Shelly

A module for shell-like / perl-like programming in Haskell. Shelly's focus is entirely on ease of use for those coming from shell scripting. However, it also tries to use modern libraries and techniques to keep things efficient.

The functionality provided by this module is (unlike standard Haskell filesystem functionality) thread-safe: each Sh maintains its own environment and its own working directory.

I highly recommend putting the following at the top of your program, otherwise you will likely need either type annotations or type conversions

 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE ExtendedDefaultRules #-}
 {-# OPTIONS_GHC -fno-warn-type-defaults #-}
 import Shelly
 import Data.Text.Lazy as LT
 default (LT.Text)

Entering Sh.

data Sh a

Instances

Monad Sh

Functor Sh

Applicative Sh

MonadIO Sh

ShellCommand (Sh ())

(~ * s Text, Show s) => ShellCommand (Sh s)

ShellCommand (Sh Text)

MonadReader (IORef State) Sh

type ShIO a = Sh a

ShIO is Deprecated in favor of Sh, which is easier to type.

shelly :: MonadIO m => Sh a -> m a

Enter a Sh from (Monad)IO. The environment and working directories are inherited from the current process-wide values. Any subsequent changes in processwide working directory or environment are not reflected in the running Sh.

shellyNoDir :: MonadIO m => Sh a -> m a

Using this entry point does not create a .shelly directory in the case of failure. Instead it logs directly into the standard error stream (stderr).

sub :: Sh a -> Sh a

Enter a sub-Sh that inherits the environment The original state will be restored when the sub-Sh completes. Exceptions are propagated normally.

silently :: Sh a -> Sh a

Create a sub-Sh in which external command outputs are not echoed and commands are not printed. See sub.

verbosely :: Sh a -> Sh a

Create a sub-Sh in which external command outputs are echoed and Executed commands are printed See sub.

escaping :: Bool -> Sh a -> Sh a

Create a sub-Sh with shell character escaping on or off. Defaults to True. Setting to False allows for shell wildcard such as * to be expanded by the shell along with any other special shell characters.

print_stdout :: Bool -> Sh a -> Sh a

Create a sub-Sh with stdout printing on or off Defaults to True.

print_commands :: Bool -> Sh a -> Sh a

Create a sub-Sh with command echoing on or off Defaults to False, set to True by verbosely

tracing :: Bool -> Sh a -> Sh a

Create a sub-Sh where commands are not traced Defaults to True. You should only set to False temporarily for very specific reasons

errExit :: Bool -> Sh a -> Sh a

named after bash -e errexit. Defaults to True. When True, throw an exception on a non-zero exit code. When False, ignore a non-zero exit code. Not recommended to set to False unless you are specifically checking the error code with lastExitCode.

Running external commands.

run :: FilePath -> [Text] -> Sh Text

Execute an external command. Takes the command name (no shell allowed, just a name of something that can be found via PATH; FIXME: setenv'd PATH is not taken into account when finding the exe name)

stdout and stderr are collected. The stdout is returned as a result of run, and complete stderr output is available after the fact using lastStderr

All of the stdout output will be loaded into memory You can avoid this but still consume the result by using run_, If you want to avoid the memory and need to process the output then use runFoldLines.

run_ :: FilePath -> [Text] -> Sh ()

the same as run, but return () instead of the stdout content stdout will be read and discarded line-by-line

runFoldLines :: a -> FoldCallback a -> FilePath -> [Text] -> Sh a

used by run. fold over stdout line-by-line as it is read to avoid keeping it in memory stderr is still being placed in memory under the assumption it is always relatively small

cmd :: ShellCommand result => FilePath -> result

variadic argument version of run. The syntax is more convenient, but more importantly it also allows the use of a FilePath as a command argument. So an argument can be a Text or a FilePath. a FilePath is converted to Text with toTextIgnore. You will need to add the following to your module:

 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE ExtendedDefaultRules #-}
 {-# OPTIONS_GHC -fno-warn-type-defaults #-}
 import Shelly
 import Data.Text.Lazy as LT
 default (LT.Text)

