Hi everyone, I am trying to put together a page showing the status of static rendering between the various free SVG packages out there. Mostly to make sure that librsvg and inkscape renders images as identically as possible.
But Inkscape refuse to render most of the w3c testsuite from the command line. I am using this command line: inkscape -z --file=$i --export-png= $NAME_NO_EXT.png --export-width=160 --export-height=120
Which generate the result you can see on the current testpage:
http://www.linuxrising.org/svg_test/test.html
Any suggestions?
Christian
On 8/20/05, christian@...693... <christian@...693...> wrote:
Any suggestions?
The reason is that Inkscape uses the width/height attrs to determine what to export, whereas these test files have them set to "100%". In this case Inkscape must take the width/height from viewBox but it does not. I will try to fix this now.
Thanks for the testing page, btw, it's very useful.
On Sat, 2005-08-20 at 21:18 +0200, christian@...693... wrote:
Hi everyone, I am trying to put together a page showing the status of static rendering between the various free SVG packages out there. Mostly to make sure that librsvg and inkscape renders images as identically as possible.
But Inkscape refuse to render most of the w3c testsuite from the command line. I am using this command line: inkscape -z --file=$i --export-png= $NAME_NO_EXT.png --export-width=160 --export-height=120
Which generate the result you can see on the current testpage:
http://www.linuxrising.org/svg_test/test.html
Any suggestions?
Are you using Bryce's rendertest code? Hopefully you both can integrate these testing methods.
Great to see more testing!
Jon
Hi Jon, I am using some very simple shell scripts that runs a for loop on the w3c testimages applying the various command line tools available to generate png's :) I have no idea what Bryce's rendertest code is or what it can do.
Christian
On Sat, 2005-08-20 at 13:32 -0700, Jon Phillips wrote:
On Sat, 2005-08-20 at 21:18 +0200, christian@...693... wrote:
Hi everyone, I am trying to put together a page showing the status of static rendering between the various free SVG packages out there. Mostly to make sure that librsvg and inkscape renders images as identically as possible.
But Inkscape refuse to render most of the w3c testsuite from the command line. I am using this command line: inkscape -z --file=$i --export-png= $NAME_NO_EXT.png --export-width=160 --export-height=120
Which generate the result you can see on the current testpage:
http://www.linuxrising.org/svg_test/test.html
Any suggestions?
Are you using Bryce's rendertest code? Hopefully you both can integrate these testing methods.
Great to see more testing!
Jon
Christian Fredrik Kalager Schaller wrote:
Hi Jon, I am using some very simple shell scripts that runs a for loop on the w3c testimages applying the various command line tools available to generate png's :) I have no idea what Bryce's rendertest code is or what it can do.
Here is some info on Bryce's rendertest http://sourceforge.net/mailarchive/forum.php?thread_id=7701948&forum_id=...
Hmm, looks interesting, I try to see if I can get the visualdiff numbers extracted and added to the webpage somehow.
Christian
On Mon, 2005-08-22 at 12:17 +0300, Nicu Buculei wrote:
Christian Fredrik Kalager Schaller wrote:
Hi Jon, I am using some very simple shell scripts that runs a for loop on the w3c testimages applying the various command line tools available to generate png's :) I have no idea what Bryce's rendertest code is or what it can do.
Here is some info on Bryce's rendertest http://sourceforge.net/mailarchive/forum.php?thread_id=7701948&forum_id=...
Hi, Ok, I have now expanded a lot of the SVG comparison tables based on the feedback I got.
I removed KSVG from the table as the version I had available was the now unmaintained ksvg1 engine (which is the one in KDE 3.4.2). Will look into adding a KSVG2 column later.
Also have split librsvg into two columns, one for libart backend and one for Cairo backend. Turned out I was using the libart backend before not the Cairo one as I originally thought.
Also added versions of the pages using big png's as requested by Nicu among others.
The pages can now be found here: http://www.linuxrising.org/svg_test/index.html
Please let me know if there are any mistakes or errors somewhere.
Christian
Known issues: There seems to be a bug in batik/rasterizer which cause it not to give the same size images as the other libraries in the big version.
Also there is a bug in the svg2png (xsvg) which cause it to quadratize the graphics.
christian@...693... wrote:
Hi everyone, I am trying to put together a page showing the status of static rendering between the various free SVG packages out there. Mostly to make sure that librsvg and inkscape renders images as identically as possible.
