It really depends on what you want your final structure to be.
If you don't mind a non-flat object structure, you can do nested _.groupBy calls:
var result = _(givenArray) //begin chain
.groupBy('a')
.mapValues(function(groupedByA) {
return _(groupedByA) //begin chain
.groupBy('b')
.mapValues(function (groupedByAAndB) {
return _.groupBy(groupedByAAndB, 'c');
})
.value(); //end chain
})
.value(); //end chain
//Then you can do things like:
result[5][4][3]; //All items where a=5, b=4, and c=3.
Downside here is that there's an extra level of nesting for each property you group by, and result[5][4] will blow up if there aren't any results where a=5 for example. (Though you could use a library like koalaesce for that)
Alternatively, if you want a flat object structure, but don't mind it being a bit ungainly to access the items, then:
var result = _.groupBy(givenArray, function (item) {
return JSON.stringify(_.pick(item, 'a','b','c'));
});
//Much simpler above, but accessed like:
result[JSON.stringify({a: 5, b:4, c:3})]
Much simpler and scales better to grouping by more things... but awkward to work with, since the keys are full JSON strings; but then you also don't have the null issue that the first option has.
You can also, just use _.values or some equivalent to turn the flat object structure into a single array of arrays. Then there's obviously no "random" access to get all items with a given value of a,b, and c, but you can always loop over it.