Remove eq and show from num class

[Anthony Clayden]

> On 8 Sep 2017 at 07:54 Nathan van Doorm wrote:

Thank you Nathan and Mario for your explanations.

> The motivation is that there are many types with sensible
> definitions for addition and multiplication etc that can't
> be instances of Eq or Show, for example functions to
> numbers (allowing us to write "sin + cos" instead of "\x
> -> sin x + cos X") or arbitrary real numbers (where
> comparison isn't necessarily computable).

The Haskell 2010 report section 6.4 Numbers starts:

"Haskell provides several kinds of numbers; ...".
Then goes on to class `Num`.

Functions `sin`, `cos` are not numbers,
so it's not sensible to put them in `Num`.

By the same token, I wouldn't expect Haskell to figure out
tan == sin / cos
so it's not sensible to put those in `Eq`.

If you want to write higher-order algebraic expressions
with the Num operators, override the Prelude.

(Perhaps this is part of a wider issue
 to do with reorganising `Num` so it has nicer
 'mathematical' structure: additive, multiplicative, etc.
 That might also cover that `Eq` is dodgy
 for floating-point or `Rational`s representations.
 But that reorg has seemed prohibitively hard.)

> 
> I'm not sure why this is being discussed on the mailing
> list rather than the Github proposals thing, ...

Quite. I've found the github proposals process for GHC
very valuable.

> but I do know that that has grown somewhat inactive

The Haskell-prime list has been fairly inactive too.
It was originally formed to identify stable features,
that were to go into a revised language standard
that was to be an evolution from H98.

"Haskell' will be a conservative refinement of Haskell 98."
it still says on the listinfo. (Note the future tense.)

That process ended with the H2010 report.
I must admit I thought (wrongly, it seems) that 
Haskell-prime rather died after that.

Somebody could start by loading up
the outstanding proposals from the Haskell-prime list.
(Herbert says this one is from 2011.)
 
> and this is as good as anything for getting the
> Haskell' committee to make a final judgement.
> 

Do we need any judgment at all?
The skies do not seem to be falling.

> This has been how it is in GHC for a long time now,
> so it really is a matter for the Haskell' committee
> rather than one of the GHC committees.

MPTCs, GADTs (for example) have been in GHC
far longer.

OK it's bit naughty GHC doesn't have a flag
for something that's not compliant to the report.
But that's a GHC issue, not a grounds for changing
the language spec.


AntC

> 
>> On 8 Sep 2017 8:35 am, "Anthony Clayden" wrote:
>> ...