Searching multiple terms on multiple fields with different boost using Index.query

[myalgo]

I would like to know to write the search query efficiently for searching multiple terms on multiple fields with different boost

I have set of fields

• title with boost 10
• heading with boost 5
• text with boost 3
• keywords with boost 6

and would like to search "foo bar" or "foo bar*"

Since build time boost is not supported in 2.1.2, I am using query time boost. However would like to use to search with and without trailing wildcard and with/without usepipeline to get the best results.

Please let me know an efficient way to write query using Index.query or Index.search.

[taysloane]

You could use something like this (tested in 2.1.0):

const query = 'foo bar';
const idx = ...;
const matches =  idx.query(q => {
    q.term(query, { fields: [ 'title' ], boost: 10 });
    q.term(query, { fields: [ 'heading' ], boost: 5 });
    q.term(query, { fields: [ 'text' ], boost: 3 });
    q.term(query, { fields: [ 'keywords' ], boost: 6 });

    q.term(query, { fields: [ 'title' ], boost: 10, wildcard: lunr.Query.wildcard.TRAILING });
    q.term(query, { fields: [ 'heading' ], boost: 5, wildcard: lunr.Query.wildcard.TRAILING });
    q.term(query, { fields: [ 'text' ], boost: 3, wildcard: lunr.Query.wildcard.TRAILING });
    q.term(query, { fields: [ 'keywords' ], boost: 6, wildcard: lunr.Query.wildcard.TRAILING });
  });
});

You can add usePipeline: false, editDistance: ... etc. as normal inside there. We're using something similar in production.

[olivernn]

What @jakeapeters has looks good, one thing to be careful of though is that the query string passed to q.term will not be split into multiple tokens by lunr. I.e. if query is the string "foo bar" it will search for "foo bar" not "foo", "bar", with the default index builder this will mean you won't get any results.

You can borrow Lunr's tokenizer, or use the regex it uses directly:

idx.query(function (q) {
  lunr.tokenizer(queryString).forEach(function (token) {
    q.term(token.toString(), { fields: [ 'title' ], boost: 10 })
    // etc....
  })
})

or, just borrow the regex:

idx.query(function (q) {

  queryString.split(lunr.tokenizer.separator).forEach(function (term) {
    q.term(term, { fields: [ 'title' ], boost: 10 })
    // etc....
  })
})

[myalgo]

thanks @jakeapeters and @olivernn.

[myalgo]

@olivernn , do you plan to support build time boost?

[olivernn]

@myalgo please open a new issue requesting build time boost.

I'm going to close this issue now as I think the original problem is solved.