This reverts commit 51dc75b769. The pretty
library is a core package since (at least) GHC 6.10.4, i.e. every single one of
our compilers ships this library by default. Adding the package to the global
namespace like this is almost certainly going to cause trouble because we'll
end up with inconsistently linked binaries whenever packages mix the user-space
pretty library with the core one.
Other tools like hsc2hs or haddock allows this, too, and some tools explicitly
set memory limits on these tools during the build (i.e. llvm-general-pure) by
exposing a GHCRTS environment variable. See [1] for a concrete example of this
issue.
[1] https://github.com/NixOS/nixpkgs/issues/2939
Unfortunately, there are still build errors with this package (see
https://github.com/ekmett/parsers/issues/39). We need the latest
version, though, because older ones won't work with the latest
attoparsec.
1) Packages formerly called haskell-haskell-platform-ghcXYZ-VVVV.X.Y.Z are
now called haskell-platform-VVVV.X.Y.Z. The latest version can be
installed by running "nix-env -i haskell-platform".
2) The attributes haskellPackages_ghcXYZ.haskellPlatform no longer exist.
Instead, we have attributes like haskellPlatformPackages."2012_4_0_0".
(The last numeric bit must be quoted when used in a Nix file, but not on
the command line to nix-env, nix-build, etc.) The latest Platform has a
top-level alias called simply haskellPlatform.
3) The haskellPackages_ghcXYZ package sets offer the latest version of every
library that GHC x.y.z can compile. For example, if 2.7 is the latest
version of QuickCheck and if GHC 7.0.4 can compile that version, then
haskellPackages_ghc704.QuickCheck refers to version 2.7.
4) All intermediate GHC releases were dropped from all-packages.nix to
simplify our configuration. What remains is a haskellPackages_ghcXYZ set
for the latest version of every major release branch, i.e. GHC 6.10.4,
6.12.3, 7.0.4, 7.2.2, 7.4.2, 7.6.3, 7.8.2, and 7.9.x (HEAD snapshot).
5) The ghcXYZPrefs functions in haskell-defaults.nix now inherit overrides
from newer to older compilers, i.e. an override configured for GHC 7.0.4
will automatically apply to GHC 6.12.3 and 6.10.4, too. This change has
reduced the redundancy in those configuration functions. The downside is
that overriding an attribute for only one particular GHC version has become
more difficult. In practice, this case doesn't occur much, though.
6) The 'cabal' builder has a brand-new argument called 'extension'. That
function is "self : super : {}" by default and users can override it to
mess with the attribute set passed to cabal.mkDerivation. An example use
would be the definition of darcs in all-packages.nix:
| darcs = haskellPackages.darcs.override {
| cabal = haskellPackages.cabal.override {
| extension = self : super : {
| isLibrary = false;
| configureFlags = "-f-library " + super.configureFlags or "";
| };
| };
| };
In this case, extension disables building the library part of the package
to give us an executable-only version that has no dependencies on GHC or
any other Haskell packages.
The 'self' argument refers to the final version of the attribute set and
'super' refers to the original attribute set.
Note that ...
- Haskell Platform packages always provide the Haddock binary that came with
the compiler.
- Haskell Platform 2009.2.0.2 is broken because of build failures in cgi and
cabal-install.
- Haskell Platform 2010.1.0.0 is broken becasue of build failures in cgi.
Now that both self and super are available to prefFun, we can use self, where appropriate to access late bound versions of most
packages.
When extensions are not used, there is no difference between self and super.
The existing knot-tying code I felt was a bit incoherent with result, finalReturn, self, refering to different various forms of the "haskellPackages" value and often
different forms in the same place.
This commit instills some object-oriented discipline to the construction of hasekllPackages using a small number of fundamental OO concepts:
* An class is a open recursive function of the form (self : fooBody) where fooBody is a set.
* An instance of a class is the fixed point of the class.
This value is sometimes refered to as an object and the values in the resulting set are sometimes refered to as methods.
* A class, foo = self : fooBody, can be extended by an extension which is a function bar = (self : super : barBody) where barBody a set of overrides for fooBody.
The result of a class extension is a new class whose value is self : foo self // bar self (foo self).
The super parameter gives access to the original methods that barBody may be overriding.
This commit turns the haskell-packages value into a "class".
The knot-tying, a.k.a the object instanitation, is moved into haskells-defaults. The "finalReturn" is no longer needed and is eliminated from the body of
haskell-packages. All the work done by prefFun is moved to haskell-defaults, so that parameter is eliminated form haskell-packages. Notice that the old prefFun took
two pameters named "self" and "super", but both parameters got passed the same value "result". There seems to have been some confusion in the old code.
Inside haskell-defaults, the haskell-packages class is extended twice before instantiation. The first extension is done using prefFun argument.
The second extension is done the extension argument, which is a renamed version of extraPrefs.
This two stage approach means that extension's super gets access to the post "perfFun" object while previously the extraPrefs only had access to the pre "prefFun"
object. Also the extension function has access to both the super (post "perfFun") object and to self, the final object. With extraPrefs, one needed to use the
"finalReturn" of the haskell packages to get access to the final object. Due to significant changes in semantics, I thought it best to replace extraPrefs with
extension so that people using extraPrefs know to update thier cod.
Lastly, all the Prefs functions have renamed the "self" parameter to "super". This is because "self" was never actually a self-reference in the object oriented sense
of the word. For example
Cabal_1_18_1_3 = self.Cabal_1_18_1_3.override { deepseq = self.deepseq_1_3_0_2; };
doesn't actually make sense from an object oriented standpoint because, barring further method overriding, the value of Cabal_1_18_1_3 would be trying to override it's
own value which simply causes a loop exception. Thankfully all these uses of self were really uses of super:
Cabal_1_18_1_3 = super.Cabal_1_18_1_3.override { deepseq = super.deepseq_1_3_0_2; };
In this notation the overriden Cabal_1_18_1_3 method calls the Cabal_1_18_1_3 of the super-class, which is a well-founded notion.
Below is an example use of using "extension" parameter
{
packageOverrides = pkgs : {
testHaskellPackages = pkgs.haskellPackages.override {
extension = self : super : {
transformers_0_4_1_0 = self.cabal.mkDerivation (pkgs: {
pname = "transformers";
version = "0.4.1.0";
sha256 = "0jlnz86f87jndv4sifg1zpv5b2g2cxy1x2575x727az6vyaarwwg";
meta = {
description = "Concrete functor and monad transformers";
license = pkgs.stdenv.lib.licenses.bsd3;
platforms = pkgs.ghc.meta.platforms;
maintainers = [ pkgs.stdenv.lib.maintainers.andres ];
};
});
transformers = self.transformers_0_4_1_0;
lensFamilyCore = super.lensFamilyCore.override { transformers = self.transformers_0_3_0_0; };
};
};
};
}
Notice the use of self in the body of the override of the transformers method which references the newly defined transformers_0_4_1_0 method.
With the previous code, one would have to instead akwardly write
transformers = super.finalReturn.transformers_0_4_1_0;
or use a rec clause, which would prevent futher overriding of transformers_0_4_1_0.
