Request for community: please, give us feedback about Percolate Queries about manticoresearch

[barryhunter]

They seem to work in limited testing done.

What I dont understand is the point of adding fields/attributes to the percolate index. Its specifally mentioned can reconfigure the index. But not sure why, or how to use it added fields

I initially thought it was so could additionally filter the queries in SELECT (kinda like the buiult in tags column) but that doesnt seem to be the case.

The docs shows adding a json atgtribute. But HOW would you use it? Can use the extra fields/attributes in INSERT, in the SELECT, or in the CALL PQ? (dont know)

... just offering this as a part that dont think explained very well in the docs.

[barryhunter]

btw, the blog post does appear to include an example. where the fields/attributes on the index, seem to be what can use in @field syntax and Filters on queries.

So is it that need to explicitly add fields/attributes to the index, BEFORE can INSERT a query into the index, AND/OR before can run a CALL PQ query (including the columns in the JSON 'document') ?

[tomatolog]

You might use attributes to filter out some documents and fields to constrain query terms.

On INSERT query got transformed to node tree and query field operators out of schema got dropped. However on index reconfigure all queries again got transformed to node tree from initial text representation.

Same for attributes. On issue matching with CALL PQ attributes our of schema got dropped and attributes these not mention at CALL PQ got their default values. Then on matching for query with filters document(s) got rejected in case it failed to pass query filters.

[tomatolog]

the syntax is same as SphinxQL for @field and filters

INSERT INTO pq_filter (query, filters) VALUES ( 'filter test', 'gid >= 10' );
INSERT INTO pq_filter (query, filters) VALUES ( 'filter test', 'gid >= 10 OR gid <= 3' );
CALL PQ ('pq_filter', ('{"title":"filter test doc1", "gid":2 }', '{"title":"filter test doc2", "gid":13}'), 1 as docs, 1 as docs_json );

[barryhunter]

So the 'columns' used in CALL PQ documents (title and gid - in your sample) - need to be defined as either field or attributes in the index itself. Thats the bit that not clear to me at least.

The text field showing in a DESCRIBE pqindex is just a single field, so that 0 AS docs_json, a single string for the document is assigned a field name.

[tomatolog]

I took example from test_321 there index declaration is

index pq_filter
{
	type = percolate
	path = data/pq_filter
	rt_field = title
	rt_attr_uint = gid
}

without docs_json option you can not map documents to different fields as stated into documentation
Fields and attributes mapping are allowed for JSON documents only.
without docs_json parser assigns fields from document with index schema defined order.

We implementing HTTP interface for PQ it will be json based and no need to encode \ decode json objects into string as you have to with SphinxQL

[airolg]

Note about Percolate Query preformance Manticoresearch vs. Elasticsearch
http://bit.ly/2nLSpqJ

[airolg]

Done, thank you