vim-jp / vimdoc-en / testing

testing - Vim Documentation

Return to main English | 日本語
testing.txt   For Vim version 8.2.  Last change: 2021 Apr 02

                  VIM REFERENCE MANUAL    by Bram Moolenaar

Testing Vim and Vim script                      testing-support

Expression evaluation is explained in eval.txt.  This file goes into details
about writing tests in Vim script.  This can be used for testing Vim itself
and for testing plugins.

1. Testing Vim                          testing
2. Test functions                       test-functions-details
3. Assert functions                     assert-functions-details

1. Testing Vim                                          testing

Vim can be tested after building it, usually with "make test".
The tests are located in the directory "src/testdir".

There are two types of tests added over time:               oldest, only for tiny and small builds
        test_something.vim      new style tests

New tests should be added as new style tests.  The test scripts are named
test_<feature>.vim (replace <feature> with the feature under test). These use
functions such as assert_equal() to keep the test commands and the expected
result in one place.
These tests are used only for testing Vim without the +eval feature.

Find more information in the file src/testdir/README.txt.

2. Test functions                               test-functions-details

test_alloc_fail({id}{countdown}{repeat})            test_alloc_fail()
                This is for testing: If the memory allocation with {id} is
                called, then decrement {countdown}, and when it reaches zero
                let memory allocation fail {repeat} times.  When {repeat} is
                smaller than one it fails one time.

                Can also be used as a method:

test_autochdir()                                        test_autochdir()
                Set a flag to enable the effect of 'autochdir' before Vim
                startup has finished.

test_feedinput({string})                                test_feedinput()
                Characters in {string} are queued for processing as if they
                were typed by the user. This uses a low level input buffer.
                This function works only when with +unix or GUI is running.

                Can also be used as a method:

test_garbagecollect_now()                        test_garbagecollect_now()
                Like garbagecollect(), but executed right away.  This must
                only be called directly to avoid any structure to exist
                internally, and v:testing must have been set before calling
                any function.

test_garbagecollect_soon()                       test_garbagecollect_soon()
                Set the flag to call the garbagecollector as if in the main
                loop.  Only to be used in tests.

test_getvalue({name})                                   test_getvalue()
                Get the value of an internal variable.  These values for
                {name} are supported:

                Can also be used as a method:

test_ignore_error({expr})                        test_ignore_error()
                Ignore any error containing {expr}.  A normal message is given
                This is only meant to be used in tests, where catching the
                error with try/catch cannot be used (because it skips over
                following code).
                {expr} is used literally, not as a pattern.
                When the {expr} is the string "RESET" then the list of ignored
                errors is made empty.

                Can also be used as a method:

test_null_blob()                                        test_null_blob()
                Return a Blob that is null. Only useful for testing.

test_null_channel()                                     test_null_channel()
                Return a Channel that is null. Only useful for testing.
                {only available when compiled with the +channel feature}

test_null_dict()                                        test_null_dict()
                Return a Dict that is null. Only useful for testing.

test_null_function()                                    test_null_function()
                Return a Funcref that is null. Only useful for testing.

test_null_job()                                         test_null_job()
                Return a Job that is null. Only useful for testing.
                {only available when compiled with the +job feature}

test_null_list()                                        test_null_list()
                Return a List that is null. Only useful for testing.

test_null_partial()                                     test_null_partial()
                Return a Partial that is null. Only useful for testing.

test_null_string()                                      test_null_string()
                Return a String that is null. Only useful for testing.

test_unknown()                                          test_unknown()
                Return a value with unknown type. Only useful for testing.

test_void()                                             test_void()
                Return a value with void type. Only useful for testing.

test_option_not_set({name})                             test_option_not_set()
                Reset the flag that indicates option {name} was set.  Thus it
                looks like it still has the default value. Use like this:
                        set ambiwidth=double
                        call test_option_not_set('ambiwidth')
                Now the 'ambiwidth' option behaves like it was never changed,
                even though the value is "double".
                Only to be used for testing!

