return bit indices set during an add

[simon-weber]

My project involves multiple nodes maintaining and syncing up bloom filters. To do this efficiently, I need to know which indices get set during an add operation. This PR adds a method to do that.

This isn't quite ready to be merged. Here are some questions for you:

• would it slow things down to consume the hashes iterator before the loop so we can pre-allocate the indices list? I'd like to avoid dynamically expanding the indices list.
• what do you want the client interface for this to be? My first thought was that add_return_bits is a more general interface and that add should be calling it, but that would introduce a memory/gc perf regression for code not using the bit indices.

[simon-weber]

Hm, actually, maybe this function should be a generator of bit indices instead. That would let the caller opt-in to the list allocation overhead if they need it.

[jaybaird]

Hey, let me take a look at this weekend and I’ll get back to you. Thanks!

On Nov 1, 2014, at 9:05 AM, Simon Weber notifications@github.com wrote:

Hm, actually, maybe this function should be a generator of bit indices instead. That would let the caller opt-in to the list allocation overhead if they need it.

— Reply to this email directly or view it on GitHub.