Distribution.Simple.Utils

Non fatal conditions that may be indicative of an error or problem.

We display these at the normal verbosity level.

Like warn, but prepend Error: … instead of Waring: … before the the message. Useful when you want to highlight the condition is an error but do not want to quit the program yet.

Useful status messages.

We display these at the normal verbosity level.

This is for the ordinary helpful status messages that users see. Just enough information to know that things are working but not floods of detail.

Display a message at normal verbosity level, but without wrapping.

Pretty-print a Doc status message at normal verbosity level. Use this if you need fancy formatting.

Display a "setup status message". Prefer using setupMessage' if possible.

More detail on the operation of some action.

We display these messages when the verbosity level is verbose

Detailed internal debugging information

We display these messages when the verbosity level is deafening

A variant of debug that doesn't perform the automatic line wrapping. Produces better output in some cases.

Perform an IO action, catching any IO exceptions and printing an error if one occurs.

Given a block of IO code that may raise an exception, annotate it with the metadata from the current scope. Use this as close to external code that raises IO exceptions as possible, since this function unconditionally wraps the error message with a trace (so it is NOT idempotent.)

Add all necessary metadata to a logging message

Wrap output with a marker if +markoutput verbosity flag is set.

NB: Why is markoutput done with start/end markers, and not prefixes? Markers are more convenient to add (if we want to add prefixes, we have to lines and then map; here's it's just some concatenates). Note that even in the prefix case, we can't guarantee that the markers are unambiguous, because some of Cabal's output comes straight from external programs, where we don't have the ability to interpose on the output.

This is used by withMetadata

Run an IO computation, returning e if it raises a "file does not exist" error.

Ignore SIGPIPE in a subcomputation.

Execute the given command with the given arguments, exiting with the same exit code if the command fails.

Execute the given command with the given arguments, returning the command's exit code.

Execute the given command with the given arguments, returning the command's exit code.

Create the process argument with proc to ensure consistent options with other rawSystem functions in this module.

Execute the given command with the given arguments, returning the command's exit code. action is executed while the command is running, and would typically be used to communicate with the process through pipes.

Create the process argument with proc to ensure consistent options with other rawSystem functions in this module.

Execute the given command with the given arguments and environment, exiting with the same exit code if the command fails.

Execute the given command with the given arguments, returning the command's output. Exits if the command exits with error.

Provides control over the binary/text mode of the output.

Execute the given command with the given arguments, returning the command's output, errors and exit code.

Optional arguments allow setting working directory, environment and command input.

Provides control over the binary/text mode of the input and output.

Execute the given command with the given arguments, returning the command's exit code.

Optional arguments allow setting working directory, environment and input and output handles.

Execute the given command with the given arguments, returning the command's exit code. action is executed while the command is running, and would typically be used to communicate with the process through pipes.

Optional arguments allow setting working directory, environment and input and output handles.

fromJust for dealing with 'Maybe Handle' values as obtained via CreatePipe. Creating a pipe using CreatePipe guarantees a Just value for the corresponding handle.

Helper to use with one of the rawSystem variants, and exit unless the command completes successfully.

Like the Unix xargs program. Useful for when we've got very long command lines that might overflow an OS limit on command line length and so you need to invoke a command multiple times to get all the args in.

Use it with either of the rawSystem variants above. For example:

xargs (32*1024) (rawSystemExit verbosity) prog fixedArgs bigArgs

Look for a program and try to find it's version number. It can accept either an absolute path or the name of a program binary, in which case we will look for the program on the path.

Same as createDirectoryIfMissing but logs at higher verbosity levels.

Copies a file without copying file permissions. The target file is created with default permissions. Any existing target file is replaced.

At higher verbosity levels it logs an info message.

Copies a bunch of files to a target directory, preserving the directory structure in the target location. The target directories are created if they do not exist.

The files are identified by a pair of base directory and a path relative to that base. It is only the relative part that is preserved in the destination.

For example:

copyFiles normal "dist/src"
   [("", "src/Foo.hs"), ("dist/build/", "src/Bar.hs")]

This would copy "src/Foo.hs" to "dist/src/src/Foo.hs" and copy "dist/build/src/Bar.hs" to "dist/src/src/Bar.hs".

This operation is not atomic. Any IO failure during the copy (including any missing source files) leaves the target in an unknown state so it is best to use it with a freshly created directory so that it can be simply deleted if anything goes wrong.

