Re: LIKE 'bla%'

Tom Lane <tgl@sss.pgh.pa.us>

From: Tom Lane <tgl@sss.pgh.pa.us>
To: steffen@city-map.de, pgsql-bugs@postgresql.org
Date: 2000-09-02T17:18:14Z
Lists: pgsql-bugs
pgsql-bugs@postgresql.org writes:
> DELETE FROM MYTABLE WHERE NAME LIKE 'Ant%';
> knowing that there existed only one record with 'Anton' as name.
> However, the code above delivered 'DELETE 0'

> DELETE FROM MYTABLE WHERE NAME LIKE 'Anto%';
> which gave 'DELETE 1'

> Isn't this strange?

Yup.  What PG version are you using, and are you running it with a
non-English LOCALE setting?  Is there an index on the NAME column?

I suspect you are running into another variant of the problem we've
had for a long time concerning how to derive upper and lower index
boundes for a LIKE string.  In ASCII locale it's pretty easy:
	name >= 'Ant' AND name < 'Anu'
can be used to scan the index for all entries that might match the
given LIKE pattern.  But in non-ASCII locales with complicated collation
rules that method tends to fail.  See the pgsql-hackers mailing lists;
latest go-round was thread
	Sigh, LIKE indexing is *still* broken in foreign locales
in early June 2000.  At the moment I don't think we know a bulletproof
solution, other than not using indexes for LIKE, which won't make people
happy either ...

			regards, tom lane