XProc-pipeline with oxygen faster than with calabash stand-alone

[martin42]

Why is a XProc-pipeline with oxygen faster than with calabash stand-alone? Or in other words, how can I change the settings in oxygen to "emulate" calabash stand-alone?

Even with small XML-input files oxygen is 5-times faster than running calabash from command line.

[Mircea]

Hello Martin,

We use the Calabash API in order to provide the XProc transformation support.
Basically there is no way to "slow" it down.

You can obtain the same result if you configure the Calabash as an external tool.
See: https://www.oxygenxml.com/doc/versions/ ... tools.html

Regards,
Mircea.

[martin42]

Thank you very much! But may I ask what makes the difference?
When I run the xsl-file
just_copy_input.xsl
calabash-stand-alone is as fast as oxygen with calabash.

Input size: 62 MB
xslt 2.0
Saxon PE 9.8.0.8 in oxygen
Saxon HE 10.2 used by calabash

The Xproc pipeline consists only of one p:xslt job.
calabash-stand-alone: 8 seconds
oxygen with calabash: 8 seconds

But as soon as I change the xsl-file from just_copy_input.xsl to a more "complex" stylesheet the performance of calabash-stand-alone is disappointing.
calabash-stand-alone: 330 seconds
oxygen with calabash: 45 seconds

Does oxygen do any optimizations behind the curtain?

[Mircea]

Hello Martin,

You are comparing two different versions of Calabash actually. The one that uses Saxon 9.8.0.8 and the one that uses Saxon 10.2.
The calabash itself has the same version ?
The last version of Calabash that supports Saxon 9.8 and 10 is 1.2.1.
If you are not using the same version of Calabash and the same version of Saxon that might be the cause.

Regards,
Mircea.

[martin42]

Yes, you are right.
Now I have tried calabash stand alone in version 1.1.19-98.
But it's the same performance difference, no changes.

[crawford47]

In my case calabash stand-alone is way more faster.

[martin42]

Thank you, crawford47! In the interest of safety, did you test calabash in version 1.1.19-98?