[Haskell-cafe] conditionally exposing modules in cabal

[Antoine Latter]

On Thu, Aug 18, 2011 at 3:33 AM, Rogan Creswick <creswick at gmail.com> wrote:
> I would like to conditionally expose a number of internal modules in a
> library depending on a cabal flag - the idea is that new features
> could be implemented externally to the library without contaminating
> the source with undesirable changes.  However, I've been unable to
> find a way to structure a cabal file that does not require repeating
> the list of modules.  For example, this works:
>

One problem to consider - a downstream user of the new features won't
know that they need to pass special flags to your module, and may not
even know that they are using your module if the dependency is a few
steps removed.

What some folks have taken to doing is to either turn of Haddock
documentation for the modules (as in Data.ByteString.Internal) or just
provide great big warning messages that the functions provided can be
used in ways which do not provide key invariants and that the contents
of the module is subject to change at any time.

Antoine

>  exposed-modules:
>    Public.Stable.Module
>  if flag(exposeInternals)
>    exposed-modules:
>       <long list of internal modules>
>  else
>    other-modules:
>       <long list of internal modules, identical to above>
>
> But I don't want to maintain two identical lists of modules in the cabal file.
>
> I've tried putting the section headers in the if's, but this doesn't
> appear to work (cabal doesn't complain, but the modules to be exposed
> are not available as expected):
>
>  exposed-modules:
>    Public.Stable.Module
>  if flag(exposeInternals) {
>  exposed-modules:
>  } else {
>  other-modules:
>  }
>    <long list of internal modules>
>
>
> Does anyone have suggestions?
>
> Thanks!
> Rogan
>
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>