(-|-) :: Sh Text -> Sh b -> Sh b

Pipe operator. set the stdout the first command as the stdin of the second. This does not create a shell-level pipe, but hopefully it will in the future. To create a shell level pipe you can always set escaping False and use a pipe | character in a command.

lastStderr :: Sh Text

The output of last external command. See run.

setStdin :: Text -> Sh ()

set the stdin to be used and cleared by the next run.

lastExitCode :: Sh Int

The exit code from the last command. Unless you set errExit to False you won't get a chance to use this: a non-zero exit code will throw an exception.

command :: FilePath -> [Text] -> [Text] -> Sh Text

bind some arguments to run for re-use. Example:

 monit = command "monit" ["-c", "monitrc"]

command_ :: FilePath -> [Text] -> [Text] -> Sh ()

bind some arguments to run_ for re-use. Example:

 monit_ = command_ "monit" ["-c", "monitrc"]

command1 :: FilePath -> [Text] -> Text -> [Text] -> Sh Text

bind some arguments to run for re-use, and expect 1 argument. Example:

 git = command1 "git" []; git "pull" ["origin", "master"]

command1_ :: FilePath -> [Text] -> Text -> [Text] -> Sh ()

bind some arguments to run for re-use, and expect 1 argument. Example:

 git_ = command1_ "git" []; git+ "pull" ["origin", "master"]

sshPairs :: Text -> [(FilePath, [Text])] -> Sh Text

run commands over SSH. An ssh executable is expected in your path. Commands are in the same form as run, but given as pairs

 sshPairs "server-name" [("cd", "dir"), ("rm",["-r","dir2"])]

This interface is crude, but it works for now.

Please note this sets escaping to False: the commands will not be shell escaped. Internally the list of commands are combined with the string && before given to ssh.

sshPairs_ :: Text -> [(FilePath, [Text])] -> Sh ()

same as sshPairs, but returns ()

Modifying and querying environment.

setenv :: Text -> Text -> Sh ()

Set an environment variable. The environment is maintained in Sh internally, and is passed to any external commands to be executed.

get_env :: Text -> Sh (Maybe Text)

Fetch the current value of an environment variable. if non-existant or empty text, will be Nothing

get_env_text :: Text -> Sh Text

Fetch the current value of an environment variable. Both empty and non-existent variables give empty string as a result.

getenv :: Text -> Sh Text

Deprecated: use get_env or get_env_text

deprecated

get_env_def :: Text -> Text -> Sh Text

Deprecated: use fromMaybe DEFAULT get_env

Fetch the current value of an environment variable. Both empty and non-existent variables give the default Text value as a result

appendToPath :: FilePath -> Sh ()

add the filepath onto the PATH env variable FIXME: only effects the PATH once the process is ran, as per comments in which TODO: use cross-platform searchPathSeparator

Environment directory

cd :: FilePath -> Sh ()

Change current working directory of Sh. This does *not* change the working directory of the process we are running it. Instead, Sh keeps track of its own working directory and builds absolute paths internally instead of passing down relative paths. This may have performance repercussions if you are doing hundreds of thousands of filesystem operations. You will want to handle these issues differently in those cases.

chdir :: FilePath -> Sh a -> Sh a

cd, execute a Sh action in the new directory and then pop back to the original directory

pwd :: Sh FilePath

Obtain the current (Sh) working directory.

Printing

echo :: Text -> Sh ()

Echo text to standard (error, when using _err variants) output. The _n variants do not print a final newline.

echo_n :: Text -> Sh ()

Echo text to standard (error, when using _err variants) output. The _n variants do not print a final newline.

echo_err :: Text -> Sh ()

Echo text to standard (error, when using _err variants) output. The _n variants do not print a final newline.

echo_n_err :: Text -> Sh ()

Echo text to standard (error, when using _err variants) output. The _n variants do not print a final newline.

inspect :: Show s => s -> Sh ()

a print lifted into Sh

inspect_err :: Show s => s -> Sh ()

a print lifted into Sh using stderr

tag :: Sh a -> Text -> Sh a

same as trace, but use it combinator style

trace :: Text -> Sh ()

internally log what occurred. Log will be re-played on failure.