Given a relative path to a file, copy it to the given directory, preserving the relative path and creating the parent directories if needed.

Install an ordinary file. This is like a file copy but the permissions are set appropriately for an installed file. On Unix it is "-rw-r--r--" while on Windows it uses the default permissions for the target directory.

Install an executable file. This is like a file copy but the permissions are set appropriately for an installed file. On Unix it is "-rwxr-xr-x" while on Windows it uses the default permissions for the target directory.

Install a file that may or not be executable, preserving permissions.

This is like copyFiles but uses installOrdinaryFile.

This is like copyFiles but uses installExecutableFile.

This is like copyFiles but uses installMaybeExecutableFile.

This installs all the files in a directory to a target location, preserving the directory layout. All the files are assumed to be ordinary rather than executable files.

Recursively copy the contents of one directory to another path.

Like doesFileExist, but also checks that the file is executable.

The path name that represents the current directory. In Unix, it's ".", but this is system-specific. (E.g. AmigaOS uses the empty string "" for the current directory.)

Drop the extension if it's one of exeExtensions, or return the path unchanged.

List of possible executable file extensions on the current build platform.

Find a file by looking in a search path. The file path must match exactly.

Find a file by looking in a search path. The file path must match exactly.

Find a file by looking in a search path with one of a list of possible file extensions. The file base name should be given and it will be tried with each of the extensions in each element of the search path.

Like findFileWithExtension but returns which element of the search path the file was found in, and the file path relative to that base directory.

Find the file corresponding to a Haskell module name.

This is similar to findFileWithExtension' but specialised to a module name. The function fails if the file corresponding to the module is missing.

Finds the files corresponding to a list of Haskell module names.

As findModuleFile but for a list of module names.

List all the files in a directory and all subdirectories.

The order places files in sub-directories after all the files in their parent directories. The list is generated lazily so is not well defined if the source directory structure changes before the list is used.

Is this directory in the system search path?

Compare the modification times of two files to see if the first is newer than the second. The first file must exist but the second need not. The expected use case is when the second file is generated using the first. In this use case, if the result is True then the second file is out of date.

Like moreRecentFile, but also checks that the first file exists.

Advanced options for withTempFile and withTempDirectory.

Use a temporary filename that doesn't already exist.

A version of withTempFile that additionally takes a TempFileOptions argument.

Create and use a temporary directory.

Creates a new temporary directory inside the given directory, making use of the template. The temp directory is deleted after use. For example:

withTempDirectory verbosity "src" "sdist." $ \tmpDir -> do ...

The tmpDir will be a new subdirectory of the given directory, e.g. src/sdist.342.

A version of withTempDirectory that additionally takes a TempFileOptions argument.

Package description file (pkgname.cabal)

Find a package description file in the given directory. Looks for .cabal files.

Like findPackageDesc, but calls die in case of error.

Like findPackageDescCwd, but calls die in case of error.

Find auxiliary package information in the given directory. Looks for .buildinfo files.

Gets the contents of a file, but guarantee that it gets closed.

The file is read lazily but if it is not fully consumed by the action then the remaining input is truncated and the file is closed.

Writes a file atomically.

The file is either written successfully or an IO exception is raised and the original file is left unchanged.

On windows it is not possible to delete a file that is open by a process. This case will give an IO exception but the atomic property is not affected.

Write a file but only if it would have new content. If we would be writing the same as the existing content then leave the file as is so that we do not update the file's modification time.

NB: Before Cabal-3.0 the file content was assumed to be ASCII-representable. Since Cabal-3.0 the file is assumed to be UTF-8 encoded.

Same as rewriteFileEx but for ByteStrings.

Decode String from UTF8-encoded BS.ByteString

Invalid data in the UTF8 stream (this includes code-points U+D800 through U+DFFF) will be decoded as the replacement character (U+FFFD).

Variant of fromUTF8BS for lazy BS.ByteStrings

Encode String to UTF8-encoded ByteString

Code-points in the U+D800-U+DFFF range will be encoded as the replacement character (i.e. U+FFFD).

Variant of toUTF8BS for lazy BS.ByteStrings

Reads a UTF8 encoded text file as a Unicode String

Reads lazily using ordinary readFile.

Reads a UTF8 encoded text file as a Unicode String

Same behaviour as withFileContents.

Writes a Unicode String as a UTF8 encoded text file.

Uses writeFileAtomic, so provides the same guarantees.

