apertium-filter-rules for lsx files about apertium-separable

[mr-martian]

I meant that I thought lsx-comp had a command-line option equivalent to lt-comp's -v option, so the recipe would just be

lsx-comp -v oci apertium-oci-fra.oci-fra.l1x oci-fra.autosep1.bin

In fact it has all the code for that and the argument just hasn't been added yet - I can do that tonight or tomorrow.

[mr-martian]

so, because command-line argument parsing is terrible, -v="oci@gascon" selects the variant =oci@gascon while -v "oci@gascon" selects the variant oci@gascon. I think --var="oci@gascon" would do oci@gascon, but my local copy isn't compiling with -- options for some reason, so I can't double check that at the moment.

[mr-martian]

It looks to me like lsx-comp should already support direction restrictions the same way lt-comp does - I don't think it filters on <pardef> though.

[hectoralos]

There is probably a bug. There isn't any syntactical error when I add a v="oci" to an <e> statement, but the output file is almost empty:

$ make
apertium-filter-rules oci apertium-oci-fra.oci-fra.l1x >.deps/oci-fra.l1x
lsx-comp lr .deps/oci-fra.l1x oci-fra.autosep1.bin
.deps/oci-fra.l1x:32: parser error : Document is empty
      pas
      ^
Error: Parse error at the end of input.

As for adding a filter to <pardef>, I'm not really sure if it could be needed. I would add it just in case if adding it at the same time as to <e> is easy.

[hectoralos]

@mr-martian, couldn't you find time for this? It would be helpful (I dare to insist because you said "tonight or tomorrow").

[mr-martian]

@hectoralos sorry I totally forgot about this. It's done and should be in the nightly builder in a few hours I think.

[hectoralos]

Unfortunately, there are no changes after installing the nightly builder:

$ make
apertium-filter-rules oci apertium-oci-fra.oci-fra.l1x >.deps/oci-fra.l1x
lsx-comp lr .deps/oci-fra.l1x oci-fra.autosep1.bin
.deps/oci-fra.l1x:32: parser error : Document is empty
      pas
      ^
Error: Parse error at the end of input.

[mr-martian]

Sorry for the lack of clarity, the change is that lsx-comp now has the same -v option as lt-comp rather than adding it to apertium-filter-rules.

[hectoralos]

Sorry, but I can't see how it works. I have to specify somewhere the filter (oci or oci@gascon, in my case). If I do

$ lsx-comp -h
USAGE: lsx-comp [ -v ] [direction] dictionary_file(s) output_bin_file
  -v, --var:           specify language variant
  -h, --help:          print this help and exit
Directions:
  lr:     left-to-right compilation
  rl:     right-to-left compilation

There is no information.

(Besides, if I run lsx-comp without anything, I get a core dump.)

If I run It-comp, I don't get any information too about any v parameter for the language variety. So I went to see what is done in the Catalan dictionaries, and I see:

$(PREFIX1C).autobil.bin: .deps/$(PREFIX1).cat.dix
        apertium-validate-dictionary $<
        lt-comp --var-right=cat lr $< $@

So it seems this is --var-right or something the like. Should I use --var-right or --var-left? I can see any significance of right of left for language in lsx-comp.

[mr-martian]

$ lt-comp -h
lt-comp v3.6.2: build a letter transducer from a dictionary
USAGE: lt-comp [-mavhlr] lr | rl dictionary_file output_file [acx_file]
  -m, --keep-boundaries:     keep morpheme boundaries
  -v, --var:                 set language variant
  -a, --alt:                 set alternative (monodix)
  -l, --var-left:            set left language variant (bidix)
  -r, --var-right:           set right language variant (bidix)
  -H, --hfst:                expect HFST symbols
  -S, --no-split:            don't attempt to split into word and punctuation transducers
Modes:
  lr:     left-to-right compilation
  rl:     right-to-left compilation

Since you're using lsx in a pair, it probably makes sense to add --var-right and --var-left in case something depends both on which variety you're translating from and which one you're translating into, but at the moment you should be able to do

<e v="oci"> ... </e>

lsx-comp --var oci lr apertium-oci-fra.oci-fra.l1x oci-fra.autosep1.bin

[hectoralos]

Excellent! It's working. Thanks a lot.

As for --var-right and --var-left, according to the pipeline, the module will be used in places where there is information about one single language. But maybe, someone could find useful to use it in another place. In any case, they don't hurt.

[hectoralos]

@mr-martian , I have a problem. I have this rule:

    <e lm="far mestièr">
      <p><l>far<s n="vblex"/></l><r>far<g><b/>mestièr</g><s n="vblex"/></r></p>
      <i><t/><d/></i>
      <par n="adv"/>
      <p><l>mestièr<s n="n"/><s n="m"/><s n="sg"/><d/></l><r></r></p>
    </e> 

I compile:

$ make
lsx-comp lr -v="oci" apertium-oci-fra.oci-fra.l1x oci-fra.autosep1.bin
main@standard 2600 3655

I run a test:

$ echo "^far<vblex><imp><p2><sg>$ ^pas<adv>$ ^mestièr<n><m><sg>$" | lsx-proc '/home/hector/apertium/apertium-oci-fra/oci-fra.autosep1.bin'
^far# mestièr<vblex><imp><p2><sg>$ ^pas<adv>$

It is working.

Now I add v="oci":

    <e lm="far mestièr" v="oci">
      <p><l>far<s n="vblex"/></l><r>far<g><b/>mestièr</g><s n="vblex"/></r></p>
      <i><t/><d/></i>
      <par n="adv"/>
      <p><l>mestièr<s n="n"/><s n="m"/><s n="sg"/><d/></l><r></r></p>
    </e> 

I compile:

$ make
lsx-comp lr -v="oci" apertium-oci-fra.oci-fra.l1x oci-fra.autosep1.bin
main@standard 2590 3638

I run the same test:

$ echo "^far<vblex><imp><p2><sg>$ ^pas<adv>$ ^mestièr<n><m><sg>$" | lsx-proc '/home/hector/apertium/apertium-oci-fra/oci-fra.autosep1.bin'
^far<vblex><imp><p2><sg>$ ^pas<adv>$ ^mestièr<n><m><sg>$

It is not working.

Any idea?

[mr-martian]

The -v=oci part has to be before the lr due to the way command-line options are parsed. If I switch the order of those it fixes the issue for me.

[hectoralos]

This is really strange because I still have the same problem after putting -v="oci" before lr:

$ make
lsx-comp -v="oci" lr apertium-oci-fra.oci-fra.l1x oci-fra.autosep1.bin
main@standard 2590 3638
lsx-comp -v="oci@gascon" lr apertium-oci-fra.oci-fra.l1x [email protected]
main@standard 2590 3638
$ echo "^far<vblex><imp><p2><sg>$ ^pas<adv>$ ^mestièr<n><m><sg>$" | lsx-proc '/home/hector/apertium/apertium-oci-fra/oci-fra.autosep1.bin'
^far<vblex><imp><p2><sg>$ ^pas<adv>$ ^mestièr<n><m><sg>$