[Xapian-devel] Document ID 0 is invalid... but not always...

[Daniel Ménard]

Note: this is rather long and not very important and I don't want to 
prevent the team from releasing version 1.0, so go on reading only if 
you have too much free time !!! ;-)

0 is not a valid document ID, never, ever, but I just found a special 
case in which xapian will create a record and return 0 for the newly 
created record.

In fact, I was "hacking", trying to store metadata in a "special" record 
with docId "-1"...
I know, this is bad, but what is interesting is what xapian does in such 
a situation...

Here is a simple PHP test :

    $db=new XapianWritableDatabase('pathtodb', 
Xapian::DB_CREATE_OR_OVERWRITE);
       
    $doc=new XapianDocument();
    $doc->set_data('metadata'); // waiting for 
http://www.xapian.org/cgi-bin/bugzilla/show_bug.cgi?id=143
    $doc->add_term('metadata');
    $db->replace_document(-1, $doc); //  or 4294967295 = (2^32)-1

    $doc=new XapianDocument();
    $doc->set_data('data');
    $doc->add_term('data');
    $docId=$db->add_document($doc); // get_lastdocid()+1 overflows, will 
return 0
    echo "doc #$docId added\n";

The two docs are added without error, but the second one will get the 
docId 0 and of course won't be accessible.

Another variant will cause xapian to overwrite an existing record :
(pseudo code) :
    new XapianWritableDatabase()
    add_document($doc); // will get id #1
    replace_document(-1, $doc);
    add_document($doc); // will get id #0
    add_document($doc); // will again get id #1, previous record is 
probably overwritten

As I said at the beginning, I don't think this is very important, and it 
does not cause me any inconvenience.
I just thought it would be better to let you know.

Thanks for reading, please go back releasing xapian 1.0 !

-- 

Daniel Ménard