Closed 414owen closed 9 months ago
This was discussed before at https://github.com/LeventErkok/sbv/issues/659
And for the reasons cited there, I'm not terribly excited about this idea. For individual users (that only have a few dependencies) extra cost should be minimal. For someone compiling hackage, I guess you already are committing so much resources so I don't see how this separation will make any change to your bottom line..
I don't expect everyone to agree with my view on this, but the cost of maintenance is really high for me for separate packages.
I think that if the test suite imports the Documentation/examples cabal target, then there won't be a problem with examples getting out of sync.
Personally, I think that reducing iteration time (building the library) is quite valuable for maintenance.
But maybe your workflow is different somehow :)
Thanks for the advice. I’ll consider it. I’m kind of fond of having documentation show up in hackage prominently, so that’s another consideration.
Sorry for the drive-by issue creation - I'm not a real user of
sbv
. I just realised, when building a large enough subset of hackage to test a change, thatsbv
seemed to take an unusually long time to compile.It looks like the example modules are exposed in the library of
sbv
. This means that the examples get linked into binaries that usesbv
and are loaded into memory when the programs are run. I think that creating anexamples
library, that depends on thesbv
library would solve this? It would also take less time for people to compilesbv
as a library,