[Haskell-cafe] conflicting variable definitions in pattern

[Lennart Augustsson]

Simplicity of pattern matching semantics, not of implementation (we
all knew how to implement it).
Miranda had non-linear patterns, but nobody really argued for them in Haskell.
If Haskell had them, I'd not argue to have them removed, but nor will
I argue to add them.

  -- Lennart

On Fri, May 15, 2009 at 1:19 PM, Conor McBride
<conor at strictlypositive.org> wrote:
>
> On 15 May 2009, at 12:07, Lennart Augustsson wrote:
>
>> In the original language design the Haskell committee considered
>> allowing multiple occurrences of the same variable in a pattern (with
>> the suggested equality tests), but it was rejected in favour of
>> simplicity.
>
> Simplicity for whom, is the question? My point is
> only that there's no technical horror to the proposal.
> It's just that, given guards, the benefit (in simplicity
> of program comprehension) of nonlinear patterns over
> explicit == is noticeable but hardly spectacular.
>
> Rumblings about funny termination behaviour, equality
> for functions, and the complexity of unification (which
> isn't the proposal anyway) are wide of the mark. This
> is just an ordinary cost-versus-benefit issue. My guess
> is that if this feature were already in, few would be
> campaigning to remove it. (By all means step up and say
> why!) As it's not in, it has to compete with other
> priorities: I'm mildly positive about nonlinear
> patterns, but there are more important concerns.
>
> Frankly, the worst consequence I've had from Haskell's
> pattern linearity was just my father's derision. He
> quite naturally complained that his programs had lost
> some of their simplicity.
>
> All the best
>
> Conor
>
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>