Version 14 (modified by simonpj, 8 years ago) (diff)


This page summarises our current proposal for packages in GHC.

The problem

A vexed question in the current design of Haskell is the issue of whether a single program can contain two modules with the same name. Currently that is absolutely ruled out, and as a result packages are fundamentally non-modular: every package must use a distinct space in the global namespace.

There are two quite separate issues, addressed in "Question 1", "Question 2" below. First we give assumptions.


Before we start, note that we take for granted the following

  • Each package has a globally-unique name, organised by some social process. This assumption is deeply built into Cabal, and lots of things would need to change if it wasn't met.
  • Module names describe purpose (what it's for, e.g. Data.Bits), whereas package names describe provenance (where it comes from, e.g. "gtkhs"). We should not mix these two up, and that is a good reason for not combining package and module names into a single grand name. One quite frequently wants to globally change provenance but not purpose (e.g. compile my program with a new version of package "foo"), without running through all the source files to change the import statements.

Question 1: Can two different packages contain a module with the same module name?

We now think that's unreasonable to answer 'no', because that means that EVERY module in EVERY package written by ANYONE must have different module names. That's like saying that every function must have different local variables, and is a serious loss of modularity. We suspect that this is something about which we can all agree.

The only sensible way to fix it is to relax the language design so that

  • A module name must be unique within its package (only)

That means that module A.B.C could exist *both* in package P1 and in P2. And both packages could be linked into the same program. Suppose for the moment that A.B.C is not exposed by both P1 and P2. Then you would say simply:

  ghc --make Main -o app

The authors of packages P1 and P2 didn't need to know about each other, and don't need to choose globally unique module names.

For this to work, GHC must incorporate the package name (and version) into the names of entities the package defines. That means that when compiling a module M you must say what package it is part of:

  ghc -c -package package-name P1 C.hs

Then C.o will contain symbols like "P1.A.B.C.f" etc. In effect, the "original name" of a function f in module M of package P is <P,M,f>.

But what if two packages expose the same module A.B.C? That takes us to Question 2.

Question 2. When you say "import A.B.C", from what package does A.B.C come?

Here GHC already has a fairly elaborate scheme (perhaps too elaborate).

  • For a start, you may or may not have a package installed.
  • Even if you do, the package may or may not be exposed by default (reasoning: you may want old versions of package X to be installed, but not in scope by default).
  • Then, you can use the -hide-package flag to hide an otherwise-exposed package, and the -package flag to expose an otherwise-hidden package.

When GHC incorporates package names in exported symbols, you will be able to expose package P1 when compiling module M (say), and expose P2 when compiling module N by manipulating these flags. Then M and N could both import module A.B.C, which would come from P1 and P2 respectively. But:

  • What if you wanted to import A.B.C from P1 and A.B.C from P2 into the same module?
  • What if you want to only replace parts of P1 (e.g., you want to use an updated version of a module in base)?
  • Compiling different modules with different flags in a way that affects the semantics (rather than, say, the optimisation level) seems undesirable.
  • To support --make in this situation we'd need to allow -package flags in the per-module OPTIONS pragmas, which isn't currently supported. (ghc --make already gathers those options together for the link step.)

The obvious solution is to allow the programmer to specify the source package in the import line, something like this:

  import A.B.C from "base" ( map, filter )

That would presumably get the most recent installed incarnation of the base package. If you want a particular version of the package, we could allow

  import A.B.C from "base-3.4" ( map, filter )

The exact syntax is unimportant. The important thing is that the programmer can specify the package in the source text.

If we adopt the idea that an import statement can specify the source package, several design choices arise:

Is the 'from <package>' compulsory?

If you want to import A.B.C, a module exported by package "foo", can you say just import A.B.C, or must you say import A.B.C from "foo"?

We think of this as rather like the question "If you import f from module M, can you refer to it as plain "f", or must you refer to it as "M.f"? The answer in Haskell 98 is that you can refer to it as plain "f" so long as plain "f" is umambiguous; otherwise you can use a qualified reference "M.f" to disambiguate.

We propose to adopt the same principle for imports. That is, you can say "import A.B.C" so long as that is umambiguous (meaning that there is only one module A.B.C exported by any installed package). If the reference to A.B.C is ambiguous, you can qualify the import by adding "from "foo"".

Package versions

We probably want some special treatment for multiple versions of the same package. What if you have both "foo-3.9" and "foo-4.0" installed, both exporting A.B.C? This is jolly useful when you want to keep install new packages, but keep old ones around so you can try your program with the older one. So we propose that this is not regarded as ambiguous: importing A.B.C gets the latest version, unless some compiler flag (-hide-package) takes it of the running.

Importing from the home package

If A.B.C is in the package being compiled (which we call "the home package"), and in an exposed package, and you say import A.B.C, do you get an "ambiguous import" error , or does the current package override. And if the former, how can you say "import A.B.C from the current package"?

The 'as P' alias

We propose to maintain the local, within-module "as P" alias mechanism unchanged. Thus:

   import A.B.C( T ) from "foo" as M
   type S = M.T -> M.T

Here, the qualified name "M.T" refers to the T imported from A.B.C in package "foo".

Qualified names

We propose that the default qualified name of an entity within a module is just the module name plus the entity name. Thus

  import A.B.C( T ) from "foo" 
  type S = A.B.C.T -> A.B.C.T

If you want to import multiple A.B.C's (from different packages) then perhaps they define different entities, in which case there is no problem:

  import A.B.C( T1 ) from "foo" 
  import A.B.C( T2 ) from "bar" 
  type S = A.B.C.T1 -> A.B.C.T2

But if they both export entities with the same name, there is no alternative to using the 'as M' mechanism:

  import A.B.C( T ) from "foo" as M1
  import A.B.C( T ) from "bar" as M2
  type S = M1.T -> M2.T

Exporting modules from other packages

It is perfectly OK to export entities, or whole modules, imported from other packages:

  module M( f, g, module Q ) where
  import A.B( f, g ) from "foo"
  import X.Y.Z from "bar" as Q


Should package names be in quotes? Probably not. They have a well-defined syntax.

It's been suggested that one might want to import several modules from one package in one go:

    from "base" import
        Prelude hiding (length)
        qualified Data.List as List

What we don't like about that is that it needs a new keyword "from". Perhaps all imports can start with the keyword import, and then we are free to use extra (context-specific) keywords. (Haskell already has several of these, such as hiding. Something like this:

    import from "base" {
        Prelude hiding (length) ;
        Control.Exception ;
        qualified Data.List as List }
    import from "foo" M( x, y )

Here the layout is explicit, but perhaps the layout rule could apply.

Indeed, we could allow this multiple form even for ordinary imports:

   import { A(f); B(g); C(S,T) }