                Can also be used as a method:

test_override({name}{val})                            test_override()
                Overrides certain parts of Vim's internal processing to be able
                to run tests. Only to be used for testing Vim!
                The override is enabled when {val} is non-zero and removed
                when {val} is zero.
                Current supported values for name are:

                name         effect when {val} is non-zero
                redraw       disable the redrawing() function
                redraw_flag  ignore the RedrawingDisabled flag
                char_avail   disable the char_avail() function
                starting     reset the "starting" variable, see below
                nfa_fail     makes the NFA regexp engine fail to force a
                             fallback to the old engine
                no_query_mouse  do not query the mouse position for "dec"
                no_wait_return  set the "no_wait_return" flag.  Not restored
                                with "ALL".
                ui_delay     time in msec to use in ui_delay(); overrules a
                             wait time of up to 3 seconds for messages
                term_props   reset all terminal properties when the version
                             string is detected
                uptime       overrules sysinfo.uptime
                ALL          clear all overrides ({val} is not used)

                "starting" is to be used when a test should behave like
                startup was done.  Since the tests are run by sourcing a
                script the "starting" variable is non-zero. This is usually a
                good thing (tests run faster), but sometimes changes behavior
                in a way that the test doesn't work properly.
                When using:
                        call test_override('starting', 1)
                The value of "starting" is saved.  It is restored by:
                        call test_override('starting', 0)

                Can also be used as a method:
                        GetOverrideVal()-> test_override('starting')

test_refcount({expr})                                   test_refcount()
                Return the reference count of {expr}.  When {expr} is of a
                type that does not have a reference count, returns -1.  Only
                to be used for testing.

                Can also be used as a method:

test_scrollbar({which}{value}{dragging})            test_scrollbar()
                Pretend using scrollbar {which} to move it to position
                {value}.  {which} can be:
                        left    Left scrollbar of the current window
                        right   Right scrollbar of the current window
                        hor     Horizontal scrollbar

                For the vertical scrollbars {value} can be 1 to the
                line-count of the buffer.  For the horizontal scrollbar the
                {value} can be between 1 and the maximum line length, assuming
                'wrap' is not set.

                When {dragging} is non-zero it's like dragging the scrollbar,
                otherwise it's like clicking in the scrollbar.
                Only works when the {which} scrollbar actually exists,
                obviously only when using the GUI.

                Can also be used as a method:
                        GetValue()->test_scrollbar('right', 0)

test_setmouse({row}{col})                             test_setmouse()
                Set the mouse position to be used for the next mouse action.
                {row} and {col} are one based.
                For example:
                        call test_setmouse(4, 20)
                        call feedkeys("\<LeftMouse>", "xt")

test_settime({expr})                                    test_settime()
                Set the time Vim uses internally.  Currently only used for
                timestamps in the history, as they are used in viminfo, and
                for undo.
                Using a value of 1 makes Vim not sleep after a warning or
                error message.
                {expr} must evaluate to a number.  When the value is zero the
                normal behavior is restored.

                Can also be used as a method:

test_srand_seed([seed])                                 test_srand_seed()
                When [seed] is given this sets the seed value used by
                srand().  When omitted the test seed is removed.

3. Assert functions                             assert-functions-details

assert_beeps({cmd})                                     assert_beeps()
                Run {cmd} and add an error message to v:errors if it does
                NOT produce a beep or visual bell.
                Also see assert_fails()assert_nobeep() and

                Can also be used as a method:

assert_equal({expected}{actual} [, {msg}])
                When {expected} and {actual} are not equal an error message is
                added to v:errors and 1 is returned.  Otherwise zero is
                returned assert-return.
                There is no automatic conversion, the String "4" is different
                from the Number 4.  And the number 4 is different from the
                Float 4.0.  The value of 'ignorecase' is not used here, case
                always matters.
                When {msg} is omitted an error in the form "Expected
                {expected} but got {actual}" is produced.
        assert_equal('foo', 'bar')
                Will result in a string to be added to v:errors:
        test.vim line 12: Expected 'foo' but got 'bar'

                Can also be used as a method, the base is passed as the
                second argument:
                        mylist->assert_equal([1, 2, 3])

assert_equalfile({fname-one}{fname-two} [, {msg}])
                When the files {fname-one} and {fname-two} do not contain
                exactly the same text an error message is added to v:errors.
                Also see assert-return.
                When {fname-one} or {fname-two} does not exist the error will
                mention that.
                Mainly useful with terminal-diff.

