copying some code from AcroForm#flatten and adding
contentStream.beginMarkedContent and contentStream.endMarkedContent myself, I
can workaround the problem, but that's less than ideal, it would be great if
this could be included in PDFBox.
<pre>
final var dict = new COSDictionary();
dict.setLong(COSName.MCID, mcid);
dict.setItem(COSName.BBOX, bBox);
dict.setItem(COSName.TYPE, COSName.BACKGROUND);
final var propList = PDPropertyList.create(dict);
contentStream.beginMarkedContent(COSName.ARTIFACT, propList);
contentStream.saveGraphicsState();
// see https://stackoverflow.com/a/54091766/1729265 for an
explanation
// of the steps required
// this will transform the appearance stream form object into the
rectangle of the
// annotation bbox and map the coordinate systems
final var transformationMatrix =
pdfbox_resolveTransformationMatrix(form, annotation, appearanceStream);
contentStream.transform(transformationMatrix);
contentStream.drawForm(fieldObject);
contentStream.restoreGraphicsState();
contentStream.endMarkedContent();
</pre>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
