Skip to content

Commit

Permalink
WIP on facet counts
Browse files Browse the repository at this point in the history
  • Loading branch information
@rbuels
rbuels committed May 17, 2012
1 parent 2a25702 commit 92fa1e8
Show file tree
Hide file tree
Showing 3 changed files with 260 additions and 85 deletions.
27 changes: 27 additions & 0 deletions faceted_track_selector.css
View file
Original file line number Diff line number Diff line change
Expand Up @@ -149,14 +149,41 @@

/* Facet selection controls */

#faceted_tracksel .facetSelect {
width: 100%;
border-spacing: 0;
}
#faceted_tracksel .facetSelect .facetValue {
padding: 0.1em 0.4em;
cursor: pointer;
}
#faceted_tracksel .facetSelect .facetValue > * {
vertical-align: top;
}
#faceted_tracksel .facetSelect .disabled {
color: gray;
}

#faceted_tracksel .facetSelect .facetValue.disabled {
display: none;
}
#faceted_tracksel .facetSelect .facetValue.disabled.selected {
display: block;
}

.tundra #faceted_tracksel .facetSelect .facetValue:hover {
background: #D2E1F1;
}

#faceted_tracksel .facetSelect .facetValue .count {
padding: 0 0.7em 0 0.4em;
color: #333;
text-align: right;
}
#faceted_tracksel .facetSelect .facetValue .value {
width: 80%;
}

#faceted_tracksel .facetSelect .selected {
background: #b1d3f6;
}
Expand Down
266 changes: 187 additions & 79 deletions js/Model/TrackMetaData.js
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,9 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
* @lends JBrowse.Model.TrackMetaData.prototype
*/
{

_noDataValue: '(no data)',

/**
* Data store for track metadata, supporting faceted
* (parameterized) searching. Keeps all of the track metadata,
Expand Down Expand Up @@ -83,11 +86,41 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
* @private
*/
_finishLoad: function() {
this.facets = this.facets.sort();

// sort the facet names
this.facets.sort();

// calculate the average bucket size for each facet index
dojo.forEach( dojof.values( this.facetIndexes.byName ), function(bucket) {
bucket.avgBucketSize = bucket.itemCount / bucket.bucketCount;
});
// calculate the rank of the facets: make an array of
// facet names sorted by bucket size, descending
this.facetIndexes.facetRank = dojo.clone(this.facets).sort(dojo.hitch(this,function(a,b){
return this.facetIndexes.byName[a].avgBucketSize - this.facetIndexes.byName[b].avgBucketSize;
}));

// sort the facet indexes by ident, so that we can do our
// kind-of-efficient N-way merging when querying
var itemSortFunction = dojo.hitch( this, '_itemSortFunc' );
dojo.forEach( dojof.values( this.facetIndexes.byName ), function( facetIndex ) {
dojo.forEach( dojof.keys( facetIndex.byValue ), function( value ) {
facetIndex.byValue[value].items = facetIndex.byValue[value].items.sort( itemSortFunction );
});
},this);

this.ready = true;
this.onReady();
},

_itemSortFunc: function(a,b) {
var ai = this.getIdentity(a),
bi = this.getIdentity(b);
return ai == bi ? 0 :
ai > bi ? 1 :
ai < bi ? -1 : 0;
},

_indexItems: function( args ) {
// get our (filtered) list of facets we will index for
var store = args.store,
Expand Down Expand Up @@ -171,25 +204,15 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
}, this);

// index the items that do not have data for this facet
var noDataValue = '(no data)';
dojo.forEach( new_facets, function(facet) {
var gotSomeWithNoData = false;
dojo.forEach( dojof.values( this.identIndex ), function(item) {
if( ! gotDataForItem[facet][this.getIdentity(item)] ) {
gotSomeWithNoData = true;
this._indexItem( facet, noDataValue, item );
this._indexItem( facet, this._noDataValue, item );
}
},this);
},this);

// calculate the rank of the facets: make an array of
// facet names sorted by smallest average bucket size,
// descending
this.facetIndexes.facetRank = this.facets.sort(dojo.hitch(this,function(a,b){
a = this.facetIndexes.byName[a];
b = this.facetIndexes.byName[b];
return b.itemCount/b.bucketCount - a.itemCount/a.bucketCount;
}));
}
},

