Linter.py

from contextlib import contextmanager
from functools import partial, partialmethod
import inspect
from itertools import chain, compress
import re
import shutil
from subprocess import check_call, CalledProcessError, DEVNULL
from types import MappingProxyType

from coalib.bears.LocalBear import LocalBear
from coala_utils.ContextManagers import make_temp
from coala_utils.decorators import assert_right_type, enforce_signature
from coalib.misc.Shell import run_shell_command
from coalib.results.Diff import Diff
from coalib.results.Result import Result
from coalib.results.RESULT_SEVERITY import RESULT_SEVERITY
from coalib.settings.FunctionMetadata import FunctionMetadata


def _prepare_options(options):
    """
    Prepares options for ``linter`` for a given options dict in-place.

    :param options:
        The options dict that contains user/developer inputs.
    """
    allowed_options = {'executable',
                       'output_format',
                       'use_stdin',
                       'use_stdout',
                       'use_stderr',
                       'config_suffix',
                       'executable_check_fail_info',
                       'prerequisite_check_command'}

    if not options['use_stdout'] and not options['use_stderr']:
        raise ValueError('No output streams provided at all.')

    if options['output_format'] == 'corrected':
        if (
                'diff_severity' in options and
                options['diff_severity'] not in RESULT_SEVERITY.reverse):
            raise TypeError('Invalid value for `diff_severity`: ' +
                            repr(options['diff_severity']))

        if 'result_message' in options:
            assert_right_type(options['result_message'], str, 'result_message')

        if 'diff_distance' in options:
            assert_right_type(options['diff_distance'], int, 'diff_distance')

        allowed_options |= {'diff_severity', 'result_message', 'diff_distance'}
    elif options['output_format'] == 'regex':
        if 'output_regex' not in options:
            raise ValueError('`output_regex` needed when specified '
                             "output-format 'regex'.")

        options['output_regex'] = re.compile(options['output_regex'])

        # Don't setup severity_map if one is provided by user or if it's not
        # used inside the output_regex. If one is manually provided but not
        # used in the output_regex, throw an exception.
        if 'severity_map' in options:
            if 'severity' not in options['output_regex'].groupindex:
                raise ValueError('Provided `severity_map` but named group '
                                 '`severity` is not used in `output_regex`.')
            assert_right_type(options['severity_map'], dict, 'severity_map')

            for key, value in options['severity_map'].items():
                assert_right_type(key, str, 'severity_map key')

                try:
                    assert_right_type(value, int, '<severity_map dict-value>')
                except TypeError:
                    raise TypeError(
                        'The value {!r} for key {!r} inside given '
                        'severity-map is no valid severity value.'.format(
                            value, key))

                if value not in RESULT_SEVERITY.reverse:
                    raise TypeError(
                        'Invalid severity value {!r} for key {!r} inside '
                        'given severity-map.'.format(value, key))

            # Auto-convert keys to lower-case. This creates automatically a new
            # dict which prevents runtime-modifications.
            options['severity_map'] = {
                key.lower(): value
                for key, value in options['severity_map'].items()}

        if 'result_message' in options:
            assert_right_type(options['result_message'], str, 'result_message')

        allowed_options |= {'output_regex', 'severity_map', 'result_message'}
    elif options['output_format'] is not None:
        raise ValueError('Invalid `output_format` specified.')

    if options['prerequisite_check_command']:
        if 'prerequisite_check_fail_message' in options:
            assert_right_type(options['prerequisite_check_fail_message'],
                              str,
                              'prerequisite_check_fail_message')
        else:
            options['prerequisite_check_fail_message'] = (
                'Prerequisite check failed.')

        allowed_options.add('prerequisite_check_fail_message')

    # Check for illegal superfluous options.
    superfluous_options = options.keys() - allowed_options
    if superfluous_options:
        raise ValueError(
            'Invalid keyword arguments provided: ' +
            ', '.join(repr(s) for s in sorted(superfluous_options)))


def _create_linter(klass, options):
    class LinterMeta(type):

        def __repr__(cls):
            return '<{} linter class (wrapping {!r})>'.format(
                cls.__name__, options['executable'])

    class LinterBase(LocalBear, metaclass=LinterMeta):

        @staticmethod
        def generate_config(filename, file):
            """
            Generates the content of a config-file the linter-tool might need.

