I'm not familiar with automatic cabal flags. Could you point me to documentation? On Thu, Apr 19, 2018 at 3:01 PM, Claude Heiland-Allen wrote:

Hi David,

On 19/04/18 18:13, David Feuer wrote:

...

Is there any way to determine whether Integer comes from integer-gmp or integer-simple? I'm playing with the idea of using the underlying representation to get more compact/efficient tries, but I need to be able to find out what that representation is. One way could be to use an automatic cabal flag. With it enabled, depend on integer-gmp and add hs-source-dirs to a directory containing your integer-gmp implementation, with it disabled, depend on integer-simple and add hs-source dirs to a directory containing your integer-simple implementation. This gives module-level control. With automatic flag I think cabal will try both and choose the setting that gives the best build plan (likely corresponding to the Integer implementation of your ghc).

Claude -- https://mathr.co.uk _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

Hi David, you could  check http://hackage.haskell.org/package/integer-logarithms I don't use different source-dirs (I could), the implementation isn't that different so  CPP didn't felt too intrusive. The http://cabal.readthedocs.io/en/latest/developing-packages.html#resolution-of... section speaks about automatic flags. TL;DR cabal solver will toggle _automatic_ flags (manual flags are set to defaults). It's recommended to make flag assignment disjoint, e.g. note how bytestring version ranges are disjoint.     if flag(bytestring-builder)       build-depends: bytestring >= 0.9.2 && < 0.10.4,                      bytestring-builder >= 0.10.4 && < 1     else       build-depends: bytestring >= 0.10.4 && < 0.11 Similarly, in `integer-logarithms` we have     flag integer-gmp       description:  integer-gmp or integer-simple       default:      True       manual:       False     library     ...      if flag(integer-gmp)         build-depends:           integer-gmp < 1.1      else         build-depends:           integer-simple           -- here we could have hs-source-dirs relying that there aren't install plan with both integer-gmp and integer-simple For very complicated example see: functor-classes-compat http://hackage.haskell.org/package/functor-classes-compat-1/functor-classes-... Hopefully these help Cheers, Oleg. On 19.04.2018 23:05, David Feuer wrote:

I'm not familiar with automatic cabal flags. Could you point me to documentation?

On Thu, Apr 19, 2018 at 3:01 PM, Claude Heiland-Allen wrote:

...

Hi David,

On 19/04/18 18:13, David Feuer wrote:

...

Is there any way to determine whether Integer comes from integer-gmp or integer-simple? I'm playing with the idea of using the underlying representation to get more compact/efficient tries, but I need to be able to find out what that representation is. One way could be to use an automatic cabal flag. With it enabled, depend on integer-gmp and add hs-source-dirs to a directory containing your integer-gmp implementation, with it disabled, depend on integer-simple and add hs-source dirs to a directory containing your integer-simple implementation. This gives module-level control. With automatic flag I think cabal will try both and choose the setting that gives the best build plan (likely corresponding to the Integer implementation of your ghc).

Claude -- https://mathr.co.uk _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

---

Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.