Re: pkg/38040 (textproc/mecab-{ipadic,jumandic} would have to provide the way to select charset)

[Naoto Morishima <naoto%morishima.net@localhost>]

Hummm, sorry for my ignorance abount MeCab API to select dictionaries.

>>>>> On Sun, 17 Feb 2008 04:50:02 +0000 (UTC), "OBATA Akio" 
>>>>> <obache%netbsd.org@localhost> said:

obache> Namazu is only package possibly using MeCab in pkgsrc and it expect
obache> that charset of the dictionary is euc-jp.

 There should be potential users who uses MeCab standalone,
 i.e. without Namazu.  The pkgsrc should not enforce them to use
 euc-jp dictionaries.

obache> MeCab already can select dictionary:

 I believe that the pkgsrc system does not have any mechanisms for a
 package to ensure some options has been set in a package it depends.
 For instance, package 'namazu' cannot ensure 'euc-jp' option has been
 set in package 'mecab-ipadic' or something, so we cannot deal with
 'euc-jp' in mecab-ipadic as an option, right?

 So there should be several solutions for it;

  - allow mecab-ipadic (or jumandic) to install multiple dictionaries
    at the same time by introducing some options.  In this case, these
    package should guarantee to install the dictionary in euc-jp; that
    is, euc-jp should not be included to the options.

     pros: no need to create multiple (and possibly fiddly) packages.
     cons: the dictionary in euc-jp is always installed regardless of
           its necessity.

  - split mecab-ipadic (or jumandic) to multiple packages according to
    its charset.  Namazu should depend to mecab-ipadic-euc-jp.

     pros: users and packages can choice dictionaries to install.
     cons: need to generate and maintain similar packages.

 Anyway, namazu should provide configuration file which considers
 which dictionary should be used.

 Any other idea?

 BTW, mecab-base in the HEAD of pkgsrc cannot be compiled in
 3.1_STABLE and Solaris w/ sunpro.  Should I send pr for them?

Regards,
Naoto Morishima