De-composite helper functions #47
Conversation
Without inter-dependencies between strategies, we can just return a strategy from a function. This has somewhat better performance because it avoids re-constructing and re-validating strategy objects (even though they're cached).
|
Thanks for the PR. I wonder if you could also comment on whether my thinking in #37 was correct. If you have any other suggestions on improvements on my hypothesis usage, that would also be helpful. It's mostly been working for me, but I'm sure I am doing some bad things. |
|
About splitting positive and negative integers? We do the same thing internally by drawing a sign bit, so there's almost no difference from using Nothing else jumps out - one of our API design goals is that obvious uses should work well, and you're not doing anything that exotic yet. Personally I'd try to use upstream strategies more (e.g. array_shapes), but that's more a matter of taste than function. Feel free to ping me for pr reviews going forward, if that would be helpful! |
|
Ah OK. But am I at least correct that splitting things in one_of guides hypothesis about the split? |
Yep, it'll take that structure into account. |
Without inter-dependencies between strategies, we can just return a strategy from a function. This has somewhat better performance because it avoids re-constructing and re-validating strategy objects (even though they're cached).
Loosely related to #46.