[
https://issues.apache.org/jira/browse/PDFBOX-2894?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14634046#comment-14634046
]
John Hewson edited comment on PDFBOX-2894 at 7/20/15 8:42 PM:
--------------------------------------------------------------
I'm trying to solve some of the memory and performance issues with 2.0 by
getting rid of SequenceRandomAccessRead, which necessitates removing
COSStreamArray too. Otherwise I'd agree that we could postpone this.
It's a pretty simple change really, because we're replacing a very complex
artificial abstraction with a simple one which represents the fact that pages
can contain arrays of streams.
Also... we can't solve this problem by deprecating PDPage#getStream(), because
that won't allow us to remove any of the classes I've mentioned. We'd just be
adding yet another layer on top.
was (Author: jahewson):
I'm trying to solve some of the memory and performance issues with 2.0 by
getting rid of SequenceRandomAccessRead, which necessitates removing
COSStreamArray too. Otherwise I'd agree that we could postpone this.
It's a pretty simple change really, because we're replacing a very complex
artificial abstraction with a simple one which represents the fact that pages
can contain arrays of streams.
> Remove COSStreamArray / SequenceRandomAccessRead
> ------------------------------------------------
>
> Key: PDFBOX-2894
> URL: https://issues.apache.org/jira/browse/PDFBOX-2894
> Project: PDFBox
> Issue Type: Improvement
> Affects Versions: 2.0.0
> Reporter: John Hewson
> Assignee: John Hewson
> Fix For: 2.0.0
>
>
> This ties in with my COSStream simplification in PDFBOX-2893.
> COSStreamArray is a troublesome abstraction, it's not a real COS object and
> it's the only COS object which can be generated _after_ parsing. Look at the
> implementation of COSStreamArray, most methods throw an exception because
> it's _not_ a COSStream - it violates the contact of the very thing it claims
> to be. Even PDPageContentStream has to use instanceof to "peer through" the
> abstraction of COSStreamArray.
> There's no reason to have this class, other than to duck-tape flaws in 1.8's
> APIs, namely that PDPage#getStream() returns a PDStream and PDFStreamParser
> expects a PDStream, yet both of these may be arrays of streams.
> We can fix this in 2.0 by getting rid of the erroneous PDPage#getStream() and
> by exposing the array of streams, rather than attempting to hide them. -This
> will also fix existing errors throughout the codebase which are associated
> with mistaking COSStreamArray for a COSStream.- We can still provide an
> InputStream API which abstracts over the array of streams, because there's
> nothing wrong with that - so users can have the same simple and convenient
> experience.
> An added benefit of doing this is that it will allow us to remove
> SequenceRandomAccessRead, a highly complex memory-holding class.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