show_command :: FilePath -> [Text] -> Text

Querying filesystem.

ls :: FilePath -> Sh [FilePath]

List directory contents. Does *not* include "." and "..", but it does include (other) hidden files.

lsT :: FilePath -> Sh [Text]

Get back [Text] instead of [FilePath]

test_e :: FilePath -> Sh Bool

Does a path point to an existing filesystem object?

test_f :: FilePath -> Sh Bool

Does a path point to an existing file?

test_d :: FilePath -> Sh Bool

Does a path point to an existing directory?

test_s :: FilePath -> Sh Bool

Does a path point to a symlink?

which :: FilePath -> Sh (Maybe FilePath)

Get a full path to an executable on PATH, if exists. FIXME does not respect setenv'd environment and uses findExecutable which uses the PATH inherited from the process environment. FIXME: findExecutable does not maintain a hash of existing commands and does a ton of file stats

Filename helpers

absPath :: FilePath -> Sh FilePath

Make a relative path absolute by combining with the working directory. An absolute path is returned as is. To create a relative path, use relPath.

(</>) :: (ToFilePath filepath1, ToFilePath filepath2) => filepath1 -> filepath2 -> FilePath

uses System.FilePath.CurrentOS, but can automatically convert a Text

(<.>) :: ToFilePath filepath => filepath -> Text -> FilePath

uses System.FilePath.CurrentOS, but can automatically convert a Text

canonic :: FilePath -> Sh FilePath

makes an absolute path. Like canonicalize, but on an exception returns absPath

canonicalize :: FilePath -> Sh FilePath

Obtain a (reasonably) canonic file path to a filesystem object. Based on canonicalizePath in system-fileio.

relPath :: FilePath -> Sh FilePath

Makes a relative path relative to the current Sh working directory. An absolute path is returned as is. To create an absolute path, use absPath

relativeTo

Arguments

:: FilePath

anchor path, the prefix

-> FilePath

make this relative to anchor path

-> Sh FilePath

make the second path relative to the first Uses stripPrefix, but will canonicalize the paths if necessary

path :: FilePath -> Sh FilePath

deprecated

hasExt :: Text -> FilePath -> Bool

flipped hasExtension for Text

Manipulating filesystem.

mv :: FilePath -> FilePath -> Sh ()

Currently a renameFile wrapper. TODO: Support cross-filesystem move. TODO: Support directory paths in the second parameter, like in cp.

rm :: FilePath -> Sh ()

Remove a file. Does fail if the file does not exist (use rm_f instead) or is not a file.

rm_f :: FilePath -> Sh ()

Remove a file. Does not fail if the file does not exist. Does fail if the file is not a file.

rm_rf :: FilePath -> Sh ()

A swiss army cannon for removing things. Actually this goes farther than a normal rm -rf, as it will circumvent permission problems for the files we own. Use carefully. Uses removeTree

cp :: FilePath -> FilePath -> Sh ()

Copy a file. The second path could be a directory, in which case the original file name is used, in that directory.

cp_r :: FilePath -> FilePath -> Sh ()

Copy a file, or a directory recursively.

mkdir :: FilePath -> Sh ()

Create a new directory (fails if the directory exists).

mkdir_p :: FilePath -> Sh ()

Create a new directory, including parents (succeeds if the directory already exists).

reading/writing Files

readfile :: FilePath -> Sh Text

(Strictly) read file into a Text. All other functions use Lazy Text. Internally this reads a file as strict text and then converts it to lazy text, which is inefficient

readBinary :: FilePath -> Sh ByteString

wraps ByteSting readFile

writefile :: FilePath -> Text -> Sh ()

Write a Lazy Text to a file.

appendfile :: FilePath -> Text -> Sh ()

Append a Lazy Text to a file.

touchfile :: FilePath -> Sh ()

Update a file, creating (a blank file) if it does not exist.

withTmpDir :: (FilePath -> Sh a) -> Sh a

Create a temporary directory and pass it as a parameter to a Sh computation. The directory is nuked afterwards.