Fix different systems silly line ending conventions

Ignore a Unicode byte order mark (BOM) at the beginning of the input

dropWhileEndLE p is equivalent to reverse . dropWhile p . reverse, but quite a bit faster. The difference between "Data.List.dropWhileEnd" and this version is that the one in Data.List is strict in elements, but spine-lazy, while this one is spine-strict but lazy in elements. That's what LE stands for - "lazy in elements".

Example:

>>> safeTail $ Data.List.dropWhileEnd (<3) [undefined, 5, 4, 3, 2, 1]*** Exception: Prelude.undefined...

>>> safeTail $ dropWhileEndLE (<3) [undefined, 5, 4, 3, 2, 1][5,4,3]

>>> take 3 $ Data.List.dropWhileEnd (<3) [5, 4, 3, 2, 1, undefined][5,4,3]

>>> take 3 $ dropWhileEndLE (<3) [5, 4, 3, 2, 1, undefined]*** Exception: Prelude.undefined...

takeWhileEndLE p is equivalent to reverse . takeWhile p . reverse, but is usually faster (as well as being easier to read).

comparing p x y = compare (p x) (p y)

Useful combinator for use in conjunction with the xxxBy family of functions from Data.List, for example:

  ... sortBy (comparing fst) ...

The isInfixOf function takes two lists and returns True iff the first list is contained, wholly and intact, anywhere within the second.

Examples

>>> isInfixOf "Haskell" "I really like Haskell."True

>>> isInfixOf "Ial" "I really like Haskell."False

For the result to be True, the first list must be finite; for the result to be False, the second list must be finite:

>>> [20..50] `isInfixOf` [0..]True

>>> [0..] `isInfixOf` [20..50]False

>>> [0..] `isInfixOf` [0..]* Hangs forever *

intercalate xs xss is equivalent to (concat (intersperse xs xss)). It inserts the list xs in between the lists in xss and concatenates the result.

Laziness

intercalate has the following properties:

>>> take 5 (intercalate undefined ("Lorem" : undefined))"Lorem"

>>> take 6 (intercalate ", " ("Lorem" : undefined))"Lorem*** Exception: Prelude.undefined

Examples

>>> intercalate ", " ["Lorem", "ipsum", "dolor"]"Lorem, ipsum, dolor"

>>> intercalate [0, 1] [[2, 3], [4, 5, 6], []][2,3,0,1,4,5,6,0,1]

>>> intercalate [1, 2, 3] [[], []][1,2,3]

Lower case string

>>> lowercase "Foobar""foobar"

Like "Data.List.union", but has O(n log n) complexity instead of O(n^2).

A right-biased version of listUnion.

Example:

>>> listUnion [1,2,3,4,3] [2,1,1][1,2,3,4,3]

>>> listUnionRight [1,2,3,4,3] [2,1,1][4,3,2,1,1]

Like nub, but has O(n log n) complexity instead of O(n^2). Code for ordNub and listUnion taken from Niklas Hambüchen's ordnub package.

Like ordNub and nubBy. Selects a key for each element and takes the nub based on that key.

A right-biased version of ordNub.

Example:

>>> ordNub [1,2,1] :: [Int][1,2]

>>> ordNubRight [1,2,1] :: [Int][2,1]

A total variant of head.

A total variant of tail.

A total variant of last.

A total variant of init.

Wraps text to the default line width. Existing newlines are preserved.

Wraps a list of words to a list of lines of words of a particular width.

isAbsoluteOnAnyPlatform and isRelativeOnAnyPlatform are like System.FilePath.isAbsolute and System.FilePath.isRelative but have platform independent heuristics. The System.FilePath exists in two versions, Windows and Posix. The two versions don't agree on what is a relative path and we don't know if we're given Windows or Posix paths. This results in false positives when running on Posix and inspecting Windows paths, like the hackage server does. System.FilePath.Posix.isAbsolute "C:\hello" == False System.FilePath.Windows.isAbsolute "/hello" == False This means that we would treat paths that start with "/" to be absolute. On Posix they are indeed absolute, while on Windows they are not.

The portable versions should be used when we might deal with paths that are from another OS than the host OS. For example, the Hackage Server deals with both Windows and Posix paths while performing the PackageDescription checks. In contrast, when we run 'cabal configure' we do expect the paths to be correct for our OS and we should not have to use the platform independent heuristics.

Append a call-site and/or call-stack based on Verbosity