pdfbox-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tilman Hausherr <THaush...@t-online.de>
Subject Re: Performance Issue with 2.0.0 SNAPSHOT latest builds
Date Sun, 15 Nov 2015 17:07:52 GMT
Am 15.11.2015 um 17:08 schrieb Sridhar So:
> Dear PDFBox team
>
> Thanks for response to my earlier query on performance issue. The work around suggested
was to use the ctor PDFPageable(pdDocument, Orientation.AUTO, true, 300) while setting PrinterJob
pjob instead of PDFPageable(pdDocument), which is supposed to force the java.awt.print.PrinterJob
pjob.print() to use rasterization to render image which is supoosed to be faster.
>
>   //pjob.setPageable(new PDFPageable(pdDocument));    // In my test this caused performance
issue ( 40 to 80 secs )
> pjob.setPageable( new PDFPageable(pdDocument, Orientation.AUTO, true, 0) );  //In my
test this is about 3 times faster but still 18 to 30 sec.

but you had dpi = 0 above. Except for the page border, this is the same 
as the 1-parameter constructor.


> Below are my observations in unit test
> Compared to 1 ( 40 to 80 sec) above, 2 was faster, but still it took 18 to 28 secs in
multiple rounds of test and 18 to 30 sec is also slow.
> If I used 300 dpi ( last argument to PDFPageable ctor, then the output file ( redirected
to MS XPS file ) quality was not good.

But was it faster at 300 dpi?

> With 600 dpi,  awt.print.Printerjob pjob.print() throws out of memory exception as below
>
> Exception in thread "Thread-3" java.lang.OutOfMemoryError: Java heap space
> 	at java.awt.image.DataBufferInt.<init>(Unknown Source)
> 	at java.awt.image.Raster.createPackedRaster(Unknown Source)
> 	at java.awt.image.DirectColorModel.createCompatibleWritableRaster(Unknown Source)
> 	at java.awt.image.BufferedImage.<init>(Unknown Source)
> 	at org.apache.pdfbox.printing.PDFPrintable.print(PDFPrintable.java:169)
> 	at sun.print.RasterPrinterJob.printPage(Unknown Source)
> 	at sun.print.RasterPrinterJob.print(Unknown Source)
> 	at GPDFBox2AppletClass.run(GPDFBox2AppletClass.java:160)
> 	at java.lang.Thread.run(Unknown Source)
> Nov 15, 2015 8:15:58 PM org.apache.pdfbox.cos.COSDocument finalize
> WARNING: Warning: You did not close a PDF Document

Then you'll need more memory. And don't forget to close your documents.

>
> Input PDF file size is 118KB, whereas the output PDF generated by PDFBox 2.0.0 latest
SNAPSHOT build ( redirecting to MS XPS printer ) is about 1 MB
> There was no alignment and font issue with 2.0 but slower
> OutPut PDF generated by PDFBox was about 500 KB much less in size with PDFBox 1.8.10
and performance was better ( 12 secs in 1.8.10)
> PDFBox 1.8.10 is faster but fonts are lighter ( it used default fonts and has thrown
exception )
>
> Please clarift the following:
> Is 300  or non zero dpi mandatory, in  pjob.setPageable( new PDFPageable(pdDocument,
Orientation.AUTO, true, 300) ) to force awt to use rasterizing.

Yes, non zero is mandatory to force rasterizing (this is mentioned in 
the javadoc).


> I saw in 2.0.0-RC1 code the other ctors of PDFPageable(pdDocument) using default values
for other 3 arguments and what made pjob.setPageable( new PDFPageable(pdDocument, Orientation.AUTO,
true, 0) ) faster, compared to pjob.setPageable(new PDFPageable(pdDocument)) which in turn
internally uses default 0 dpi as per RC1 source code. Am i missing something. Whether ctor
implementations are different in SNAPSHOT code post 2.0.0-RC1?

The best is you just download the source code.. the API wasn't changed 
recently, but something related to rotation was recently changed.

> Why pjob.setPageable( new PDFPageable(pdDocument, Orientation.AUTO, true, 600) ) causes
PrinterJob.print() to throw out of memroy exception?  Is there any way to increase the dpi
without increasing the memory footprint?

No. A 600dpi raster image is 4x as large as a 300dpi raster image. So 
the memory has to come from somewhere.

> Why even with pjob.setPageable( new PDFPageable(pdDocument, Orientation.AUTO, true, 0
) ), the print is slow 18 to 28 sec and with pjob.setPageable( new PDFPageable(pdDocument,
Orientation.AUTO, true, 0 ) ), there is only marginal improvement in performance of about
15 to 18 sec.
> Why the printable image or PDFBox output PDF file size increases to 1MB, while the original
input PDF file is only 115 KB?

Because PDF is a more efficient format for vector graphics than a raster 
image. You could as well say that a .TXT file of Hamlet's "To be or not 
to be" monologue has a size of less than 2 KB. But a screenshot of its 
display will be much larger.

> PDFBox 1.8.10 was able to generate smaller PDF file and performance is better ( is it
at the cost of fonts as fonts are lighter? ).  Will smaller initial size of PDF to PrinterJob.print()
will make it to print faster?

No, it depends of the contents.

> Will seting java property pdfbox.fontcache to a folder with write permission, will remove
admin privelage requirement to run PDFBox 2.0.0 to build font cache first time and store in
disk subsequent time

Yes

> Is there a way to improve or roadmap to improve performace in 2.0.0, as 2.0.0 has lot
of fixes and improvements over 1.8.10, but 2.0 performance is main blocker for many of users.
> Thanks a lot in advance for PDFBox team

No there is no "roadmap". We try to do our best. The slow printing with 
some files is a known problem, see the issue I mentioned earlier.

We may find improvements by looking at the file. But there is no guuarantee.

