String != [Char]

[Duncan Coutts]

On 17 March 2012 01:44, Greg Weber <greg at gregweber.info> wrote:
> the text library and Text data type have shown the worth in real world
> Haskell usage with GHC.
> I try to avoid String whenever possible, but I still have to deal with
> conversions and other issues.
> There is a lot of real work to be done to convert away from [Char],
> but I think we need to take it out of the language definition as a
> first step.

I'm pretty sure the majoirty of people would agree that if we were
making the Haskell standard nowadays we'd make String type abstract.

Unfortunately I fear making the change now will be quite disruptive,
though I don't think we've collectively put much effort yet into
working out just how disruptive.

In principle I'd support changing to reduce the number of string types
used in interfaces. From painful professional experience, I think that
one of the biggest things where C++ went wrong was not having a single
string type that everyone would use (I once had to write a C++
component integrating code that used 5 different string types). Like
Python 3, we should have two common string types used in interfaces:
string and bytes (with implementations like our current Text and
ByteString).

BTW, I don't think taking it out of the langauge would be a helpful
step. We actually want to tell people "use *this* string type in
interfaces", not leave everyone to make their own choice. I think
taking it out of the language would tend to encourage everyone to make
their own choice.

Duncan