Control.Monad.Writer.Strict not as strict as expected

[Matt Brown]

Thanks for the clarification.  Though the lazy version suits my
present need well, I wonder if a variant strict in its monoid would be
useful as well.  If nothing else, it might help highlight that the
Strict version is not, something I didn't realize until reading the
source.
-matt

On Thu, Sep 15, 2011 at 2:02 PM, Edward Kmett <ekmett at gmail.com> wrote:
> On Thu, Sep 15, 2011 at 4:43 PM, Matt Brown <matt at softmechanics.net> wrote:
>>
>> Hello all,
>>
>> I've been debugging a space leak, and believe I've traced it to a
>> chain of unevaluated calls to mappend in the bind operator for
>> Control.Monad.Writer.Strict.  I had expected these calls to be
>> evaluated strictly by the strict Writer, but it doesn't seem to be the
>> case.  Am I understanding this correctly?
>
> The Strict in Control.Monad.Writer.Strict refers to the fact that the tuple
> is matched strictly -- not the Monoid.
>
>>
>> If so, is this the intended
>> behavior?
>
> The Strict/Lazy writer split has nothing to do with the strictness of your
> monoid, but rather whether or not operations such as:
>> fmap f ~(w, a) = (w, f a)
>
> or
>> fmap f (w, a) = (w, f a)
> are used throughout the API. We default to Lazy because slightly more code
> can terminate (at least in the case of State), lthough some would challenge
> that due to the behavior in the presence of bottoms the Strict versions are
> more correct.
> Neither of these are, or are intended to be, the monad you are looking for.
> -Edward Kmett
>