Safe Haskell | None |
---|---|
Language | Haskell2010 |
A module for producing forward-defined build systems, in contrast to standard backwards-defined build systems such as shake. Based around ideas from fabricate. As an example:
import Development.Shake import Development.Shake.Forward import Development.Shake.FilePath main =shakeArgsForward
shakeOptions
$ do contents <-readFileLines
"result.txt"cache
$cmd
"tar -cf result.tar" contents
Compared to backward-defined build systems (such as normal Shake), forward-defined build systems tend to be simpler for simple systems (less boilerplate, more direct style), but more complex for larger build systems (requires explicit parallelism, explicit sharing of build products, no automatic command line targets). As a general approach for writing forward-defined systems:
- Figure out the sequence of system commands that will build your project.
- Write a simple
Action
that builds your project. - Insert
cache
in front of most system commands. - Replace most loops with
forP
, where they can be executed in parallel. - Where Haskell performs real computation, if zero-build performance is insufficient, use
cacheAction
.
All forward-defined systems use AutoDeps
, which requires fsatrace
to be on the $PATH
.
You can obtain fsatrace
from https://github.com/jacereda/fsatrace. You must set
shakeLintInside
to specify where AutoDeps
will look for dependencies - if you want all dependencies
everywhere use [""]
.
This module is considered experimental - it has not been battle tested. There are now a few possible alternatives in this space:
- Pier http://hackage.haskell.org/package/pier/docs/Pier-Core-Artifact.html (built on Shake).
- Rattle https://github.com/ndmitchell/rattle (by the same author as Shake).
- Stroll https://github.com/snowleopard/stroll.
Synopsis
- shakeForward :: ShakeOptions -> Action () -> IO ()
- shakeArgsForward :: ShakeOptions -> Action () -> IO ()
- forwardOptions :: ShakeOptions -> ShakeOptions
- forwardRule :: Action () -> Rules ()
- cache :: (forall r. CmdArguments r => r) -> Action ()
- cacheAction :: (Typeable a, Binary a, Show a, Typeable b, Binary b, Show b) => a -> Action b -> Action b
- cacheActionWith :: (Typeable a, Binary a, Show a, Typeable b, Binary b, Show b, Typeable c, Binary c, Show c) => a -> b -> Action c -> Action c
Documentation
shakeForward :: ShakeOptions -> Action () -> IO () Source #
Run a forward-defined build system.
shakeArgsForward :: ShakeOptions -> Action () -> IO () Source #
Run a forward-defined build system, interpreting command-line arguments.
forwardOptions :: ShakeOptions -> ShakeOptions Source #
Given a ShakeOptions
, set the options necessary to execute in forward mode.
forwardRule :: Action () -> Rules () Source #
cache :: (forall r. CmdArguments r => r) -> Action () Source #
Apply caching to an external command using the same arguments as cmd
.
cache $ cmd "gcc -c" ["foo.c"] "-o" ["foo.o"]
This command will be cached, with the inputs/outputs traced. If any of the
files used by this command (e.g. foo.c
or header files it imports) then
the command will rerun.
cacheAction :: (Typeable a, Binary a, Show a, Typeable b, Binary b, Show b) => a -> Action b -> Action b Source #
Cache an action, given a key and an Action
. Each call in your program should specify a different
key, but the key should remain consistent between runs. Ideally, the Action
will gather all its dependencies
with tracked operations, e.g. 'readFile''. However, if information is accessed from the environment
(e.g. the action is a closure), you should call cacheActionWith
being explicit about what is captured.
cacheActionWith :: (Typeable a, Binary a, Show a, Typeable b, Binary b, Show b, Typeable c, Binary c, Show c) => a -> b -> Action c -> Action c Source #
Like cacheAction
, but also specify which information is captured by the closure of the Action
. If that
information changes, the Action
will be rerun.