Add IsList instances

[Bodigrim]

Resolves #106, as suggested by @dcoutts in #114 (and supersedes the latter).
CC @andrewthad

[sjakobi]

How does this behave for literals that don't fit into Word8, e.g.

[ 256 ] :: ByteString

or

[ -1 ] :: ByteString

?

[Bodigrim]

Similar to Data.ByteString.pack.

[-1] throws warning: [-Woverflowed-literals] Literal -1 is out of the GHC.Word.Word8 range 0..255, as expected. Surprisingly enough, [256] does not throw such warning, when OverloadedLists are enabled, and silently returns "\NUL".

This seems to be a bug in GHC:

> [256] :: [Data.Word.Word8]

<interactive>:1:2: warning: [-Woverflowed-literals]
    Literal 256 is out of the GHC.Word.Word8 range 0..255
[0]
> :set -XOverloadedLists
> [256] :: [Data.Word.Word8]
[0]

[sjakobi]

@Bodigrim That looks worth reporting!

as suggested by @dcoutts in #116 (and supersedes the latter).

dcoutts didn't comment on #116. Did you mean a different issue?

[Bodigrim]

Sure, filed as https://gitlab.haskell.org/ghc/ghc/issues/18172

Good catch. Sorry, I meant #114.

[andrewthad]

The only thing arguably missing is a more efficient implementation of fromListN (rather than the default), but I don't really care one way or the other since I don't use overloaded lists.

[Bodigrim]

I did not define fromListN, because it is difficult to make it both strictly more efficient than fromList and at the same time strictly conforming to the definition "The fromListN function takes the input list's length as a hint".

That said, I believe this is ready for review. @hvr @cartazio

[cartazio]

@Bodigrim im inclined that we should make FromList N validating. Rather than a hint. This was the take away about semantic mismatch on the libraries list. And since bytestring and vector are used in data loading contexts this is even more important!

[cartazio]

We should have a new discussion on the libraries list to discuss this. I think the hint semantic was the wrong choice at the time. ESP since the api then was about a desugaring trick rather than a user api

[Bodigrim]

@cartazio sure, I remember that discussion on the libraries list and I totally agree that it does not make sense to treat n as a hint. But as long as this statement remains unchanged in GHC.Exts, I hesitate to ignore it.

I mean, anyone who cares about performance can reach for unsafePackLenBytes directly.

[cartazio]

I mean we should circulate the semantic change in docs on libraries list. But understood

[cartazio]

Wrt base

[Bodigrim]

@hsyl20 could you please review?

[hsyl20]

LGTM