On 01/29/2017 09:47 PM, Andy Lester wrote: >> The point is the conceptual complexity involved, and how it would >> impact our ability to switch to better implementations for our >> associative arrays. > Those are fine arguments, and I'm not not taking a stand one way or the other. I just think that if you're going to make the argument that we "pay a penalty" then that penalty should be quantified, or else people can't reasonably consider it. > > Also, we will want to explain in the upcoming release notes that "hash lookups are now X.X% faster, and hash stores are Y.Y% faster." I think Andy's comment leads to my comments: If we can prove a difficulty in maintaining the code and/or difficulty in introducing new functionality and/or a major improvement in speed, this can be justified. At the same time, we need to estimate the number of breakages and how bad they would be. We should also look at optionally providing such functionality in a separate module, maybe using ties (or some other magic) to provide users with alternatives to it. I appreciate starting the discussion on this and reaching further to the community to illicit more comments.Thread Previous | Thread Next