But Inkscape refuse to render most of the w3c testsuite from the command line. I am using this command line: inkscape -z --file=$i --export-png= $NAME_NO_EXT.png --export-width=160 --export-height=120
Which generate the result you can see on the current testpage:
http://www.linuxrising.org/svg_test/test.html
Any suggestions?
Christian
Keep in mind that neither Inkscape nor librsvg render SVG completely or totally correctly. The best renderers that we know of, are Batik and Adobe. We should not consider librsvg to be a benchmark, but compare both librsvg and Inkscape to those, and try to emulate them in their good qualities, else work toward the SVG spec directly.
Bob
I think you should add Deerpark(Beta 2) rendering to the mix; I have been doing so informally and I find inkscape's rendering of large svg files to be only slightly slower than batik's rendering, and Deerpark's rendering was also correct in my tests but notably slower than either batik or inkscape, but acceptably fast, especially for most web uses. I was happy to see that inkscape's rendering of complex svg files has improved notably since the last time I did this several releases ago.
./firefox some_file.svg
displays some_file.svg for the Deerpark Beta2 release of Firefox. If you can figure out how to add Firefox(Deerpark) to the web page it would be very useful for this important renderer, It would be very interesting to run Deerpark through all your tests.
christian@...693... wrote:
Hi everyone, I am trying to put together a page showing the status of static rendering between the various free SVG packages out there. Mostly to make sure that librsvg and inkscape renders images as identically as possible.
But Inkscape refuse to render most of the w3c testsuite from the command line. I am using this command line: inkscape -z --file=$i --export-png= $NAME_NO_EXT.png --export-width=160 --export-height=120
Which generate the result you can see on the current testpage:
http://www.linuxrising.org/svg_test/test.html
Any suggestions?
Christian
SF.Net email is Sponsored by the Better Software Conference & EXPO September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf _______________________________________________ Inkscape-devel mailing list Inkscape-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/inkscape-devel
Hi, I would like to add Firefox, but after talking to one of the devs this weekend I don't think there is a nice way to do so for me currently. I did add a page which embeds the actual svg image though, so if you use deerpark to view that page you will get a 'firefox' collumn.
Christian
On Sat, 2005-08-20 at 17:57 -0500, inkblotter wrote:
I think you should add Deerpark(Beta 2) rendering to the mix; I have been doing so informally and I find inkscape's rendering of large svg files to be only slightly slower than batik's rendering, and Deerpark's rendering was also correct in my tests but notably slower than either batik or inkscape, but acceptably fast, especially for most web uses. I was happy to see that inkscape's rendering of complex svg files has improved notably since the last time I did this several releases ago.
./firefox some_file.svg
displays some_file.svg for the Deerpark Beta2 release of Firefox. If you can figure out how to add Firefox(Deerpark) to the web page it would be very useful for this important renderer, It would be very interesting to run Deerpark through all your tests.
christian@...693... wrote:
Hi everyone, I am trying to put together a page showing the status of static rendering between the various free SVG packages out there. Mostly to make sure that librsvg and inkscape renders images as identically as possible.
But Inkscape refuse to render most of the w3c testsuite from the command line. I am using this command line: inkscape -z --file=$i --export-png= $NAME_NO_EXT.png --export-width=160 --export-height=120
Which generate the result you can see on the current testpage:
http://www.linuxrising.org/svg_test/test.html
Any suggestions?
Christian
SF.Net email is Sponsored by the Better Software Conference & EXPO September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf _______________________________________________ Inkscape-devel mailing list Inkscape-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/inkscape-devel
That's cool. In general there are at least two tyes of testing that I think we should have. The static redering tests that we now have, which one way or another should include the Deerpark results. The other is a performance testing with the same rederers, perhaps less ambitious, that would compare redering times between the players.
I have been playing with the wonderful clip art browser, rough around the edges but still wonderful, and it provides a lot of real-world examples useful for both types of testing, but moreso for static redering testing. It would be nice if we could get all these rederers displaying all the svg clipart in OCAL correctly, where correctly doesn't include emmitting streams of warnings or exceptions either. This is currently not at all true for various reasons.
Christian Fredrik Kalager Schaller wrote:
Hi, I would like to add Firefox, but after talking to one of the devs this weekend I don't think there is a nice way to do so for me currently. I did add a page which embeds the actual svg image though, so if you use deerpark to view that page you will get a 'firefox' collumn.
