Safe Haskell | None |
---|---|
Language | Haskell2010 |
This module provides a simplified interface. If you want more, see Test.Tasty.Silver.Advanced.
Note about filenames. They are looked up in the usual way, thus relative
names are relative to the processes current working directory.
It is common to run tests from the package's root directory (via cabal
test
or cabal install --enable-tests
), so if your test files are under
the tests/
subdirectory, your relative file names should start with
tests/
(even if your test.hs
is itself under tests/
, too).
Note about line endings. The best way to avoid headaches with line endings (when running tests both on UNIX and Windows) is to treat your golden files as binary, even when they are actually textual.
This means:
- When writing output files from Haskell code, open them in binary mode
(see
openBinaryFile
,withBinaryFile
andhSetBinaryMode
). This will disable automatic\n -> \r\n
conversion on Windows. For convenience, this module exportswriteBinaryFile
which is just likewriteFile
but opens the file in binary mode. When usingByteString
s note that Data.ByteString and Data.ByteString.Lazy use binary mode forwriteFile
, while Data.ByteString.Char8 and Data.ByteString.Lazy.Char8 use text mode. - Tell your VCS not to do any newline conversion for golden files. For
git check in a
.gitattributes
file with the following contents (assuming your golden files have.golden
extension):
*.golden -text
On its side, tasty-golden reads and writes files in binary mode, too.
Why not let Haskell/git do automatic conversion on Windows? Well, for
instance, tar
will not do the conversion for you when unpacking a release
tarball, so when you run cabal install your-package --enable-tests
, the
tests will be broken.
As a last resort, you can strip all \r
s from both arguments in your
comparison function when necessary. But most of the time treating the files
as binary does the job.
- goldenVsFile :: TestName -> FilePath -> FilePath -> IO () -> TestTree
- goldenVsProg :: TestName -> FilePath -> FilePath -> [String] -> Text -> TestTree
- goldenVsAction :: TestName -> FilePath -> IO a -> (a -> Text) -> TestTree
- printProcResult :: (ExitCode, Text, Text) -> Text
- findByExtension :: [FilePath] -> FilePath -> IO [FilePath]
Documentation
:: TestName | test name |
-> FilePath | path to the «golden» file (the file that contains correct output) |
-> FilePath | path to the output file |
-> IO () | action that creates the output file |
-> TestTree | the test verifies that the output file contents is the same as the golden file contents |
Compare a given file contents against the golden file contents. Assumes that both text files are utf8 encoded.
:: TestName | test name |
-> FilePath | path to the golden file |
-> FilePath | executable to run. |
-> [String] | arguments to pass. |
-> Text | stdin |
-> TestTree |
Compares a given file with the output (exit code, stdout, stderr) of a program. Assumes that the program output is utf8 encoded.
:: TestName | test name |
-> FilePath | path to the «golden» file (the file that contains correct output) |
-> IO a | action that returns a text-like value. |
-> (a -> Text) | Converts a value to it's textual representation. |
-> TestTree | the test verifies that the returned textual representation is the same as the golden file contents |
Compare something text-like against the golden file contents. For the conversion of inputs to text you may want to use the Data.Text.Encoding or/and System.Process.Text modules.
printProcResult :: (ExitCode, Text, Text) -> Text Source #
Converts the output of a process produced by e.g. System.Process.Text to a textual representation. Stdout/stderr are written seperately, any ordering relation between the two streams is lost in the translation.
Find all files in the given directory and its subdirectories that have the given extensions.
It is typically used to find all test files and produce a golden test per test file.
The returned paths use forward slashes to separate path components, even on Windows. Thus if the file name ends up in a golden file, it will not differ when run on another platform.
The semantics of extensions is the same as in takeExtension
. In
particular, non-empty extensions should have the form ".ext"
.
This function may throw any exception that getDirectoryContents
may
throw.
It doesn't do anything special to handle symlinks (in particular, it probably won't work on symlink loops).
Nor is it optimized to work with huge directory trees (you'd probably want to use some form of coroutines for that).