Lucene-like searching through JSON objects in JavaScript

1. Go through your records, to create a one time index by combining all searchable fields in a single string field called index.

2. Store these indexed records in an Array.

3. Partition the Array on index .. like all a's in one array and so on.

4. Use the javascript function indexOf() against the index to match the query entered by the user and find records from the partitioned Array.

That was the easy part but, it will support all simple queries in a very efficient manner because the index does not have to be re-created for every query and indexOf operation is very efficient. I have used it for searching up to 2000 records. I used a pre-sorted Array. Actually, that's how Gmail and yahoo mail work. They store your contacts on browser in a pre-sorted array with an index that allows you to see the contact names as you type.

This also gives you a base to build on. Now you can write an advanced query parsing logic on top of it. For example, to support a few simple conditional keywords like - AND OR NOT, will take about 20-30 lines of custom JavaScript code. Or you can find a JS library that will do the parsing for you the way Lucene does.

For a reference implementation of above logic, take a look at how ZmContactList.js sorts and searches the contacts for autocomplete.

You might want to check FullProof, it does exactly that:https://github.com/reyesr/fullproof

Have you tried CouchDB?

Edit:

How about something along these lines (also see http://jsfiddle.net/7tV3A/1/):

var filtered_collection = [];var query = 'foo';$.each(collection, function(i,e){    $.each(e, function(ii, el){        if (el == query) {            filtered_collection.push(e);        }    });});

The (el == query) part of course could/should be modified to allow more flexible search patterns than exact match.