full-text indexing, locales, triggers, SPI & more fun

Started by Charlie Hornbergerover 25 years ago2 messages
#1Charlie Hornberger
charlie@pressflex.com

I've been doing some poking at the full-text indexing code in
/contrib/fulltextindex to try to get it to work with non-ASCII locales
(among other things), but I'm having a bit of trouble trying to figure
out how to properly parse non-ASCII strings from inside the fti()
trigger function (which is written in C).

My problem is this:

I want to aggregate text in multiple languages in a single full-text index
much like the current structure used by the current fti() function. In order
to correctly parse the strings, however, I've got to know what locale
they're written in/for (otherwise, isalpha() thinks that characters such as
the Hungarian letter u" -- that's a 'u' with a double acute accent -- aren't
very alphabetic.)

My initial thinking (which could certainly be very wrong) is that the
easiest way to get around this would be to allow client apps to set their
LC_ALL environment variables, and then to have the new fti() function use
that locale while doing string manipulation.

But the way I'm doing things, it doesn't appear that the LC_ALL environment
variable is available. (Maybe it was never meant to be ... but I'm not a
very skilled C programmer, and I don't know the first thing about the SPI
interface, so please forgive me if I'm asking why the sun doesn't rise in
the west more often ;-)).

Here's what's happening:

bash# LC_ALL=hu_HU
bash# export LC_ALL
bash# psql test
Welcome to psql, the PostgreSQL interactive terminal.

Type: \copyright for distribution terms
\h for help with SQL commands
\? for help on internal slash commands
\g or terminate with semicolon to execute query
\q to quit

test=# INSERT INTO ttxt (t1) values ('FELEL�SS�G�');
INSERT 513377 1
test=#select * from ttxt_fti;
string | id
--------+--------
felel | 513377
ss | 513377
(2 rows)

Which isn't quite what I'm looking for ;-).

Inside the C source of fti(), I added a call to getenv("LC_ALL") to make
sure that LC_ALL really isn't set:

locale = getenv("LC_ALL");
elog(NOTICE,"Locale is '%s'\n",locale);

And sure enough, it outputs:

NOTICE: Locale is '(null)'

If, on the other hand, I do:

setlocale("LC_ALL","hu_HU")

inside fti(), everything works out perfectly:

test=# INSERT INTO ttxt (t1) values ('FELEL�SS�G�');
INSERT 513410 1
test=# select * from ttxt_fti;
string | id
-------------+--------
felel�ss�g� | 513410
(1 row)

Any ideas?

Cheers,
Charlie

P.S. I only subscribe to the hackers digest, so please CC me with your
replies... Thanks!

#2Karel Zak
zakkr@zf.jcu.cz
In reply to: Charlie Hornberger (#1)
Re: full-text indexing, locales, triggers, SPI & more fun

On Wed, 31 May 2000, Charlie Hornberger wrote:

I want to aggregate text in multiple languages in a single full-text index
much like the current structure used by the current fti() function. In order
to correctly parse the strings, however, I've got to know what locale
they're written in/for (otherwise, isalpha() thinks that characters such as
the Hungarian letter u" -- that's a 'u' with a double acute accent -- aren't
very alphabetic.)

My initial thinking (which could certainly be very wrong) is that the
easiest way to get around this would be to allow client apps to set their
LC_ALL environment variables, and then to have the new fti() function use
that locale while doing string manipulation.

But the way I'm doing things, it doesn't appear that the LC_ALL environment
variable is available. (Maybe it was never meant to be ... but I'm not a
very skilled C programmer, and I don't know the first thing about the SPI
interface, so please forgive me if I'm asking why the sun doesn't rise in
the west more often ;-)).

The PostgreSQL set in main() next locale catg. (if you compile it with locale
support)

#ifdef USE_LOCALE
setlocale(LC_CTYPE, "");
setlocale(LC_COLLATE, "");
setlocale(LC_MONETARY, "");
#endif

If you need in your routines ctype.h's functions a solution is a set LANG
env. :
# LANG=Czech (for me)
# start_postmaster

It works very well, and you not need any other setting.

IMHO use setlocale(LC_ALL, ..) is a very strong hard to backend, because
a example all float data will crashed.

If you (still:-) need all locales see pg_locale.c in pg's sources in
utils/atd and usage of these routines in formatting.c (to_char()) which use
full locale for numbers.

But don't remember - you must always return all to state before LC_ALL.

Karel