Comments on: You Lazy Thunk! http://blog.tmorris.net/you-lazy-thunk/ The weblog of Tony Morris Sat, 19 May 2012 03:16:30 +0000 http://wordpress.org/?v=2.6 By: Tony Morris http://blog.tmorris.net/you-lazy-thunk/#comment-1012 Tony Morris Fri, 18 May 2007 01:19:43 +0000 http://blog.tmorris.net/you-lazy-thunk/#comment-1012 Thanks Chad, you're right Thanks Chad, you’re right

]]> By: Chad Scherrer http://blog.tmorris.net/you-lazy-thunk/#comment-1011 Chad Scherrer Wed, 16 May 2007 22:37:53 +0000 http://blog.tmorris.net/you-lazy-thunk/#comment-1011 I think you mean "normal form", without the "weak head". WHNF evaluates a data type to its top-level constructor, so a list to either [] or (_:_), but does not evaluate the elements of the list. Normal form forces evaluation of all the elements. I think you mean “normal form”, without the “weak head”. WHNF evaluates a data type to its top-level constructor, so a list to either [] or (_:_), but does not evaluate the elements of the list. Normal form forces evaluation of all the elements.

]]> By: Ricky Clarkson http://blog.tmorris.net/you-lazy-thunk/#comment-1010 Ricky Clarkson Wed, 16 May 2007 05:17:38 +0000 http://blog.tmorris.net/you-lazy-thunk/#comment-1010 It should be noted that the runtimes and/or compilers are generally free to optimise two calls to f() to one call, as long as they preserve side effects and results. Usually, side effects are the real reason why calling f() twice is bad - the writer you mention probably knows this, but has still got efficiency bouncing around his head. I'm guessing that the article you read was about macro expansions, in which case it is definitely problematic to evaluate an argument multiple times in a macro expansion, if evaluating the argument has side effects. Macros that evaluate an argument multiple times, and hence leak their implementation details, should be annotated as such - typical C style is to make them uppercase - encouraging caution in calling. It should be noted that the runtimes and/or compilers are generally free to optimise two calls to f() to one call, as long as they preserve side effects and results.

Usually, side effects are the real reason why calling f() twice is bad - the writer you mention probably knows this, but has still got efficiency bouncing around his head.

I’m guessing that the article you read was about macro expansions, in which case it is definitely problematic to evaluate an argument multiple times in a macro expansion, if evaluating the argument has side effects. Macros that evaluate an argument multiple times, and hence leak their implementation details, should be annotated as such - typical C style is to make them uppercase - encouraging caution in calling.

]]>