Re: Issue with libhyphen

Dear Bert,

I am glad of it. Yes, it is not a bug. Thanks for the feedback.

Best regards,

László

2014-03-24 15:57 GMT+01:00 Bert Frees <bertfrees@gmail.com>:
> Dear László,
>
> Thanks for your answer. You are right, substrings.pl turns out to be exactly
> what I need.
>
> So I guess this is not a bug. I missed the section about the substrings
> script in the README, although it is made obvious enough with warnings etc.
> :)
>
> Cheers,
> Bert
>
>
> 2014-03-24 15:19 GMT+01:00 Németh László <nemeth@numbertext.org>:
>
> > Dear Bert,
> >
> > It seems for me, that your hyphenation patterns haven't been processed
> > by the substrings.pl script of libhyphen distribution. Could you check
> > it?
> >
> > Thanks for your bug report,
> >
> > Best regards,
> > László
> >
> > 2014-03-24 14:44 GMT+01:00 Bert Frees <bertfrees@gmail.com>:
> > > Dear László and others,
> > >
> > > We think we may have found a bug in libhyphen. It could be that it is
> > > just a
> > > limitation of the algorithm, but anyway it's an issue for us.
> > >
> > > The problem is that some patterns in a dictionary are ignored in some
> > > cases,
> > > namely when the match string of that pattern is a part of the match
> > > string
> > > of another pattern, and more specifically when it's not just a prefix.
> > >
> > > Let me clarify that with an example. When a dictionary consists of these
> > > two
> > > patterns, the word `foobar' is not hyphenated because the first pattern
> > > is
> > > ignored:
> > >
> > > oo1b
> > > foob
> > >
> > > The second, longer pattern doesn't even have to match, as the second
> > > example
> > > shows (the first pattern is still ignored):
> > >
> > > oo1b
> > > foobz
> > >
> > > I have a patch that solves part of the problem:
> > >
> > >
> > > https://github.com/bertfrees/libhyphen-nar/blob/adc2b74a19469e4dc93777fcdb82e36e566a0472/src/patches/bug.patch
> > >
> > > With this patch the given examples will be handled correctly, but in
> > > other
> > > situations it will still fail, such as here:
> > >
> > > oo1b
> > > foobaz
> > >
> > > Have I indeed found a bug here, and does my patch make sense, or am I
> > > just
> > > expecting too much and are we hitting the limits of the algorithm?
> > >
> > > Thanks for considering,
> > > Bert Frees