On Tue, September 19, 2006 13:27, Agustin Martin wrote: > On Wed, Sep 13, 2006 at 02:52:26PM +0300, Martin-Éric Racine wrote: >> >> On Fri, September 8, 2006 18:13, Stanislav Maslovski said: > >> > The word list /usr/share/aspell/ru.cwl.gz contains "138420" as the >> very >> > first word, which is a mistake. Here is the complete warning message I >> got >> > from "dpkg-reconfigure aspell-ru" (in russian) >> > >> > ------------------------------------------------------------------------- >> > aspell-autobuildhash: processing: ru [ru] >> > ðÒÅÄÕÐÒÅÖÄÅÎÉÅ: The word "138420" is invalid. óÉÍ×ÏÌ '1' (U+31) ÎÅ >> ÍÏÖÅÔ >> > ÐÏÑ×ÌÑÔØÓÑ × ÎÁÞÁÌÅ ÓÌÏ×Á. ðÒÏÐÕÓËÁÅÔÓÑ ÓÌÏ×Ï. >> > ------------------------------------------------------------------------- >> >> That is perfectly inoffensive. The aspell dictionary is generated from >> the >> myspell dictionary, which starts with a wordcount number. Aspell >> discards >> it when it builds the hash, as you saw, but it is needlessly verbose >> about >> it. > > You can use 'sed 1d' to strip first line and make this less verbose. This > is > what I do for the esperanto dictionary (with eo changed to ru): > > cat ru.dic | sed 1d | LC_COLLATE=C sort -u | prezip > ru.cwl
Except that, unless I'm mistaken, using the C collation rules for languages other than English languages is wrong. -- Martin-Éric Racine http://q-funk.iki.fi
