> dbahena@tpv.com.mx writes:
> > ventasge2000=# create table t1 (c1 char(10),c2 varchar(10));
> > CREATE
> > ventasge2000=# insert into t1 values('XXX666','XXX666');
> > INSERT 182218 1
> > ventasge2000=# select * from t1 where c1 ~ '666$';
> > c1 | c2
> > ----+----
> > (0 rows)
> > ventasge2000=# select * from t1 where c2 ~ '666$';
> > c1 | c2
> > ------------+--------
> > XXX666 | XXX666
> > (1 row)
>
> > Doesn't regular expressions(in particular the $ metachar)
> work properly
> > with char columns????
>
> I see no bug there --- you've forgotten about the trailing spaces in
> the char(10) column. Try c1 ~ '666 *$' if you want to match against a
> variable amount of padding in a char(N) column.
No, imho char(n) is defined to return trailing blanks but be insensitive to
the actual amount of trailing spaces. Thus I do see that this behavior can
be interpreted as a bug here.
In char(n) speak 'ab' = 'ab ' is supposed to be true.
Imho a change from char(6) to char(8) should only require more storage
space in a client program, but be otherwise transparent,
which it currently is not.
Imho a similar problem is that char_length does not return the count to
the last non space character, which imho also is a bug.
Andreas