exiting the program

exit :: Int -> Sh ()

exit 0 means no errors, all other codes are error conditions

errorExit :: Text -> Sh ()

echo a message and exit with status 1

quietExit :: Int -> Sh ()

for exiting with status > 0 without printing debug information

terror :: Text -> Sh a

fail that takes a Text

Exceptions

catchany :: IO a -> (SomeException -> IO a) -> IO a

A helper to catch any exception (same as ... catch (e :: SomeException) -> ...).

catch_sh :: Exception e => Sh a -> (e -> Sh a) -> Sh a

Same as a normal catch but specialized for the Sh monad.

finally_sh :: Sh a -> Sh b -> Sh a

Same as a normal finally but specialized for the Sh monad.

data ShellyHandler a

You need to wrap exception handlers with this when using catches_sh.

Constructors

forall e . Exception e => ShellyHandler (e -> Sh a)

catches_sh :: Sh a -> [ShellyHandler a] -> Sh a

Same as a normal catches, but specialized for the Sh monad.

catchany_sh :: Sh a -> (SomeException -> Sh a) -> Sh a

Catch an exception in the Sh monad.

convert between Text and FilePath

toTextIgnore :: FilePath -> Text

silently uses the Right or Left value of Filesystem.Path.CurrentOS.toText

toTextWarn :: FilePath -> Sh Text

fromText :: Text -> FilePath

Utilities.

(<$>) :: Functor f => (a -> b) -> f a -> f b

An infix synonym for fmap.

(<$$>) :: Functor m => (b -> c) -> (a -> m b) -> a -> m c

A functor-lifting function composition.

whenM :: Monad m => m Bool -> m () -> m ()

A monadic-conditional version of the when guard.

unlessM :: Monad m => m Bool -> m () -> m ()

A monadic-conditional version of the unless guard.

time :: Sh a -> Sh (Double, a)

Run a Sh computation and collect timing information.

Re-exported for your convenience

liftIO :: MonadIO m => forall a. IO a -> m a

Lift a computation from the IO monad.

when :: Monad m => Bool -> m () -> m ()

Conditional execution of monadic expressions. For example,

       when debug (putStr "Debugging\n")

will output the string Debugging\n if the Boolean value debug is True, and otherwise do nothing.

unless :: Monad m => Bool -> m () -> m ()

The reverse of when.

data FilePath

Instances

Eq FilePath

Data FilePath

Ord FilePath

Show FilePath

Typeable FilePath

IsString FilePath

Monoid FilePath

NFData FilePath

ToFilePath FilePath

ShellArg FilePath

ShellArg FilePath

internal functions for writing extensions

get :: Sh State

put :: State -> Sh ()

find functions

find :: FilePath -> Sh [FilePath]

List directory recursively (like the POSIX utility find). listing is relative if the path given is relative. If you want to filter out some results or fold over them you can do that with the returned files. A more efficient approach is to use one of the other find functions.

findWhen :: (FilePath -> Sh Bool) -> FilePath -> Sh [FilePath]

find that filters the found files as it finds. Files must satisfy the given filter to be returned in the result.

findFold :: (a -> FilePath -> Sh a) -> a -> FilePath -> Sh a

Fold an arbitrary folding function over files froma a find. Like findWhen but use a more general fold rather than a filter.

findDirFilter :: (FilePath -> Sh Bool) -> FilePath -> Sh [FilePath]

find that filters out directories as it finds Filtering out directories can make a find much more efficient by avoiding entire trees of files.

findDirFilterWhen

Arguments

:: (FilePath -> Sh Bool)

directory filter

-> (FilePath -> Sh Bool)

file filter

-> FilePath

directory

-> Sh [FilePath]

similar findWhen, but also filter out directories Alternatively, similar to findDirFilter, but also filter out files Filtering out directories makes the find much more efficient

findFoldDirFilter :: (a -> FilePath -> Sh a) -> a -> (FilePath -> Sh Bool) -> FilePath -> Sh a

like findDirFilterWhen but use a folding function rather than a filter The most general finder: you likely want a more specific one