Ilya Grigorik

Do you really need to call it with rule_set.train_data? Looking at the code, we memoize @accuracy so I think you can just cal r.accuracy and get the value computed during...

Guessing, you're on an older version of ruby? This was introduced in https://github.com/igrigorik/decisiontree/commit/13aed0b2ae57e994e3366fedf15530d86c47bccd. /cc @cheerfulstoic

Right, figures... Sum is 2.4+ http://blog.bigbinary.com/2016/11/02/ruby-2-4-introduces-enumerable-sum.html. @cheerfulstoic I propose we revert the sum update until 2.3 is deprecated.

@elestrade nice catch. The code _definitely_ needs some careful refactoring. If you spot any other gotchas like that, let me know (pull's are welcome too :))

Do you have an example test case to reproduce this behavior? I'm still a bit unclear on how this path is triggered.

/cc @martoche @eriwen @boxofrad, who've contributed most of the gzip streaming logic. @drbrain thanks for the heads up. It's not clear to me yet if this will break any of...

Gotcha, thanks Eric. We'll investigate..

Is there a tl;dr on the motivation and impact of this change in core? Not sure I fully grok either. That said, happy to take/review PR's, if anyone is willing.....

Interesting, thanks for the heads up. As you can tell, not something on my radar.. but if anyone is interested in making a PR, I'd be happy to review.

I think belongs on EM repo... as I can't do much until we have that.