On Sun, Feb 19, 2012 at 05:23:42PM +0100, Ernest Adrogué wrote: > Hello, > > It seems it does not accept l + <middle dot> + l either. > Or at least hunspell does not: > > $ echo 'intel·ligent'| hunspell -d ca_ES -m > intel > > ligent > > It splits it into two words. I added the following line in the affix > file: > > WORDCHARS · > > and apparently now it works (the line above must go before the first > TRY): > > $ echo 'intel·ligent'| hunspell -d ca_ES -m > intel·ligent st:intel·ligent
Hi Ernest, thanks for the info myspell-ca puts middle dot in the TRY section. Not sure if this is supposed to imply that all chars above are declared as possible parts of a word, or for '·', ' (and in general for anything but '-', but better declare this too) we need to declare it explicitly at WORDCHARS. Anyway, we can easily add it to the aff file, although it will only work for standalone hunspell. Did you contact upstream (Joan Moratinos) about this or tried some of softcatala resources (forums, ...)? > There are still problems though, because both Abiword and Firefox > still think it's two words. Communication here is done in a word basis, if Abiword or Firefox wrongly split word, "intel" and "ligent" will be asked as two different words and dealt with accordingly. Seems that this has been reported at least to Mozilla, #257073 - Spelling checker mark as misspelled words with a special character in Catalan https://bugzilla.mozilla.org/show_bug.cgi?id=257073 Regarding Abiword, problem seems similar. While it uses a compatibility layer (enchant+libenchant), the enchant test program (enchant itself) seems to wrongly tokenize word when sending it for spellchecking, $ $ echo 'intel·ligent' | enchant -l -d ca_ES intel ligent This happens when I use as default spellchecker any of ispell, aspell or myspell (hunspell), so I think enchant (and may be Abiword too) are to blame here first. I admit I do not know if there is something we can do about this problem in Abiword. Suggestions are welcome. Regarding Openoffice, I have seen references that new dict works with it, http://www.cau.cat/blog/versi_de_proves_220_del_diccionari_ortogr_fic_lliure_per_a_la_llengua_catalana I looked at a recent version of the dictionary (newer than that mentioned there) and does not contain explicit WORDCHARS, so I am confused, unless when calling Ooo.org passed extra wordchars info when calling libhunspell. Regarding the original bug report, since that is upstream choice, I was considering to close the bug report. I do not think it worths adding a new UTF8-only catalan dict just for 'l·l' in the two chars form. What do you think, Jordi? -- Agustin -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