Tilman


>
> Regards
> Sridhar Sowmiyanarayanan
> Tata Consultancy Services
> Website: http://www.tcs.com
> ____________________________________________
> Experience certainty.	IT Services
> Business Solutions
> Consulting
> ____________________________________________
>
>
> -----Sridhar So/BLR/TCS wrote: -----
> To: users@pdfbox.apache.org
> From: Sridhar So/BLR/TCS
> Date: 11/14/2015 07:01PM
> Subject: Re: Performance Issue with 2.0.0 SNAPSHOT latest builds
>
> Dear PDFBox team
>
> Thanks for your response to below query.
>
> I set the Java property value pdfbox.fontcache to JDK lib directory where fonts are ther
 ( D:\\Softwares\\Java\\jre7\\lib\\fonts ).
> I set the java property within the code before PDFBox API calls using
> System.setProperty("pdfbox.fontcache", "D:\\Softwares\\Java\\jre7\\lib\\fonts") ;
> First time execution it took 140sec and created the cache file .pdfbox in the given directory
> Subsequent time it took same 24 to 60 secs and no improvement in performance.
> The difference noticed was, earlier  .pdfbox file was created in user.home directory,
now given directory but on performance side no noticable difference.
>
> My OS is MS Windows 7 home basic, since the user id has admin privilage, did not have
admin issue.
> In another machine/work machine with admin user id/prvilage, PDFBox could create font
cache, but the performance was slow ( 30 sec + )
>
> We use only Ariel and Time New Roman. ÿ
> If the default font is chosen as Ariel in 1.8.10 code, that should be fine as 1.8.10
has performance, but fonts are lighter.
> Alternatively if 2.0.0 ÿis optimized for performance, that will be great.ÿ
>
> FYI
> PDF file ( redireting the print to MS XPS Printer type file ), the size of PDF file generated
in 1.8.10 is about 151 KB, whereas the size of PDF file created using 2.0 is 1152 KB ( about
10 times larger ).ÿ
> It job.print() which takes lot of time. ÿTime profile output below
>
> System.getproperty of pdfbox.fontcache = D:\Softwares\Java\jre7\lib\fonts
> System.getproperty of user.home = C:\Users\Rangarajan
> Nov 14, 2015 6:58:22 PM org.apache.pdfbox.cos.COSDocument finalize
> WARNING: Warning: You did not close a PDF Document
> ÿPDDocument load time = 30ms PrinterJob creation time = 10ms job.setPageable ( new PDFPageable(pdDocument)
time = 0ms job.print(); Printing Time = ÿ58129ms Total time = 58.169 secondsÿ
>
>
>
> Regards
> Sridhar Sowmiyanarayanan
> Tata Consultancy Services
> Website: http://www.tcs.com
> ____________________________________________
> Experience certainty.	IT Services
> Business Solutions
> Consulting
> ____________________________________________
>
>
> -----Sridhar So/BLR/TCS wrote: -----
> To: users@pdfbox.apache.org
> From: Sridhar So/BLR/TCS
> Date: 11/14/2015 01:36AM
> Cc: users-owner@pdfbox.apache.org
> Subject: Performance Issue with 2.0.0 SNAPSHOT latest builds
>
> Subject line changed.
>
>
> Sridhar Sowmiyanarayanan
> Tata Consultancy Services
> Website: http://www.tcs.com
> ____________________________________________
> Experience certainty.	IT Services
> Business Solutions
> Consulting
> ____________________________________________
>
>
> -----Sridhar So/BLR/TCS wrote: -----
> To: users@pdfbox.apache.org
> From: Sridhar So/BLR/TCS
> Date: 11/14/2015 01:32AM
> Cc: users-owner@pdfbox.apache.org
> Subject: Re: Returned post for users@pdfbox.apache.org
>
> Dear PDFBox Developers/Contributors
>
>
> Thanks for reply, I tested with latest SNAPSHOT buildsÿ
> https://repository.apache.org/content/groups/snapshots/org/apache/pdfbox/preflight-app/2.0.0-SNAPSHOT/
------> build 1823
> https://repository.apache.org/content/groups/snapshots/org/apache/pdfbox/pdfbox-app/2.0.0-SNAPSHOT/
---------> build 1800 and 1801
> ÿFont cache rebuild is not happening every time and it is able to use from local store,
however build 1800 & 1801 requires Admin privilage.ÿ
> ÿCompared to PDFBox version 1.8.10, 2.0.0 SNAPSHOT builds performance is slow ÿ ( 1.8.10
about 13 seconds, but 2.0.0 SNAPSHOT 1800, 1801 takes 30 seconds. )
>
> ÿMeasured the time difference between 1.8.10 and 2.0.0 SNAPSHOT builds with following
code
> ÿGetting 13 to 19 seconds in 1.8.10, whereas in 2.0.0SNAPSHOT builds 22 to 36 secondsÿ
> ÿÿ
> ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ ÿ long t1 = ÿSystem.currentTimeMillis() ;
> 			pdDocument = PDDocument.load(is);	
> 			long t2 = ÿSystem.currentTimeMillis() ÿ;
