Re: Fulltext search configuration

Hi Mohamed.

I don't know where you get the dictionary - I unsuccessfully tried the
OpenOffice one by myself (the Ayaspell one), and I had no arabic
stopwords file.

Renaming the file is supposed to be enough (I did it successfully for
Thailandese dictionary) - the ".aff'" file becoming the ".affix" one.
When I tried to create the dictionary:

CREATE TEXT SEARCH DICTIONARY ar_ispell (
TEMPLATE = ispell,
DictFile = ar_utf8,
AffFile = ar_utf8,
StopWords = english
);

I had an error:

ERREUR: mauvais format de fichier affixe pour le drapeau
CONTEXTE : ligne 42 du fichier de configuration «
/usr/share/pgsql/tsearch_data/ar_utf8.affix » : « PFX Aa Y 40

(which means Bad format of Affix file for flag, line 42 of configuration
file)

Do you have an error when creating your dictionary?

Daniel

Mohamed a écrit :
> I have ran into some problems here.
>
> I am trying to implement arabic fulltext search on three columns.
>
> To create a dictionary I have a hunspell dictionary and and arabic
> stop file.
>
> CREATE TEXT SEARCH DICTIONARY hunspell_dic (
> TEMPLATE = ispell,
> DictFile = hunarabic,
> AffFile = hunarabic,
> StopWords = arabic
> );
>
>
> 1) The problem is that the hunspell contains a .dic and a .aff
> file but the configuration requeries a .dict and .affix file. I
> have tried to change the endings but with no success.
>
> 2) ts_lexize('hunspell_dic', 'ARABIC WORD') returns nothing
>
> 3) How can I convert my .dic and .aff to valid .dict and .affix ?
>
> 4) I have read that when using dictionaries, if a word is not
> recognized by any dictionary it will not be indexed. I find that
> troublesome. I would like everything but the stop words to be
> indexed. I guess this might be a step that I am not ready for yet,
> but just wanted to put it out there.
>
>
>
> Also I would like to know how the process of the fulltext search
> implementation looks like, from config to search.
>
> Create dictionary, then a text configuration, add dic to
> configuration, index columns with gin or gist ...
>
> How does a search look like? Does it match against the gin/gist index.
> Have that index been built up using the dictionary/configuration, or
> is the dictionary only used on search frases?
>
> / Moe
>
>
>
>