Omit filtered-out groupings of map-reduce

Not entirely sure whether to file this under crossfilter or dc.js, and this phenomenon probably happens by design, but could someone explain to me the reasoning behind this:

1. See fiddle http://jsfiddle.net/amergin/23yRD/ for a simple example
2. In the fiddle, two dimensions are created for the samples. Create a filter on typeDimension and on the other dimension you should have two sample rows matching the “visa” filter.
3. Group the tipDimension and do reduce providing the custom init/add/remove functions. Result:

console.log( JSON.stringify( reducedGroup.top(Infinity) ) );
[{"key":100,"value":{"count":1,"tip":100,"total":200,"quantity":1,"set":0}},{"key":200,"value":{"count":1,"tip":200,"total":300,"quantity":1,"set":1}},{"key":0,"value":{"count":0}}] 

1. In the result, the third grouping seems to be for those filtered-out samples. From the structure it can be seen that it has only gone through reduceInitial function and therefore is without payload.

My question is this: is there a way to omit this third grouping or easily avoid it being formed? The problem comes clear on the scatterplot: the keyAccessor & valueAccessor both return a point (undefined,undefined) which is plotted on the top-left corner. What am I missing here?

[gordonwoodhull]

I documented the technique here: https://github.com/dc-js/dc.js/wiki/FAQ#filter-the-data-before-its-charted

Hope to create a crossfilter wrappers library to support this sort of thing. I don’t think it should be supported directly by dc.js, so I’m closing this issue.

[gordonwoodhull]

Right now I’m responding to another question on the list about this very frequently requested feature. I forgot to mention that you can also create a “pseudo-group”; since dc.js really only calls the .all() method on a group, it’s pretty easy to create another object that wraps the group and implements its own .all() that filters the results.

Pretty hacky, and it would be great to support this directly, but a workaround for now…