search

giellalt / bugzilla-dummy

0 stars 0 forks source link

Pruning the default filter set (Bugzilla Bug 1717) #1566

Closed albbas closed 11 years ago

albbas commented 11 years ago

This issue was created automatically with bugzilla2github

Bugzilla Bug 1717

Date: 2013-10-05T10:26:25+02:00 From: Trond Trosterud <> To: Sjur Nørstebø Moshagen <> CC: sjur.n.moshagen

Last updated: 2013-10-11T16:56:14+02:00

albbas commented 11 years ago

Comment 8535

Date: 2013-10-05 10:26:25 +0200 From: Trond Trosterud <>

We should consider whether the default set of filters to be compiled should be pruned. (src/filters/Makefile.am)

The problematic filter is remove-illegal-derivation-strings.xfst.

Compiling the out-of-the-box dummy fileset without this filter takes

real 0m16.990s user 0m27.543s sys 0m2.620s

With it the operation takes

real 7m52.636s user 8m0.839s sys 0m2.908s

(oh, well, two cups of coffee....)

The filter itself is

~[ $[ %+Der1 ? %+Der1 ] | $[ %+Der2 ? %+Der1 ] | $[ %+Der2 ? %+Der2 ] | $[ %+Der3 ? %+Der1 ] ...

i.e., a pretty Saami-specific filter. So my suggestion is to leave it out of the default list of filters to be compiled in the src/filters/Makefile.am setup.

albbas commented 11 years ago

Comment 8536

Date: 2013-10-08 15:32:40 +0200 From: Sjur Nørstebø Moshagen <>

This specific filter is only an issue with certain older versions of the Xerox tools. With a recent version of the Xerox tools this filter takes about the same time to compile:

real 0m12.834s user 0m9.170s sys 0m1.037s

(tested on the XServe, including the problematic filter).

That is, the solution is to upgrade the Xerox tools.

But yes, it is a pretty language-specific filter, so it may make sense to move it to the relevant languages only.

albbas commented 11 years ago

Comment 8568

Date: 2013-10-11 16:56:14 +0200 From: Sjur Nørstebø Moshagen <>

(In reply to comment #1)

But yes, it is a pretty language-specific filter, so it may make sense to move it to the relevant languages only.

The filter has been removed from all languages but SME, SMA and SMJ, and should not pose a time issue for any other language, irrespective of the version of the Xerox tools installed.

For the mentioned Sámi languages - if you still have issues compiling this filter (ie it takes more than a minute, even up to hours), then please consider installing the newest version of the Xerox tools, or skip Xerox and switch to HFST.