            The contents generated from this function are written to a
            temporary file and the path is provided inside
            ``create_arguments()``.

            By default no configuration is generated.

            You can provide additional keyword arguments and defaults. These
            will be interpreted as required settings that need to be provided
            through a coafile-section.

            :param filename:
                The name of the file currently processed.
            :param file:
                The contents of the file currently processed.
            :return:
                The config-file-contents as a string or ``None``.
            """
            return None

        @staticmethod
        def create_arguments(filename, file, config_file):
            """
            Creates the arguments for the linter.

            You can provide additional keyword arguments and defaults. These
            will be interpreted as required settings that need to be provided
            through a coafile-section.

            :param filename:
                The name of the file the linter-tool shall process.
            :param file:
                The contents of the file.
            :param config_file:
                The path of the config-file if used. ``None`` if unused.
            :return:
                A sequence of arguments to feed the linter-tool with.
            """
            raise NotImplementedError

        @staticmethod
        def get_executable():
            """
            Returns the executable of this class.

            :return:
                The executable name.
            """
            return options['executable']

        @classmethod
        def check_prerequisites(cls):
            """
            Checks whether the linter-tool the bear uses is operational.

            :return:
                True if operational, otherwise a string containing more info.
            """
            if shutil.which(cls.get_executable()) is None:
                return (repr(cls.get_executable()) + ' is not installed.' +
                        (' ' + options['executable_check_fail_info']
                         if options['executable_check_fail_info'] else
                         ''))
            else:
                if options['prerequisite_check_command']:
                    try:
                        check_call(options['prerequisite_check_command'],
                                   stdout=DEVNULL,
                                   stderr=DEVNULL)
                        return True
                    except (OSError, CalledProcessError):
                        return options['prerequisite_check_fail_message']
                return True

        @classmethod
        def _get_create_arguments_metadata(cls):
            return FunctionMetadata.from_function(
                cls.create_arguments,
                omit={'self', 'filename', 'file', 'config_file'})

        @classmethod
        def _get_generate_config_metadata(cls):
            return FunctionMetadata.from_function(
                cls.generate_config,
                omit={'filename', 'file'})

        @classmethod
        def _get_process_output_metadata(cls):
            metadata = FunctionMetadata.from_function(cls.process_output)

            if options['output_format'] is None:
                omitted = {'self', 'output', 'filename', 'file'}
            else:
                # If a specific output format is provided, function signatures
                # from process_output functions should not appear in the help.
                omitted = set(chain(metadata.non_optional_params,
                                    metadata.optional_params))

            metadata.omit = omitted
            return metadata

        @classmethod
        def get_metadata(cls):
            merged_metadata = FunctionMetadata.merge(
                cls._get_process_output_metadata(),
                cls._get_generate_config_metadata(),
                cls._get_create_arguments_metadata())
            merged_metadata.desc = inspect.getdoc(cls)
            return merged_metadata

        def _convert_output_regex_match_to_result(self,
                                                  match,
                                                  filename,
                                                  severity_map,
                                                  result_message):
            """
            Converts the matched named-groups of ``output_regex`` to an actual
            ``Result``.

            :param match:
                The regex match object.
            :param filename:
                The name of the file this match belongs to.
            :param severity_map:
                The dict to use to map the severity-match to an actual
                ``RESULT_SEVERITY``.
            :param result_message:
                The static message to use for results instead of grabbing it
                from the executable output via the ``message`` named regex
                group.
            """
            # Pre process the groups
            groups = match.groupdict()

            if 'severity' in groups:
                try:
                    groups['severity'] = severity_map[
                        groups['severity'].lower()]
                except KeyError:
                    self.warn(
                        repr(groups['severity']) + ' not found in '
                        'severity-map. Assuming `RESULT_SEVERITY.NORMAL`.')
                    groups['severity'] = RESULT_SEVERITY.NORMAL
            else:
                groups['severity'] = RESULT_SEVERITY.NORMAL

            for variable in ('line', 'column', 'end_line', 'end_column'):
                groups[variable] = (None
                                    if groups.get(variable, None) is None else
                                    int(groups[variable]))

            if 'origin' in groups:
                groups['origin'] = '{} ({})'.format(klass.__name__,
                                                    groups['origin'].strip())

            # Construct the result.
            return Result.from_values(
                origin=groups.get('origin', self),
                message=(groups.get('message', '').strip()
                         if result_message is None else result_message),
                file=filename,
                severity=groups['severity'],
                line=groups['line'],
                column=groups['column'],
                end_line=groups['end_line'],
                end_column=groups['end_column'],
                additional_info=groups.get('additional_info', '').strip())

        def process_output_corrected(self,
                                     output,
                                     filename,
                                     file,
                                     diff_severity=RESULT_SEVERITY.NORMAL,
                                     result_message='Inconsistency found.',
                                     diff_distance=1):
            """
            Processes the executable's output as a corrected file.

            :param output:
                The output of the program as a string.
            :param filename:
                The filename of the file currently being corrected.
            :param file:
                The contents of the file currently being corrected.
            :param diff_severity:
                The severity to use for generating results.
            :param result_message:
                The message to use for generating results.
            :param diff_distance:
                Number of unchanged lines that are allowed in between two
                changed lines so they get yielded as one diff. If a negative
                distance is given, every change will be yielded as an own diff,
                even if they are right beneath each other.
            :return:
                An iterator returning results containing patches for the
                file to correct.
            """
            for diff in Diff.from_string_arrays(
                file,
                output.splitlines(keepends=True)).split_diff(
                    distance=diff_distance):
                yield Result(self,
                             result_message,
                             affected_code=diff.affected_code(filename),
                             diffs={filename: diff},
                             severity=diff_severity)

        def process_output_regex(
                self, output, filename, file, output_regex,
                severity_map=MappingProxyType({
                    'critical': RESULT_SEVERITY.MAJOR,
                    'c': RESULT_SEVERITY.MAJOR,
                    'fatal': RESULT_SEVERITY.MAJOR,
                    'fail': RESULT_SEVERITY.MAJOR,
                    'f': RESULT_SEVERITY.MAJOR,
                    'error': RESULT_SEVERITY.MAJOR,
                    'err': RESULT_SEVERITY.MAJOR,
                    'e': RESULT_SEVERITY.MAJOR,
                    'warning': RESULT_SEVERITY.NORMAL,
                    'warn': RESULT_SEVERITY.NORMAL,
                    'w': RESULT_SEVERITY.NORMAL,
                    'information': RESULT_SEVERITY.INFO,
                    'info': RESULT_SEVERITY.INFO,
                    'i': RESULT_SEVERITY.INFO,
                    'note': RESULT_SEVERITY.INFO,
                    'suggestion': RESULT_SEVERITY.INFO}),
                result_message=None):
            """
            Processes the executable's output using a regex.

            :param output:
                The output of the program as a string.
            :param filename:
                The filename of the file currently being corrected.
            :param file:
                The contents of the file currently being corrected.
            :param output_regex:
                The regex to parse the output with. It should use as many
                of the following named groups (via ``(?P<name>...)``) to
                provide a good result:

                - line - The line where the issue starts.
                - column - The column where the issue starts.
                - end_line - The line where the issue ends.
                - end_column - The column where the issue ends.
                - severity - The severity of the issue.
                - message - The message of the result.
                - origin - The origin of the issue.
                - additional_info - Additional info provided by the issue.

