looking for inspiration on 'compare'

[zioigor]

Bowing to higher level of knowledge, I am working with solid-rock DB_File module, but something cracked my faith: i tried to ovverride $DB_BTREE->{'compare'} routine with a String::Approx 'adist' one. Belief and RTFM suggested to create a 'callback' with an easy template, something like ... "hey man these are 2 parameters, do some kind of comparison and give me back -1, 0 or 1" as the 'cmp' example on documentation..." so, after checking out the only working example routine with case-insensitive comparison on manual, i tried mine:

# -------------------------
sub DBFile_Compare_Distance
#--------------------------
{
    my ($k1, $k2) = @_ ;
    
        my $dist = adist ( uc $k1 ,  uc $k2 );
    my $val = undef;
    if(     $dist < -1 )    { $val = -1; }
    elsif(    $dist > 1 )    { $val = 1; }
    else            { $val = 0; }
    $val;
}

- .db file gets created and i checked the binaries
- all data seem to fit
BUT
when i try to read out key/value pairs, i got "undef" on values for which comparison function returned -1.
Any clue on how to make a Compare function different from the one on manual?
many prayers are rising towards the Monastery
thanks igor

[fizbin]

when i try to read out key/value pairs, i got "undef" on values for which comparison function returned -1.

This statement doesn't make sense. A comparison function is passed two keys. Yet when you read out key/value pairs, you get one key at a time. You can't look at a key and say "Ah, for this key, the comparison function returned -1". You can look at a pair of keys and say "For this pair of keys, in this order, the comparison function returned -1." But as it stands, your statement makes no sense.

[PodMaster]

BUT
when i try to read out key/value pairs, i got "undef" on values for which comparison function returned -1.

That sounds fishy, are you sure you're not storing undef? A custom compare subroutine should not effect the values stored (and its most likely that its not).

Any clue on how to make a Compare function different from the one on manual?

You already have one, so just like that.

[zioigor]

Hi, yes i am logging every step before and after the ->put() infact I do agree with what you say. But every check i made goes in the direction u suggest. I do print with Dumper the tied hash, and those read return undef only for -1 comapre values. On binary file i checked and data are there. I mentioned the minus one just because is the only clue i have ... on non working values ...

[PodMaster]

...On binary file i checked and data are there. I mentioned the minus one just because is the only clue i have ... on non working values ...

So which do you think is more likely, that there is something seriously broken with a very old and thoroughly tested module, or that you're doing something boneheaded? You need to submit a self contained example which demonstrates your bug.

[jZed]

This is of no help in figuring out your current problem but may be of interest for the future. Were you aware that the DBI distribution now includes DBD::DBM? This module provides a DBI/SQL front-end to any DBM file including DB_File and BerkeleyDB either with or without MLDBM and using your choice of serializer. Naturally, SQL provides a number of comparison operators so, at a minimum, you could check your file (without converting it) with DBD::DBM, just to make sure it has in it what you expect.