Linter Bears - Advanced Feature Reference

Often linters are no easy tools. To squeeze out the last bit of functionality and efficiency, @linter provides some advanced features ready for use.

Supplying Configuration Files with generate_config

Sometimes tools require a configuration file to run. @linter supports that easily by overriding generate_config().

@linter(executable='...')
class MyBear:
    @staticmethod
    def generate_config(filename, file):
        config_file = ("value1 = 1\n"
                       "value2 = 2")
        return config_file

The string returned by this method is written into a temporary file before invoking create_arguments().

The path of the temporary configuration file can be accessed inside create_arguments() via the config_file parameter:

@linter(executable='...')
class MyBear:
    @staticmethod
    def generate_config(filename, file):
        config_file = ("value1 = 1\n"
                       "value2 = 2")
        return config_file

    @staticmethod
    def create_arguments(filename, file, config_file):
        return "--use-config", config_file

If you return None, no configuration file is generated. A common case where to explicitly return None is when you want to expose a setting for the user to use his own tool-specific config. In case the user specifies such a config file, we can avoid generating one again with generate_config to reduce I/O load.

@linter(executable='...')
class MyBear:
    @staticmethod
    def generate_config(filename, file,
                        user_config: str='',
                        setting_a: bool=False):
        if user_config:
            return None
        else:
            return 'A={}'.format(setting_a)

    def create_arguments(filename, file, config_file,
                         user_config: str=''):
        return (filename, '--use-config',
                user_config if user_config else config_file)

Note

By default, no configuration file is generated.

Custom Processing Functions with process_output

Inside @linter only a few output formats are supported. And they can’t be combined for different output streams. To specify an own output parsing/processing behaviour, process_output can be overridden.

@linter(executable='my_tool')
class MyBear:
    def process_output(self, output, filename, file):
        pass

The output variable contains the string output from the executable. Depending on how you use the use_stdout and use_stderr parameters from @linter, output can contain either a tuple or a plain string: If use_stdout and use_stderr are both True, a tuple is placed with (stdout, stderr). If only one of them is True, a string is passed (containing the output stream chosen).

Inside process_output you need to yield results according to the executable output. It is also possible to combine the built-in capabilities. There are several functions accessible with the naming scheme process_output_<output-format>.

  • process_output_regex: Extracts results using a regex.

    @linter(executable='my_tool',
            use_stdout=False,
            use_stderr=True)
    class MyBear:
        # Assuming the tool puts some issue messages into stderr.
        def process_output(self, output, filename, file):
            # output is a string, as we activated just ``use_stderr``
            map = {'info': RESULT_SEVERITY.INFO,
                   'warn': RESULT_SEVERITY.NORMAL,
                   'error': RESULT_SEVERITY.MAJOR}
            regex = "(?P<line>\d+):(?P<message>(?P<severity>[WEI]).*)"
            yield from self.process_output_regex(stderr,
                                                 filename,
                                                 file,
                                                 regex,
                                                 map)
    

    A static message to use for results instead of grabbing it from the executable output (via the message named regex group) can also be provided using the result_message parameter.

  • process_output_corrected: Extracts results (with patches) by using a corrected version of the file processed.

    @linter(executable='my_tool',
            use_stdout=True,
            use_stderr=False)
    class MyBear:
        # Assuming the tool puts a corrected version of the file into stdout.
        def process_output(self, output, filename, file):
            # output is a string, as we activated just ``use_stdout``
            yield from self.process_output_corrected(
                                stdout,
                                filename,
                                file,
                                diff_severity=RESULT_SEVERITY.NORMAL,
                                diff_distance=2)
    

    The diff_distance parameter takes the number of unchanged lines allowed in between two changed lines so they get yielded as a single diff. If -1 is given, every change will be yielded as an own diff.

  • process_output_unified_diff: Extracts results (with patches) by processing a unified diff.

    @linter(executable='my_tool',
            use_stdout=True,
            use_stderr=True)
    class MyBear:
        # Assuming the tool puts a unified diff into stdout
        # and additional issue messages (that can't be fixed automatically)
        # into stderr, let's combine both streams!
        def process_output(self, output, filename, file):
            # output is now a tuple, as we activated both, ``use_stdout`` and
            # ``use_stderr``.
            stdout, stderr = output
            yield from self.process_output_unified_diff(stdout,
                                                        filename,
                                                        file)
            regex = "(?P<message>.*)"
            yield from self.process_output_regex(stderr,
                                                 filename,
                                                 file,
                                                 regex)
    

JSON output is also very common:

@linter(executable='my_tool')
class MyBear:
    def process_output(self, output, filename, file):
        for issue in json.loads(output):
            yield Result.from_values(origin=self,
                                     message=issue["message"],
                                     file=filename)

Additional Prerequisite Check

@linter supports doing an additional executable check before running the bear, together with the normal one (checking if the executable exists). For example, this is useful to test for the existence of external modules (like Java modules).

To enable this additional check with your commands, use the prerequisite_check_command parameter of @linter.

@linter(executable='...'
        prerequisite_check_command=('python3', '-c', 'import my_module'))
class MyBear:
    pass

If the default error message does not suit you, you can also supply prerequisite_check_fail_message together with prerequisite_check_command.

@linter(executable='...'
        prerequisite_check_command=('python3', '-c', 'import my_module'),
        prerequisite_check_fail_message='my_module does not exist.')
class MyBear:
    pass