Eclipse Community Forums: BIRT » How to use iText for PDF generation in M3 Release instead of FOP

Home » Archived » BIRT » How to use iText for PDF generation in M3 Release instead of FOP

How to use iText for PDF generation in M3 Release instead of FOP [message #101556]

Thu, 22 December 2005 05:03

Eclipse User

Originally posted by: ashwin.patil.cash-tech.com

Hi All
I'd like to ask what neccessary configuration steps must be done in order
PDF production went through the new PDF emitter instead of FOP. I've
deployed M3 and FOP seems to be active by default.If i remove fop the PDF
is not getting generated .I have put itext-1.3.jar in
\plugins\org.eclipse.birt.report.engine.emitter.fo_1.0.1/lib but what i
found that the M3 release is still using FOP .

Waiting for the Reply.
Regards
Ashwin

Re: How to use iText for PDF generation in M3 Release instead of FOP [message #103895 is a reply to message #101556]

Wed, 04 January 2006 07:56

Eclipse User

Hello,

I never made official M3 to render PDF via iText library. A few days ago I
downloaded recent pre-RC night build (from 31.12.2005) and to my
satisfaction this version DOES USE iText by default. Nice :-)

However the performance is below my expectation. I expected increase of
speed compared to FOP, but unfortunatelly - and to my suprise the recent
PDF emitter performs significantly worse than the old FOP emitter. The
reason is not known to me.

|=========================================================== ================|
| BIRT-release | Report | TimeConsumed | PageCount | Comment
|
|=========================================================== ================|
| M3 (6.12.) | custom | ~16 seconds | 36 |
appendix-A |
| pre-RC (31.12.) | custom | ~60 seconds | 30 |
appendix-B |
|----------------------------------------------------------- ----------------|

I used custom report and custom ODA plugin for data, custom console
application for starting print job (BIRT engine engaged through public
API) @ Sun JDK 1.4 and 1.5.

appendix-A:
- experienced interference of page body with page footer (partly overlaps)
- page number element print "1" on all pages

appendix-B
- slightly different optics
- row border drawn incorectly (expected: border should occupy whole row
width, actual: only a fraction of row is framed by border)
- page numbers works well
- print date/timestamp on footer not identical for all pages (expected:
date/time value should be same on all pages, actual: timestamp is derived
from system clock for each page individually)

Re: How to use iText for PDF generation in M3 Release instead of FOP [message #104655 is a reply to message #103895]

Thu, 05 January 2006 03:53

Eclipse User

Hi again,

I realized that BIRT refered by me as pre-RC is actually RC (same nightly
build). I performed the tests once again with the most recent nightly
build (N20060104) and I am still experiencing significant performance
penalty compared to older FOP way (M3 ~4x faster than RC in my scenario).
I also tried to copy FO emitter from M3 to RC so I have a decent printing
performance as earlier, but it does not seem to work anymore in RC
(NullPointer exception arised during printing).

I have uploaded PDF files from my M3-RC benchmark to web for case you're
interested.
http://www.volny.cz/tvala/birt/

I guess now i am going to create the most primitive report layout -> one
grid with one column then redo benchark (that prints 512 rows) once again.

Any feedback is welcome.

Cheers,
Tomas

M3 vs RC, PDF generator benchmark [message #104722 is a reply to message #104655]

Thu, 05 January 2006 04:48

Eclipse User

new report created
1x custom data source
1x custom data set featuring 1 column

Dataset printed in grid, 512 rows provided by ODA plugin. Result here:
http://www.volny.cz/tvala/birt/benchmark/

2.0 M3 ** Time taken for print job 10891 milliseconds
2.0 RC ** Time taken for print job 25000 milliseconds

Could anyone competent confirm that these are normal results or something
goes wrong on my side? In case i am experiencing a known issue of RC, can
i expected an extra printing performance in 2.0 FINAL (I mean equal or
better performance than M3 with FOP as absolute minimum)?

Thanx in advance for any reaction :)

Tomas

Re: M3 vs RC, PDF generator benchmark [message #105072 is a reply to message #104722]

Thu, 05 January 2006 15:22

Eclipse User

Tomas,

Thanks for the great efforts to test out these reports. We are looking
into the performance issues, but can not guarantee that the performance
will match FO in 2.0 final release.

