W3C home > Mailing lists > Public > xproc-dev@w3.org > June 2020

Re: Calabash slow from command line

From: Imsieke, Gerrit, le-tex <gerrit.imsieke@le-tex.de>
Date: Sat, 27 Jun 2020 19:44:40 +0200
To: xproc-dev@w3.org
Message-ID: <7147a9c0-a16c-dc0b-af91-8ce0e082250e@le-tex.de>
How do you do the directory listing in XSLT and XProc? Each time with 
collection() from within XSLT, or do you use p:directory-list, maybe 
recursively? I experienced some performance issues when (recursively) 
listing directories with thousands of files in it using p:directory-list.

On 27.06.2020 18:45, Robert Stuart wrote:
> Oxygen does take quite a while to start up.
> 
> eliminating my pipeline itself from the timing I’m not sure that I can blame just the startup of the JVM maybe the time to load Calabash is huge compared to just Saxon? Oxygen is certainly much longer than just Saxon.
> For example doing
> java -cp "../BuildDependencies/xmlcalabash-1.2.1-99/*:../BuildDependencies/xmlcalabash-1.2.1-99/lib/*” com.xmlcalabash.drivers.Main
> takes about 11 seconds to return the help info
> where
> java -jar ../BuildDependencies/SaxonHE9-9-1-7J/saxon9he.jar
> is around .75 seconds to return it’s help info
> 
> I’ll look at piperack but I think I’d have to rework my pipeline a bit to use it. Right now it grabs a directory listing of a bunch of files and iterates over them. It also outputs interim results. My first quick read of piperack sounds like I would have to move the looping outside of XProc and maybe split the piple into several so that the interim results could be returned from a standalone pipeline.
> Am I misreading piperack?
> 
> 
>> On Jun 27, 2020, at 10:12, Norman Tovey-Walsh <ndw@nwalsh.com> wrote:
>>
>> Robert Stuart <bobstuart@mac.com> writes:
>>> I run a simple pipeline to run 3 xsl on 10 files  from inside Oxygen and it takes ~9 seconds.
>>> I run the same from the command line using
>>> java -cp ../BuildDependencies/xmlcalabash-1.2.1-99/*.jar com.xmlcalabash.drivers.Main test.xpl
>>> and it takes ~51 seconds
>>>
>>> Is this expected for some reason? Do I have something not “right”?
>>
>> How long does Oxygen take to start up on your machine? The only
>> practical difference between those two should be the JVM startup tax. If
>> you’re of a mind to experiment a little, you could try configuring the
>> pipeline to run in piperack[1] which would also mitigate the JVM startup
>> overhead.
>>
>>                                         Be seeing you,
>>                                           norm
>>
>> [1] https://xmlcalabash.com/docs/reference/piperack.html
Received on Saturday, 27 June 2020 17:45:01 UTC

This archive was generated by hypermail 2.4.0 : Saturday, 27 June 2020 17:45:02 UTC