datadriven

README

Data-Driven Tests for Go

This repository implements an extension of Table-Driven Testing. Instead of building and iterating over a table in the test code, the input is further separated into files (or inline strings). For certain classes of tests, this can significantly reduce the friction involved in writing and reading these tests.

Documentation

Index

• func ClearResults(path string) error
• func RunTest(t *testing.T, path string, f func(t *testing.T, d *TestData) string)
• func RunTestFromString(t *testing.T, input string, f func(t *testing.T, d *TestData) string)
• func Verbose() bool
• func Walk(t *testing.T, path string, f func(t *testing.T, path string))
• type CmdArg
  • func ParseLine(line string) (cmd string, cmdArgs []CmdArg, err error)
  • func (arg CmdArg) Scan(t *testing.T, i int, dest interface{})
  • func (arg CmdArg) String() string
• type TestData
  • func (td TestData) Fatalf(tb testing.TB, format string, args ...interface{})
  • func (td *TestData) HasArg(key string) bool
  • func (td *TestData) ScanArgs(t *testing.T, key string, dests ...interface{})

Functions

func ClearResults

func ClearResults(path string) error

func RunTest

func RunTest(t *testing.T, path string, f func(t *testing.T, d *TestData) string)

RunTest invokes a data-driven test. The test cases are contained in a separate test file and are dynamically loaded, parsed, and executed by this testing framework. By convention, test files are typically located in a sub-directory called "testdata". Each test file has the following format:

<command>[,<command>...] [arg | arg=val | arg=(val1, val2, ...)]...
<input to the command>
----
<expected results>

The command input can contain blank lines. However, by default, the expected results cannot contain blank lines. This alternate syntax allows the use of blank lines:

<command>[,<command>...] [arg | arg=val | arg=(val1, val2, ...)]...
<input to the command>
----
----
<expected results>

<more expected results>
----
----

To execute data-driven tests, pass the path of the test file as well as a function which can interpret and execute whatever commands are present in the test file. The framework invokes the function, passing it information about the test case in a TestData struct.

The function must returns the actual results of the case, which RunTest() compares with the expected results. If the two are not equal, the test is marked to fail.

Note that RunTest() creates a sub-instance of testing.T for each directive in the input file. It is thus unsafe/invalid to call e.g. Fatal() or Skip() on the parent testing.T from inside the callback function. Use the provided testing.T instance instead.

It is possible for a test to test for an "expected error" as follows:

• run the code to test
• if an error occurs, report the detail of the error as actual output.
• place the expected error details in the expected results in the input file.

It is also possible for a test to report an _unexpected_ test error by calling t.Error().

func RunTestFromString

func RunTestFromString(t *testing.T, input string, f func(t *testing.T, d *TestData) string)

RunTestFromString is a version of RunTest which takes the contents of a test directly.

func Verbose

func Verbose() bool

Verbose returns true iff -datadriven-quiet was not passed.

func Walk

func Walk(t *testing.T, path string, f func(t *testing.T, path string))

Walk goes through all the files in a subdirectory, creating subtests to match the file hierarchy; for each "leaf" file, the given function is called.

This can be used in conjunction with RunTest. For example:

 datadriven.Walk(t, path, func (t *testing.T, path string) {
   // initialize per-test state
   datadriven.RunTest(t, path, func (t *testing.T, d *datadriven.TestData) string {
    // ...
   }
 }

Files:
  testdata/typing
  testdata/logprops/scan
  testdata/logprops/select

If path is "testdata/typing", the function is called once and no subtests
are created.

If path is "testdata/logprops", the function is called two times, in
separate subtests /scan, /select.

If path is "testdata", the function is called three times, in subtest
hierarchy /typing, /logprops/scan, /logprops/select.

Types

type CmdArg

type CmdArg struct {
	Key  string
	Vals []string
}

CmdArg contains information about an argument on the directive line. An argument is specified in one of the following forms:

• argument
• argument=value
• argument=(values, ...)

func ParseLine

func ParseLine(line string) (cmd string, cmdArgs []CmdArg, err error)

ParseLine parses a line of datadriven input language and returns the parsed command and CmdArgs.

func (CmdArg) Scan

func (arg CmdArg) Scan(t *testing.T, i int, dest interface{})

Scan attempts to parse the value at index i into the dest.

func (CmdArg) String

func (arg CmdArg) String() string

type TestData

type TestData struct {
	// Pos is a file:line prefix for the input test file, suitable for
	// inclusion in logs and error messages.
	Pos string

	// Cmd is the first string on the directive line (up to the first whitespace).
	Cmd string

	// CmdArgs contains the k/v arguments to the command.
	CmdArgs []CmdArg

	// Input is the text between the first directive line and the ---- separator.
	Input string
	// Expected is the value below the ---- separator. In most cases,
	// tests need not check this, and instead return their own actual
	// output.
	// This field is provided so that a test can perform an early return
	// with "return d.Expected" to signal that nothing has changed.
	Expected string
}

TestData contains information about one data-driven test case that was parsed from the test file.

func (TestData) Fatalf

func (td TestData) Fatalf(tb testing.TB, format string, args ...interface{})

Fatalf wraps a fatal testing error with test file position information, so that it's easy to locate the source of the error.

func (*TestData) HasArg

func (td *TestData) HasArg(key string) bool

HasArg checks whether the CmdArgs array contains an entry for the given key.

func (*TestData) ScanArgs

func (td *TestData) ScanArgs(t *testing.T, key string, dests ...interface{})

ScanArgs looks up the first CmdArg matching the given key and scans it into the given destinations in order. If the arg does not exist, the number of destinations does not match that of the arguments, or a destination can not be populated from its matching value, a fatal error results. If the arg exists multiple times, the first occurrence is parsed.

For example, for a TestData originating from

cmd arg1=50 arg2=yoruba arg3=(50, 50, 50)

the following would be valid:

var i1, i2, i3, i4 int var s string td.ScanArgs(t, "arg1", &i1) td.ScanArgs(t, "arg2", &s) td.ScanArgs(t, "arg3", &i2, &i3, &i4)