> 			PrinterJob job = PrinterJob.getPrinterJob();
> 		 ÿ ÿjob.setPageable(new PDFPageable(pdDocument)); ÿ// ÿVersion 2.0.0 SNAPSHOT builds
> 			//job.setPageable(new PDPageable(pdDocument)); // ÿVersion 1.8.10ÿ
> 		 ÿ ÿjob.print();	
> 			long t3 = ÿSystem.currentTimeMillis() ÿ;
> 			//printWithPaper(pdDocument, "A4") ;
> 			System.out.println ( " PDDocument load time = " ÿ+ÿ
> 				String.valueOf(t2 - t1) + " Printing Time = ÿ" + String.valueOf(t3-t2 ) +
> 				" Total time = " + String.valueOf( (t3 - t1)/1000.0 ) + " seconds " ÿ) ;
> ÿ
>
>
> Do we haveÿ
> Performance fix available in 2.0.0 SNAPSHOTS, if so please give full path and which build?
> Is there a fix available where ÿAdmin privelage is not required?
> Thanks a lot for your reply in advance.
>
> FYI
> ----ÿ
> Alignment issue is still there in 2.0.0 and my attempt to adjust alignment using below
code takes lot of time.ÿ
>
> PageFormat pageFormat = new PageFormat();
> ÿ ÿ ÿ ÿ pageFormat.setOrientation(PageFormat.PORTRAIT);
> ÿ ÿ ÿ ÿ Paper paper= pageFormat.getPaper();
>
> 	if ("SLEEK".equalsIgnoreCase(receiptType)) {
> ÿ ÿ ÿ ÿ ÿ ÿpaperWidth = 3.14;
> ÿ ÿ ÿ ÿ ÿ ÿ paperHeight = 50;
> ÿ ÿ ÿ ÿ } else if ("LETTER".equalsIgnoreCase(receiptType)) {
> ÿ ÿ ÿ ÿ ÿ ÿ paperWidth = 8.5;
> ÿ ÿ ÿ ÿ ÿ ÿ paperHeight = 11;
> ÿ ÿ ÿ ÿ } else if ("LEGAL".equalsIgnoreCase(receiptType)) {
> ÿ ÿ ÿ ÿ ÿ ÿ paperWidth = 8.5;
> ÿ ÿ ÿ ÿ ÿ ÿ paperHeight = 14;
> ÿ ÿ ÿ ÿ }
> ÿ ÿ ÿ ÿ ÿ else
> ÿ ÿ ÿ ÿ {
> ÿ ÿ ÿ ÿ ÿ ÿ paperWidth = 8.3;
> ÿ ÿ ÿ ÿ ÿ ÿ paperHeight = 11.7;
> ÿ ÿ ÿ ÿ }
>
> ÿ ÿ ÿ ÿ paper.setSize(paperWidth * 72.0, paperHeight * 72.0);
> ÿ ÿ ÿ ÿ paper.setImageableArea(-2000, 0, paper.getWidth(), paper.getHeight());
> ÿ ÿ ÿ ÿ pageFormat.setPaper(paper);
>
>
> ÿ ÿ ÿ ÿ PrinterJob job = PrinterJob.getPrinterJob();ÿ
> 	job.setPageable(new PDFPageable(document));ÿ
>
> ÿ ÿ ÿ ÿ Book book = new Book();ÿ
> 	book.append(new PDFPrintable(document), getPageFormat(receiptType) , document.getNumberOfPages());ÿ
> 	
> 	job.setPageable(book);ÿ
> 	job.print();
>
>
> Regards
> Sridhar Sowmiyanarayanan
> Tata Consultancy Services
> Website: http://www.tcs.com
> ____________________________________________
> Experience certainty.	IT Services
> Business Solutions
> Consulting
> ____________________________________________
>
>
>
>
>
> -
>
> To:	users@pdfbox.apache.org
> cc:	commits@pdfbox.apache.org
> Subject:	Re: Speedup Font Cache: Performance Issue in PDFBox 2.0.0-RC1
>
> Dear PDFBox Community
>
> Adding Commits, John and Tilman in the mail request
>
> Details are in below mail.
>
> In brief again
>
> In our system, PDF files generated in server and is sent to client and in client applet
code uses PDFBox to print.
> With PDFBox 1.8.10, we have alignment issue as some characters were missing on left side.
> With PDFBox 2.0.0-RC1, we have faced performance issue ( slow )
>
> Do we have fix or patch availableÿ
> either in 2.0.0-RC1 for performance ÿOR
> alignment issue in 1.8.10
> Our PDF documents uses True Type Fonts, mostly Ariel Unicode
>
> Thanks a lot for your help and support.
>
> Regards
> Sridhar Sowmiyanarayanan
> Tata Consultancy Services
> Website: http://www.tcs.com
> ____________________________________________
> Experience certainty. IT Services
> Business Solutions
> Consulting
> ____________________________________________
>
>
> -----Sridhar So/BLR/TCS wrote: -----
> To: users@pdfbox.apache.org
> From: Sridhar So/BLR/TCS
> Date: 11/12/2015 06:23PM
> Subject: Speedup Font Cache: Performance Issue in PDFBox 2.0.0-RC1
>
> Dear PDFBox Developers/Contributors
>
> I am unable to subscribe to users mailing list as the link tries to open Outlook not
the page to subscribe, hence a seperate mail on similar/same issue discussed.
>
> Issue:
> -------ÿ
> PDFBox2.0.0-RC1 is very slow in printinng ( taking 35 to 50 seconds ) ÿas it tries to
load fonts each time with the following message
>
> WARNING: New fonts found, font cache will be re-built
> Nov 12, 2015 3:17:26 PM org.apache.pdfbox.pdmodel.font.FileSystemFontProvider <init>
> WARNING: Building font cache, this may take a while
> Nov 12, 2015 3:17:32 PM org.apache.pdfbox.pdmodel.font.FileSystemFontProvider saveCache
> WARNING: Finished building font cache, found 522 fonts
>
>
> Is the fix or patch available to avoid slow performance due to above ( rebuilding font
cache ÿevery tme ) ?ÿ
> If the fix not available in 2.0.0-RC1,ÿIs there any way to fix alignment issue in PDFBox
1.8.10? as 1.8.10 left margin is too low and first few characters are found cut in printout.
>
> With PDFBox1.8.10, there is no performance issue, but alignment in prontout is not proper.
ÿÿ
> With PDFBox 2.0.0-RC1, we are facing performance issue.
>
> PDFDocument used has Ariel ÿUnicode or True Type Fonts.ÿ
>
> Similar discussion thread is pasted below, as I was unable to reply to same discussion
thread, nor able to subscribe to users mailing list, hence a seperate mail.
>
> Regards
> Sridhar
>
> Subject:	Re: Speedup Font Cache	
> From:	John Hewson (jo...@jahewson.com)
> Date:	Oct 21, 2015 5:26:41 pm
> List:	org.apache.pdfbox.users
>
> On 21 Oct 2015, at 09:43, Maruan Sahyoun <sahy...@fileaffairs.de> wrote:
>
> Hi,
>
> Am 21.10.2015 um 18:40 schrieb Tilman Hausherr <THau...@t-online.de>:
>
> Am 21.10.2015 um 14:10 schrieb Roberto Nibali:
> Hi John
>
> On Wed, Oct 21, 2015 at 12:35 AM, John Hewson <jo...@jahewson.com> wrote:
>
> Yes, I&#8217;m able to replicate that issue on Windows. It&#8217;s apparently
related
> to administrator ownership of that registry key&#8217;s parent node. Looks like
> it&#8217;ll be necessary to log in as admin and create that key with user access.
> I guess that&#8217;s far from ideal?
>
> The whole issue also happens on MacOSX. When you introduce this on-disk
> cache a couple of months back, it worked fine, however one of the recent
> changes to SVN must have wrecked the initially intended functionality. Not
> only is the font caching setup 5-10 times as long as it used to be, it also
> does not seem to persist it anymore. Version used:
>
> $ svn info | grep -i changed
> Last Changed Author: tilman
> Last Changed Rev: 1709647
> Last Changed Date: 2015-10-20 19:04:02 +0200 (Tue, 20 Oct 2015)
>
> Running my test tool indicates:
>
> Oct 21, 2015 2:08:29 PM
> org.apache.pdfbox.pdmodel.font.FileSystemFontProvider loadCache
> WARNING: New fonts found, font cache will be re-built
> Oct 21, 2015 2:08:29 PM
> org.apache.pdfbox.pdmodel.font.FileSystemFontProvider <init>
> WARNING: Building font cache, this may take a while
> Oct 21, 2015 2:08:39 PM
> org.apache.pdfbox.pdmodel.font.FileSystemFontProvider saveCache
> WARNING: Finished building font cache, found 654 fonts
> [INFO, ctx=./ccalt.pdf]: Opening Source ./ccalt.pdf
> [INFO, ctx=./ccalt.pdf]: Opening Template ./cctemp.pdf
> [INFO, ctx=./ccalt.pdf]: Writing Output ./ccmig.pdf
> [INFO, ctx=./ccalt.pdf]: Completed in 15037.02ms
>
> This used to be anything between 1200ms and 2300ms and once it was
> persisted onto disk, it was rather fast in subsequent calls. Unfortunately,
> SVN does not provide the handy tool of "git bisect" to quickly find out
> which change actually caused this regression.
>
> There were only 4 changes since then, so it might be worth a try to just revert
> that file.
>
> (I can't help; for me, it has always been slow.)
>
> Could it be that 1) you installed new stuff on your computer, 2) that MacOS has
> many of its fonts in .ttc files? In Windows there are only 10.
>
> on my OS X I have 92 ttc files (out of 384) :-)
>
> Yep, OS X uses ttc much more heavily than Windows and some of those are big
> Asian fonts which PDFBox parses relatively slowly.
>
> &#8212; John
>
> BR
> Maruan
>
> Tilman
>
> Let me know if you need any further input.
>
> Cheers
> Roberto
>
>
>
> Regards
> Sridhar Sowmiyanarayanan
> Tata Consultancy Services
> Website:ÿhttp://www.tcs.com
> ____________________________________________
> Experience certainty. IT Services
> Business Solutions
> Consulting
> ____________________________________________
> =====-----=====-----=====
> Notice: The information contained in this e-mail
> message and/or attachments to it may contain
> confidential or privileged information. If you are
> not the intended recipient, any dissemination, use,
> review, distribution, printing or copying of the
> information contained in this e-mail message
> and/or attachments to it are strictly prohibited. If
> you have received this communication in error,
> please notify us by reply e-mail or telephone and
> immediately and permanently delete the message
> and any attachments. Thank you
>
>


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@pdfbox.apache.org
For additional commands, e-mail: users-help@pdfbox.apache.org


Mime
View raw message