Christian
On Sat, 2005-08-20 at 17:57 -0500, inkblotter wrote:
I think you should add Deerpark(Beta 2) rendering to the mix; I have been doing so informally and I find inkscape's rendering of large svg files to be only slightly slower than batik's rendering, and Deerpark's rendering was also correct in my tests but notably slower than either batik or inkscape, but acceptably fast, especially for most web uses. I was happy to see that inkscape's rendering of complex svg files has improved notably since the last time I did this several releases ago.
./firefox some_file.svg
displays some_file.svg for the Deerpark Beta2 release of Firefox. If you can figure out how to add Firefox(Deerpark) to the web page it would be very useful for this important renderer, It would be very interesting to run Deerpark through all your tests.
christian@...693... wrote:
Hi everyone, I am trying to put together a page showing the status of static rendering between the various free SVG packages out there. Mostly to make sure that librsvg and inkscape renders images as identically as possible.
But Inkscape refuse to render most of the w3c testsuite from the command line. I am using this command line: inkscape -z --file=$i --export-png= $NAME_NO_EXT.png --export-width=160 --export-height=120
Which generate the result you can see on the current testpage:
http://www.linuxrising.org/svg_test/test.html
Any suggestions?
Christian
OK, I fixed this in CVS. Christian, please compile CVS (after a few hours, when anon CVS is updated) and rerun your tests. By the way, your page does not include all tests - I see no text tests there.
Attention everyone: Now, when svg width/height are set in % and at the same time there's a viewBox attribute, the canvas size on load is determined relative to the size of the viewBox (i.e. equal to viewBox when width=height=100%). Before, the canvas in this case was always A4. I think this is the correct behavior, but let me know if you disagree.
Christian,
Thanks for updating the test page with Inkcape CVS renderings. I have a few comments:
- Can you make the images larger? It's difficult to compare in such small size. If the page has to scroll horizontally I think it's OK.
- The W3C images are scaled in the browser, which makes them look ugly and makes comparing with Inkscape's AA images more difficult. Can you prescale them with AA?
- Inkscape renderings show "broken image" often. I just tested masking-path-04-b, struct-image-01-t and render-groups-01-b, they show correct imported bitmaps for me in Inkscape, both in GUI and in commandline rendering. Can you check what's wrong with your setup to fix this?
Hi Bulia, Thanks for the quick fix for the command line rendering of the test files, much appreciated.
I would prefer not to make the images larger, the current size is based on the size of the majority of the test png images from the w3c webpage. I have scaled down the few images that where larger now using Gimp so hopefully the image quality is better.
Fix the broken image problem, my svg generation script was being to clever for its own good :)
Anyway, if you want I can make the png images links to the w3c original testimages, that what you at least have access to the higher resolution version for those images who have such. Or I can make it a link to the html harnessed testpage. Tell me what you prefer.
Sincerely, Christian
On Sun, 2005-08-21 at 17:09 -0300, bulia byak wrote:
Christian,
Thanks for updating the test page with Inkcape CVS renderings. I have a few comments:
- Can you make the images larger? It's difficult to compare in such
small size. If the page has to scroll horizontally I think it's OK.
- The W3C images are scaled in the browser, which makes them look ugly
and makes comparing with Inkscape's AA images more difficult. Can you prescale them with AA?
- Inkscape renderings show "broken image" often. I just tested
masking-path-04-b, struct-image-01-t and render-groups-01-b, they show correct imported bitmaps for me in Inkscape, both in GUI and in commandline rendering. Can you check what's wrong with your setup to fix this?
On 8/22/05, Christian Fredrik Kalager Schaller <christian@...693...> wrote:
I would prefer not to make the images larger, the current size is based on the size of the majority of the test png images from the w3c webpage. I have scaled down the few images that where larger now using Gimp so hopefully the image quality is better.
Well, at pages linked from http://www.w3.org/Graphics/SVG/Test/20030813/htmlframe/full-index.html all images are 480x360. What if you link each w3c image to the corresponding test page (such as http://www.w3.org/Graphics/SVG/Test/20030813/htmlframe/full-pservers-grad-01...), and for other renderers, link each image to a separate 480x360 version? Then the images in the table would work as thumbnails - convenient to quickly assess rendering but giving access to the full version to examine details.
participants (8)
-
unknown＠example.com -
Bob Jamison -
bulia byak -
Christian Fredrik Kalager Schaller -
inkblotter -
inkbottle -
Jon Phillips -
Nicu Buculei