Subroutine argument apparently lost in loop

I have a CGI script pulling bibliography data from a 开发者_如何学编程BibTeX file, building HTML from it. It uses CGI::Ajax to call the subroutine below with one or two arguments. Most of the time, it will be a search term that is passed as $s, but if I pass a string through my HTML form, the subroutine will not be entirely happy with it. There is a foreach loop checking the entries and jumping over the entries that do not match. Now I can print the argument outside this loop alright, but the loop itself won’t print anything for $s, nor will it find any entries matching it. If within the loop $s were simply empty, the subroutine would print the entire bibliography, but this is not the case.

Basically it is as if $s passed as an argument breaks the loop, whereas an explicit definition in the subroutine works fine.

Here is a simplified version of my code. Please excuse sloppy or ignorant coding, I’m just dabbling in Perl.

sub outputBib {
    my ( $self,$s,$kw ) = @_;
    my @k;
    @k = ('foo','bar'); # this is fine
    @k = keys (%{$self->{_bib}}); # this causes problems
    foreach my $k (@k) {
        $output .= "Key = $k<br/>";
        $output .= "Search Term = $s<br/>";
    }
    return $output;
}

The problem seems to be the array built from the keys of the $self->{_bib} hash. It is odd that

1. the loop is fine when $s is not passed through CGI::Ajax. All elements are processed.
2. as soon as the subroutine is called with $s, the loop does not return anything.
3. if @k is defined as a straightforward array, the loop works and $s can be printed within the loop;

I build $self->{_bib} like so:

sub parseBib {
    my ( $self ) = @_;
    while (my $e = new Text::BibTeX::Entry $self->{_bibFileObject}) {
        next unless $e->parse_ok;
        my %entry_hash;
        $entry_hash{'title'} = $e->get('title');
        $entry_hash{'keywords'} = $e->get('keywords');
        $self->{_bib}{$e->key} = \%entry_hash;
    }
}

Any ideas? Thanks.

My first suggestion would be to use warn/print STDERR to verify on the live running copy that, when called via CGI::Ajax, all of your variables ($self, $s, $kw, $self->{_bib}) have the values that you're expecting. Although I'm a big fan of CGI::Ajax, it does a fair bit of magic behind your back and it may not be calling outputBib in quite the way you think it is.

Also keep in mind that CGI operates on a per-request model, not per-page. Are you perhaps populating $self->{_bib} when you send the initial page (and also doing all of your successful tests in that environment), then expecting it to still be there when the AJAX requests come in? If so, you're out of luck - you'll need to rebuild it in the AJAX handler, either within outputBib or earlier in your code, before you call ->build_html and hand it off to CGI::Ajax.