Re: [Haskell-cafe] proper way to generate a random data in criterion

On Wed, Oct 19, 2011 at 12:21 PM, Gregory Collins wrote:

On Wed, Oct 19, 2011 at 5:03 PM, Johan Tibell wrote:

...

It does. You need to use evaluate to have ensure actually be evaluated.

I'm almost certain you're wrong about this. The bang pattern on the return from ensure (!r1 <- ensure $ ...) forces r1 to WHNF, which goes through deepseq, and thus the whole list is forced. See https://gist.github.com/1299380 for a short counterexample.

I should have paid more attention; I missed the bangs on the bindings. I still recommend the pattern I linked in my previous email. If you want to do it they way you currently do use let !foo = xs `deepseq` xs no return needed.