BIRT 2.0 engine API now supports multiple emitters outputting same format,
i.e., PDF. The IRenderTask and IRunAndRenderTask now support setEmitterID
method, so that you can pick an emitter. So by modifying viewer, you may
keep two emitters coexist. ANother way to keep both is to modify the FO
emitter, for example, to output a format FOPDF. Then the two can coexist
without the setEmitterID call.

Hope this helps.

Stanley Wang
BIRT Engine

Re: M3 vs RC, PDF generator benchmark [message #105186 is a reply to message #104722]

Thu, 05 January 2006 18:06

Eclipse User

Originally posted by: swang.actuate.c

Tomas,

We ran some tests with a JDBC data source, and found out that the two PDF
emitters are more comparable in performance. We can not run your report
because you use your own ODA driver. Do you mind creating another report
using, for example, the sample DB, and see what performance numbers you
can get? Or if you do not mind, can you share the ODA driver so that we
can run your tests?

Stanley Wang
BIRT Engine

Re: M3 vs RC, PDF generator benchmark [message #105341 is a reply to message #105072]

Fri, 06 January 2006 05:14

Eclipse User

Hello Stanley,

Thanks a lot for information. I will look closely on setEmitter mechanism
in near future. Having support for two or more coexisting PDF (or whatever
else) emitters is nice feature since I prefer to order BIRT engine to use
certain emitter for certain task.

One question: are there any plans to implement a working Page-Number
mechanism in FOP PDF Emitter? Having no page numbers prevents me from
using it for reports longer than one page, unfortunately. That's business
requirement to have proper page numbers.

I have a concerns regarding not 100% precise WYSIWYG previews and PDF
finals. Each PDF emitter means different document optics. It's complicated
or even impossible to design professional top notch report that looks
perfect when printed through different emitter. From my perspective it's
big effort to tailor report for specific emitter and consequently maintain
it later when a new emitter is available or updated. Simply the relation
between report design and particular emitter is too close. One example:
interpretation of item margin and padding is way too different in old and
new emitter. Howewer I fully understand that things are still under
development and I am looking forward to fortcoming releses, for sure! :)

Cheers,
Tomas

Re: M3 vs RC, PDF generator benchmark [message #105356 is a reply to message #105186]

Fri, 06 January 2006 05:19

Eclipse User

Yes indeed, I will change my report to use standard XML or perhaps ODBC
plugin and redo tests. Later...

For political and technical reasons I can't share my plugin at the moment.

Tomas

New benchmark done through standard XML ODA driver [message #105467 is a reply to message #105356]

Fri, 06 January 2006 08:05

Eclipse User

Hi Stanley,

I changed my trivial report to use XML ODA datasource. The performance gap
between M3 a RC is still considerable.

M3 ** Time taken for print job 14703 milliseconds
RC ** Time taken for print job 41140 milliseconds

All files:
http://www.volny.cz/tvala/birt/benchmark/xml-driver/

I have noticed there's a difference in implementation of XML plugin in
2.0RC. Reports (precisely XML datasources and datasets) created in M3 no
longer function in RC. That's way i have to made a specific report for M3
and RC, though both contain the same body, only datasources are defined
accordingly to requirements of M3 vs RC release. You will have to copy
"VehicleSpecAndValue - ICV PVS Output Example 1.0 options.xml" into C:\
then run report the way you are used to. I am curious for result of your
measurement.

Thanks for interest.

Tomas

Re: M3 vs RC, PDF generator benchmark [message #105665 is a reply to message #104722]

Fri, 06 January 2006 15:41

Eclipse User

For what its worth, I'm having similar issues with PDF generation. I'm
using a standard JDBC connection to SQL Server. I've confirmed that
Tomcat is monopolizing CPU when PDF is generated-it isn't a data issue.
The query in SQL Server is very efficient. On M2, run time was about 10
sec, is now 30-40 sec.

Thanks-

Jett Hanna

Previous Topic:	Feature Request: Form-Elements
Next Topic:	Preview v. Deploy Formatting Differences

Goto Forum:

-=] Back to Top [=-

Current Time: Sun Jun 01 07:21:54 EDT 2025

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter