Greets,
I expect to commit a few changes to the public API in a bit, and wanted to
document the rationales and open an opportunity for discussion.
First, there are four methods and one constructor which should have been
officially exposed a while ago; we'll correct the oversight and make them
public.
TermQuery_Get_Term()
TermQuery_Get_Field()
Doc_Set_Doc_ID()
Doc_Get_Doc_ID()
Arch_new()
Second, Searcher_Fetch_Doc() will be changed to return a HitDoc rather than an
Obj. I'm also going to remove the "score" and "offset" parameters, leaving
only "doc_id". Since this brings the argument count down to one, the Perl
bindings will now use a single positional arg instead of labeled params.
# Before:
my $doc = $searcher->fetch_doc( doc_id => $doc_id );
# After:
my $doc = $searcher->fetch_doc($doc_id);
Similarly, DocReader_Fetch(), which takes three params and returns an Obj, will
be replaced by DocReader_Fetch_Doc(), which takes only a doc_id and returns a
HitDoc.
These Fetch_Doc() changes close the book on an unsuccessful experiment:
allowing the Fetch_Doc() stack to return arbitrary objects. It was this
ill-advised flexibility which necessitated the addition of parameters to
Fetch_Doc(); we'll be returning to the Lucene model for fetching documents,
which is simpler for doing simple things, yet still makes hard things possible:
# Before ($doc can be an arbitrary object):
my $doc = $searcher->fetch_doc(
doc_id => $doc_id,
offset => $offset,
score => $score,
);
# After ($doc isa HitDoc):
my $doc = $searcher->fetch_doc($doc_id - $offset);
$doc->set_doc_id($doc_id);
$doc->set_score($score);
Lastly, I'm going to redact HeatMap (used by Highlighter) as a public class.
It's not complete or polished enough to do what it sets out to do right now.
We need to think its interface through a little more.
Marvin Humphrey
