[
https://issues.apache.org/jira/browse/PDFBOX-2592?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14270028#comment-14270028
]
Andreas Lehmkühler commented on PDFBOX-2592:
--------------------------------------------
I guess I may have some background about that "magic voodoo holding references"
part.
John is totally right about gc, but he didn't put the user into account. There
were several cases where users keep references to all PDDocument instances they
are working on. As long as those instances are referenced, the corresponding
COSDocument is referenced as well as all the other referenced objects.
Therefore those objects can't be removed by the gc. To free those objects
*earlier* by the gc we set all involved references to null and clear all lists.
This makes the whole thing more independent of the user behaviour as long as he
closed the pdf.
> Allow sharing of COS objects between different documents
> --------------------------------------------------------
>
> Key: PDFBOX-2592
> URL: https://issues.apache.org/jira/browse/PDFBOX-2592
> Project: PDFBox
> Issue Type: Improvement
> Components: PDModel
> Affects Versions: 2.0.0
> Reporter: John Hewson
> Assignee: John Hewson
>
> A number of users on the mailing list have asked about how to import pages
> from other PDFs as forms, our current solution is LayerUtility, which is
> depends on PDFCloneUtility. Both these classes are surprisingly complex for
> what should be a simple task.
> The two main tasks which these classes perform is copying the page's
> COSStream and cloning every relevant COS object. However, there seems to be
> no real need to do any of this copying and cloning - there's nothing about
> any of the COS objects which is specific to a given document. While a
> COSStream can share the same backing file as the COSDocument, this isn't a
> problem for COSWriter, even then we need only make sure that an exception is
> thrown if a COSStream is used after its parent COSDocument is closed.
> Note that there *is* one artificial dependency between COSDictionary and
> COSArrays and their parent COSDocument, that is that calling close() on the
> COSDocument clears the contents of all child COSDictionary and COSArrays.
> However, there's no need for this, it seems to have come about due to some
> long past confusion regarding how garbage collection works in Java - we all
> know that it's not necessary to set objects to null or clear lists when we
> are done with them.
> I propose that we get rid of the unnecessary object and list clearing in
> COSDocument#close() and add some checks to COSStream to throw user-friendly
> exceptions when reading from a closed backing stream. This will allow us to
> directly share COS objects between different COSDocuments, allowing simple "x
> = y" copying and making LayerUtility and PDFCloneUtility unnecessary. Instead
> of:
> {code}
> COSStream pageStream = (COSStream)page.getStream().getCOSObject();
> PDStream newStream = new PDStream(targetDoc,
> pageStream.getUnfilteredStream(), false);
> PDFormXObject form = new PDFormXObject(newStream);
> PDResources pageRes = page.getResources();
> PDResources formRes = new PDResources();
> PDFCloneUtility cloner = new PDFCloneUtility(document);
> cloner.cloneMerge(pageRes, formRes);
> form.setResources(formRes);
> {code}
> We could have:
> {code}
> PDFormXObject form = new PDFormXObject(page.getStream());
> form.setResources(page.getResources());
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