Expand Down Expand Up @@ -239,6 +262,16 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
return this._fetchCount;
},


/**
* @param facetName {String} facet name
* @returns {Object}
*/
getFacetCounts: function( facetName ) {
var context = this._fetchFacetCounts[ facetName ] || this._fetchFacetCounts[ '__other__' ];
return context[facetName];
},

/**
* Get an array of the text names of the facets that are defined
* in this track metadata.
Expand Down Expand Up @@ -270,8 +303,9 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
if( !index ) return {};

var stats = {};
dojo.forEach( ['itemCount','bucketCount'], function(attr) { stats[attr] = index[attr];});
stats.avgBucketSize = stats.itemCount / stats.bucketCount;
dojo.forEach( ['itemCount','bucketCount','avgBucketSize'],
function(attr) { stats[attr] = index[attr]; }
);
return stats;
},

Expand Down Expand Up @@ -324,68 +358,10 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
}
},this);

var textFilter = query.text;
var textFilter = this._compileTextFilter( query.text );
delete query.text;

var results;

// if we don't actually have any facets specified in the
// query, the results are just all the items
if( ! dojo.some( dojof.values(query), function(q){ return q.length > 0;}) ) {
results = dojof.values( this.identIndex );
}
else {
// start with an initial set of items that have the desired
// value for the most specific (smallest avg index bucket
// size) facet that was specified
results = (function() {
var highestRankedFacet,
queryValues = [],
index;
dojo.some( this.facetIndexes.facetRank, function(facetName) {
if( query[facetName] ) {
highestRankedFacet = facetName;
queryValues = query[highestRankedFacet];
index = this.facetIndexes.byName[highestRankedFacet];
return queryValues.length > 0;
}
return false;
},this);

delete query[highestRankedFacet];

var set = [];
dojo.forEach( queryValues, function(val) {
var bucket = index.byValue[val];
if( bucket )
set.push.apply( set, bucket.items );
},this);
return set;
}).call(this);

// and filter this starting set for the other facets
dojo.forEach( dojof.keys(query), function(facetName) {
var desired_values = query[facetName] || [];
if( ! desired_values.length )
return;
results = dojo.filter(results, function(item) {
var value = this.getValue(item,facetName);
return dojo.some( desired_values, function(desired) {
return desired == value;
},this);
},this);
},this);
}

// filter with the text filter, if we have it
if( typeof textFilter != 'undefined' ) {
var filter = this._compileTextFilter( textFilter );
results = dojo.filter( results, function(item) {
return dojo.some( this.facets, function(facetName) {
return filter( this.getValue( item, facetName ) );
},this);
},this);
}
var results = this._doQuery( query, textFilter );

this._fetchCount = results.length;

Expand All @@ -394,6 +370,133 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
this.onFetchSuccess();
},

/**
* @private
*/
_doQuery: function( /**Object*/ facetQuery, /**Function*/ textFilter ) {
// algorithm pseudocode:
//
// * for each individual facet, get a set of tracks that
// matches its selected values. sort each set by the
// tracks' unique identifier.
// * while still need to go through all the items in the filtered sets:
// - if all the facets have the same track first in their sorted set:
// add it to the core result set.
// count it in the global counts
// - if all the facets *but one* have the same track first:
// this track will need to be counted in the
// 'leave-out' counts for the odd facet out. count it.
// - shift the lowest-labeled track off of whatever facets have it at the front


var results = []; // array of items that completely match the query

// construct the filtered sets (arrays of items) for each of
// our search criteria
var filteredSets = [];
if( textFilter ) {
filteredSets.push( dojo.filter( dojof.values( this.identIndex ), textFilter ) );
filteredSets[0].facetName = 'Contains text';
filteredSets[0].myOffset = 0;
}
filteredSets.push.apply( filteredSets,
dojo.map( dojof.keys( facetQuery ), function( facetName ) {
var values = facetQuery[facetName];
var items = [];
dojo.forEach( values, function(value) {
items.push.apply( items, this.facetIndexes.byName[facetName].byValue[value].items );
},this);
items.myOffset = 0;
items.facetName = facetName;
items.sort( dojo.hitch( this, '_itemSortFunc' ));
return items;
},this)
);

// init counts
var facetMatchCounts = {};
var countItem = function( item, facetName ) {
var facetEntry = facetMatchCounts[facetName];
if( !facetEntry ) facetEntry = facetMatchCounts[facetName] = {};
dojo.forEach( this.facets, function(attrName) {
var value = this.getValue( item, attrName, this._noDataValue );
var attrEntry = facetEntry[attrName];
if( !attrEntry ) {
attrEntry = facetEntry[attrName] = {};
attrEntry[value] = 0;
}
attrEntry[value] = ( attrEntry[value] || 0 ) + 1;
},this);
};

if( ! filteredSets.length ) {
results = dojof.values( this.identIndex );
} else {
// calculate how many item records total we need to go through
var leftToProcess = 0;
dojo.forEach( filteredSets,
function(s) { leftToProcess += s.length;} );

// do a sort of N-way merge of the filtered sets
while( leftToProcess ) {

// look at the top of each of our sets, seeing what items
// we have there. group the sets by the identity of their
// topmost item.
var setsByTopIdent = {}, uniqueIdents = [], ident, item;
dojo.forEach(filteredSets, function(set,i) {
item = set[ set.myOffset ];
ident = item ? this.getIdentity( item ) : '(at end of set)';
if( setsByTopIdent[ ident ] ) {
setsByTopIdent[ ident ].push( set );
} else {
setsByTopIdent[ ident ] = [set];
uniqueIdents.push( ident );
}
},this);
if( uniqueIdents.length == 1 ) {
// each of our matched sets has the same item at the
// top. this means it is part of the core result set.
results.push( item );
} else {
uniqueIdents.sort();
ident = uniqueIdents[0] == '(at end of set)' ? uniqueIdents[1] : uniqueIdents[0];
if( uniqueIdents.length == 2
&& setsByTopIdent[ ident ].length == 1 ) {
// all of the matched sets except one has the same
// item on top, and it is the lowest-labeled item

var leftOutSet = setsByTopIdent[ ident ][0];
countItem.call( this, leftOutSet[ leftOutSet.myOffset ], leftOutSet.facetName );
}
}
dojo.forEach( setsByTopIdent[ ident ], function(s) { s.myOffset++; leftToProcess--; });
}
}

// dojo.forEach( dojof.keys(facetMatchCounts), function(category) {
// dojo.forEach( results, function(item) {
// countItem.call(this, item, category);
// },this);
// },this);
dojo.forEach( results, function(item) {
countItem.call(this, item, '__other__' );
},this);
this._fetchFacetCounts = facetMatchCounts;
console.log( facetMatchCounts );
// in the case of just one filtered set, the 'leave-one-out'
// count for it is actually the count of all results, so we
// need to make a special little count of that attribute for
// the global result set.
// if( filteredSets.length == 1 ) {
// dojo.forEach( dojof.values( this.identIndex ), function(item) {
// countItem.call( this, item, filteredSets[0].facetName, filteredSets[0].facetName );
// },this);
// }

return results;
},

/**
* Event hook called once when the store is initialized and has
* an initial set of data loaded.
Expand All @@ -412,6 +515,8 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
* @private
*/
_compileTextFilter: function( textString ) {
if( textString === undefined )
return null;

// parse out words and quoted words, and convert each into a regexp
var rQuotedWord = /\s*["']([^"']+)["']\s*/g;
Expand All @@ -437,11 +542,14 @@ dojo.declare( 'JBrowse.Model.TrackMetaData', null,
wordREs.push( new RegExp(currentWord,'i') );
}

// return a function that returns true if all of the words
// match the string, but in any order
return function( text ) {
return dojof.every( wordREs, function(re) { return re.test(text); } );
};
// return a function that takes on item and returns true if it
// matches the text filter
return dojo.hitch(this, function(item) {
return dojo.some( this.facets, function(facetName) {
var text = this.getValue( item, facetName );
return dojof.every( wordREs, function(re) { return re.test(text); } );
},this);
});
},

getFeatures: function() {
Expand Down

0 comments on commit 92fa1e8

Please sign in to comment.