                Can also be used as a method:

assert_exception({error} [, {msg}])                     assert_exception()
                When v:exception does not contain the string {error} an error
                message is added to v:errors.  Also see assert-return.
                This can be used to assert that a command throws an exception.
                Using the error number, followed by a colon, avoids problems
                with translations:
                          call assert_false(1, 'command should have failed')
                          call assert_exception('E492:')

assert_fails({cmd} [, {error} [, {msg} [, {lnum} [, {context}]]]])
                Run {cmd} and add an error message to v:errors if it does
                NOT produce an error or when {error} is not found in the
                error message.  Also see assert-return.

                When {error} is a string it must be found literally in the
                first reported error. Most often this will be the error code,
                including the colon, e.g. "E123:".
                        assert_fails('bad cmd', 'E987:')

                When {error} is a List with one or two strings, these are
                used as patterns.  The first pattern is matched against the
                first reported error:
                        assert_fails('cmd', ['E987:.*expected bool'])
                The second pattern, if present, is matched against the last
                reported error.
                If there is only one error then both patterns must match. This
                can be used to check that there is only one error.
                To only match the last error use an empty string for the first
                        assert_fails('cmd', ['', 'E987:'])

                If {msg} is empty then it is not used.  Do this to get the
                default message when passing the {lnum} argument.

                When {lnum} is present and not negative, and the {error}
                argument is present and matches, then this is compared with
                the line number at which the error was reported. That can be
                the line number in a function or in a script.

                When {context} is present it is used as a pattern and matched
                against the context (script name or function name) where
                {lnum} is located in.

                Note that beeping is not considered an error, and some failing
                commands only beep.  Use assert_beeps() for those.

                Can also be used as a method:

assert_false({actual} [, {msg}])                        assert_false()
                When {actual} is not false an error message is added to
                v:errors, like with assert_equal().
                Also see assert-return.
                A value is false when it is zero. When {actual} is not a
                number the assert fails.
                When {msg} is omitted an error in the form
                "Expected False but got {actual}" is produced.

                Can also be used as a method:

assert_inrange({lower}{upper}{actual} [, {msg}])     assert_inrange()
                This asserts number and Float values.  When {actual}  is lower
                than {lower} or higher than {upper} an error message is added
                to v:errors.  Also see assert-return.
                When {msg} is omitted an error in the form
                "Expected range {lower} - {upper}, but got {actual}" is

assert_match({pattern}{actual} [, {msg}])
                When {pattern} does not match {actual} an error message is
                added to v:errors.  Also see assert-return.

                {pattern} is used as with =~: The matching is always done
                like 'magic' was set and 'cpoptions' is empty, no matter what
                the actual value of 'magic' or 'cpoptions' is.

                {actual} is used as a string, automatic conversion applies.
                Use "^" and "$" to match with the start and end of the text.
                Use both to match the whole text.

                When {msg} is omitted an error in the form
                "Pattern {pattern} does not match {actual}" is produced.
        assert_match('^f.*o$', 'foobar')
                Will result in a string to be added to v:errors:
        test.vim line 12: Pattern '^f.*o$' does not match 'foobar'

                Can also be used as a method:

assert_nobeep({cmd})                                    assert_nobeep()
                Run {cmd} and add an error message to v:errors if it
                produces a beep or visual bell.
                Also see assert_beeps().

                Can also be used as a method:

assert_notequal({expected}{actual} [, {msg}])
                The opposite of assert_equal(): add an error message to
                v:errors when {expected} and {actual} are equal.
                Also see assert-return.

                Can also be used as a method:
                        mylist->assert_notequal([1, 2, 3])

assert_notmatch({pattern}{actual} [, {msg}])
                The opposite of assert_match(): add an error message to
                v:errors when {pattern} matches {actual}.
                Also see assert-return.

                Can also be used as a method:

assert_report({msg})                                    assert_report()
                Report a test failure directly, using {msg}.
                Always returns one.

                Can also be used as a method:

assert_true({actual} [, {msg}])                         assert_true()
                When {actual} is not true an error message is added to
                v:errors, like with assert_equal().
                Also see assert-return.
                A value is TRUE when it is a non-zero number.  When {actual}
                is not a number the assert fails.
                When {msg} is omitted an error in the form "Expected True but
                got {actual}" is produced.

                Can also be used as a method: