Document how and why P came to be and the decisions made

[erikd-ambiata]

At ICFP it was suggested that people using custom preludes for commercial work should document how their preludes differ from the official one which may help guide the development of the official prelude.

[jacobstanley]

I don't know if it was intentional or not, but it has decoupled us from churn in the base libraries and has hence provided us with stability across many major versions of GHC.

[erikd-ambiata]

The only base problems I can think of are:

• Applicative becoming a superclass of Monad in ghc 7.10.
• Foldable/Traversable in ghc 8.0.

both of which are in P. Is there anything else?

[jacobstanley]

Yeah, those are the culprits, and for the most part we can compile across 7.8, 7.10, 8.0 without any CPP in our libraries. I wouldn't say that's the reason for P though, just a happy coincidence.

I think it's more about exposing more general versions of things, and hiding unsafe things. The readme outlines the principles pretty well:

Guidance - "Should it go in p"

• Is something that is very general & useful, with an obvious semantic and can be implemented from base - YES.
• If it is unsafe - NO.
• Is there a more general version - NO.
• Is it something available in base in only some GHC versions - YES.

[HuwCampbell]

Symbolic mappend and semigroup's (<>) in future. At the moment we're exporting symbolic mappend, but we should be able to change it, which will make using 8.2 and lower less painful.