Hi,
Your issue has been fixed. Try with a snapshot or wait for 2.0.9.
https://issues.apache.org/jira/browse/PDFBOX-4060
Tilman
Am 09.01.2018 um 19:59 schrieb Tilman Hausherr:
Am 09.01.2018 um 09:05 schrieb Tilman Hausherr:
Am 09.01.2018 um 07:10 schrieb Soon Keong Tan:
My team is having some problems with the image rendering speed of
certain
PDF file. For most of the pdf files we are handling, it only took
seconds
to create an image of the file but for certain pdf, it took more than 6
minutes.
We have tried the following version of pdfbox-app-x.x.x.jar, and it
seems
that 1.8.x is more efficient at rendering the image.
(1)1.8.13 - 1.5 mins
(2)2.0.5 - 6.18 mins
(3)2.0.8 - 6.35 mins
However, due to the problem that we had with some files where some
Japanese
characters were not correctly rendered using 1.8.13, we had to use
2.0.5 as
the production version.
I tried inserting some debug code in the PDFToImage class (ver2.0.5) to
determine the bottle-necked process, and it seems
"renderer.renderImageWithDPI" was taking up most of the time.
==========================
Java version: 1.7.0_72
PDFBox version: 2.0.5
Command line: java -jar ./pdfbox-app-2.0.5.jar PDFToImage -time
-startPage
1 -endPage 1 ./sample_slow.pdf
File: https://goo.gl/WEMM2X
==========================
The full version of the PDF is quite large, so the linked file above
is the
cropped version (the page which we are having problem rendering). The
cropped version is created using PDFSplit command line.
This is my first time using the mailing list, should I just create a
JIRA
ticket requesting help instead of addressing the mailing list regarding
this problem?
It's fine to post to the mailing list first.
I had a quick look on your file... it has 1999 probably identical
separation colorspaces that are just a black or white value. These
map to a CMYK colorspace.
I'll look more later this week.
Did you set / try the two settings mentioned here?
https://pdfbox.apache.org/2.0/getting-started.html
I ran the profiler and the cause of the slowness is different... there
is a large jpeg file (5349 x 3806) that uses a DeviceN colorspace
which in turn is based on CMYK. There's some slowness due to the type0
convert function from DeviceN to CMYK, but this is less than 10%. Most
time is from converting the CMYK to RGB, one pixel at a time. (Because
of the DeviceN colorspace we can't use bulk conversion, which may or
may not be faster)
I've opened a JIRA issue
(https://issues.apache.org/jira/browse/PDFBOX-4060) but don't expect
this to be fixed soon. CMYK and ICC colorspaces are our weak spot :-(
Tilman
Tilman
Any help is deeply appreciated. Thank you in anticipation.
Regards,
Soon Keong Tan
----------------------
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
