Xapian discuss - Sep 2013 - having trouble with prefixes

I've got a small test database setup with one record.
$ delve -r 1 -V /tmp/1/
Values for record #1: 0:DD4F2162FFFF0E43741A4A1C2B8EC0E7
1:./Text_page_scan_2.jpg 2:jpg 3:.jpg
Term List for record #1: E:.jpg P:./Text_page_scan_2.jpg
Q:DD4F2162FFFF0E43741A4A1C2B8EC0E7 T:jpg

The terms were added with lines like this:
	doc.add_term(string("P:") + path);

Problem is, I can't seem to run a query that returns the document using
any of the terms. Here is the outline of the code that runs the queries
I'm trying to run:

	Database db(db_path.string());
	QueryParser queryparser;
	Stem stemmer("english");
	//queryparser.set_stemmer(stemmer);
	queryparser.set_database(db);
	queryparser.add_prefix("type", "T");
	queryparser.add_prefix("md5sum", "Q");
	queryparser.add_prefix("path", "P");
	queryparser.add_prefix("extension", "E");
	//maybe set stemming strategy here (in query parser)?
	queryparser.set_stemming_strategy(QueryParser::STEM_NONE);
	Query query(queryparser.parse_query(full_string));
	cout<<"Query is
'"<<full_string<<"'"<<endl;
	Enquire enquire(db);
	enquire.set_query(query);
	MSet match_set(enquire.get_mset(0, 10));
	for_each(match_set.begin(), match_set.end(),
	         [&db](docid id) {
		         print_doc_info(db.get_document(id));
	         });

I expected the following query to work,
md5sum:DD4F2162FFFF0E43741A4A1C2B8EC0E7
but it returns nothing. Same for all the other terms and prefixes. Terms
without prefixes seem to be working normally. I set stemming to NONE on
everything.

All I want is a way to ask xapian to return a list of all documents with
specific paths and/or md5sums.

thanks for any tips,
Chris

On Sun, Sep 01, 2013 at 10:37:59PM -0400, Christopher Harvey
wrote:
> I've got a small test database setup with one record.
> $ delve -r 1 -V /tmp/1/
> Values for record #1: 0:DD4F2162FFFF0E43741A4A1C2B8EC0E7
1:./Text_page_scan_2.jpg 2:jpg 3:.jpg
> Term List for record #1: E:.jpg P:./Text_page_scan_2.jpg
Q:DD4F2162FFFF0E43741A4A1C2B8EC0E7 T:jpg
> 
> The terms were added with lines like this:
> 	doc.add_term(string("P:") + path);
Just add the prefix "P" here.
> Problem is, I can't seem to run a query that returns the document using
> any of the terms. Here is the outline of the code that runs the queries
> I'm trying to run:
> 
> 	Database db(db_path.string());
> 	QueryParser queryparser;
> 	Stem stemmer("english");
> 	//queryparser.set_stemmer(stemmer);
> 	queryparser.set_database(db);
> 	queryparser.add_prefix("type", "T");
> 	queryparser.add_prefix("md5sum", "Q");
> 	queryparser.add_prefix("path", "P");
Or if you really want that colon in there, add the prefix as "P:"
here.
> 	queryparser.add_prefix("extension", "E");
> 	//maybe set stemming strategy here (in query parser)?
> 	queryparser.set_stemming_strategy(QueryParser::STEM_NONE);
> 	Query query(queryparser.parse_query(full_string));
> 	cout<<"Query is
'"<<full_string<<"'"<<endl;
If you print out query.get_description() it should be clearer what's
going on.

Cheers,
    Olly

Olly Betts <olly at survex.com> writes:
> On Sun, Sep 01, 2013 at 10:37:59PM -0400, Christopher Harvey wrote:
>> I've got a small test database setup with one record.
>> $ delve -r 1 -V /tmp/1/
>> Values for record #1: 0:DD4F2162FFFF0E43741A4A1C2B8EC0E7
1:./Text_page_scan_2.jpg 2:jpg 3:.jpg
>> Term List for record #1: E:.jpg P:./Text_page_scan_2.jpg
Q:DD4F2162FFFF0E43741A4A1C2B8EC0E7 T:jpg
>> 
>> The terms were added with lines like this:
>> 	doc.add_term(string("P:") + path);
>
> Just add the prefix "P" here.
>
>> Problem is, I can't seem to run a query that returns the document
using
>> any of the terms. Here is the outline of the code that runs the queries
>> I'm trying to run:
>> 
>> 	Database db(db_path.string());
>> 	QueryParser queryparser;
>> 	Stem stemmer("english");
>> 	//queryparser.set_stemmer(stemmer);
>> 	queryparser.set_database(db);
>> 	queryparser.add_prefix("type", "T");
>> 	queryparser.add_prefix("md5sum", "Q");
>> 	queryparser.add_prefix("path", "P");
>
> Or if you really want that colon in there, add the prefix as "P:"
here.
works! Thanks!

well, more precisely it works for everything except path names. I was
reading the docs here:
http://xapian.org/docs/queryparser.html

and saw the paragraph on punctuation:
---
A phrase surrounded with double quotes ("") matches documents
containing
that exact phrase. Hyphenated words are also treated as phrases, as are
cases such as filenames and email addresses (e.g. /etc/passwd or
president at whitehouse.gov).
---

using almost exactly the same code as last time, I fed it the following
query:
path:"foo.bar"
printing query.get_description() produces the following:
Query is 'Xapian::Query((P:foo:(pos=1) PHRASE 2 P:bar:(pos=2)))'
I was expecting a query with one term, since the "foo.bar" was
quoted. Consequently I can't match terms stored in the database that
have "." characters.

thanks again!
Chris