[Haskell-cafe] Numerics and Warnings

[Aleksey Khudyakov]

On 10 April 2013 22:25, Roman Cheplyaka <roma at ro-che.info> wrote:
> * Barak A. Pearlmutter <barak at cs.nuim.ie> [2013-04-10 15:38:35+0100]
>> In fiddling around with some numeric code in Haskell, I noticed some
>> issues.  Basically, you get warnings if you write
>>
>>   energy mass = mass * c^2
>>
>> but not if you write
>>
>>   energy mass = mass * c * c
>>
>> which seems a bit perverse.
>> Some more examples are below.
>>
>> I understand the inference issues that cause this, but common innocuous
>> cases could---and I would argue, should---be addressed in ad-hoc ways.
>
> Hi Barak,
>
> In a sense, defaulting in Haskell *is* a mechanism to address common
> innocuous cases in an ad-hoc way (although it still has a relatively
> simple and easy to understand semantics).
>
This IS rather annoying problem for numeric code. Raising value to positive
power is quite common operation yet ^ operator generally couldn't be used
because it leads to warning about type defaulting (rightfully) and one
wants to keep code warning free. Actually it's problem with warnings and
I don't think adding some ad-hoc rules for generating warning is necessarily
bad idea