                The groups ``line``, ``column``, ``end_line`` and
                ``end_column`` don't have to match numbers only, they can
                also match nothing, the generated ``Result`` is filled
                automatically with ``None`` then for the appropriate
                properties.
            :param severity_map:
                A dict used to map a severity string (captured from the
                ``output_regex`` with the named group ``severity``) to an
                actual ``coalib.results.RESULT_SEVERITY`` for a result.
            :param result_message:
                The static message to use for results instead of grabbing it
                from the executable output via the ``message`` named regex
                group.
            :return:
                An iterator returning results.
            """
            for match in re.finditer(output_regex, output):
                yield self._convert_output_regex_match_to_result(
                    match, filename, severity_map=severity_map,
                    result_message=result_message)

        if options['output_format'] is None:
            # Check if user supplied a `process_output` override.
            if not callable(getattr(klass, 'process_output', None)):
                raise ValueError('`process_output` not provided by given '
                                 'class {!r}.'.format(klass.__name__))
                # No need to assign to `process_output` here, the class mixing
                # below automatically does that.
        else:
            # Prevent people from accidentally defining `process_output`
            # manually, as this would implicitly override the internally
            # set-up `process_output`.
            if hasattr(klass, 'process_output'):
                raise ValueError('Found `process_output` already defined '
                                 'by class {!r}, but {!r} output-format is '
                                 'specified.'.format(klass.__name__,
                                                     options['output_format']))

            if options['output_format'] == 'corrected':
                _process_output_args = {
                    key: options[key]
                    for key in ('result_message', 'diff_severity',
                                'diff_distance')
                    if key in options}

                _processing_function = partialmethod(
                    process_output_corrected, **_process_output_args)

            else:
                assert options['output_format'] == 'regex'

                _process_output_args = {
                    key: options[key]
                    for key in ('output_regex', 'severity_map',
                                'result_message')
                    if key in options}

                _processing_function = partialmethod(
                    process_output_regex, **_process_output_args)

            def process_output(self, output, filename, file):
                """
                Processes the output of the executable and yields results
                accordingly.

                :param output:
                    The output of the executable. This can be either a string
                    or a tuple depending on the usage of ``use_stdout`` and
                    ``use_stderr`` parameters of ``@linter``. If only one of
                    these arguments is ``True``, a string is placed (containing
                    the selected output stream). If both are ``True``, a tuple
                    is placed with ``(stdout, stderr)``.
                :param filename:
                    The name of the file currently processed.
                :param file:
                    The contents of the file (line-splitted).
                """
                if isinstance(output, str):
                    output = (output,)

                for string in output:
                    yield from self._processing_function(
                        string, filename, file)

        @classmethod
        @contextmanager
        def _create_config(cls, filename, file, **kwargs):
            """
            Provides a context-manager that creates the config file if the
            user provides one and cleans it up when done with linting.

            :param filename:
                The filename of the file.
            :param file:
                The file contents.
            :param kwargs:
                Section settings passed from ``run()``.
            :return:
                A context-manager handling the config-file.
            """
            content = cls.generate_config(filename, file, **kwargs)
            if content is None:
                yield None
            else:
                with make_temp(
                        suffix=options['config_suffix']) as config_file:
                    with open(config_file, mode='w') as fl:
                        fl.write(content)
                    yield config_file

        def run(self, filename, file, **kwargs):
            # Get the **kwargs params to forward to `generate_config()`
            # (from `_create_config()`).
            generate_config_kwargs = FunctionMetadata.filter_parameters(
                self._get_generate_config_metadata(), kwargs)

            with self._create_config(
                    filename,
                    file,
                    **generate_config_kwargs) as config_file:
                # And now retrieve the **kwargs for `create_arguments()`.
                create_arguments_kwargs = (
                    FunctionMetadata.filter_parameters(
                        self._get_create_arguments_metadata(), kwargs))

                args = self.create_arguments(filename, file, config_file,
                                             **create_arguments_kwargs)

                try:
                    args = tuple(args)
                except TypeError:
                    self.err('The given arguments '
                             '{!r} are not iterable.'.format(args))
                    return

                arguments = (self.get_executable(),) + args
                self.debug("Running '{}'".format(' '.join(arguments)))

                output = run_shell_command(
                    arguments,
                    stdin=''.join(file) if options['use_stdin'] else None,
                    cwd=self.get_config_dir())

                output = tuple(compress(
                    output,
                    (options['use_stdout'], options['use_stderr'])))
                if len(output) == 1:
                    output = output[0]

                process_output_kwargs = FunctionMetadata.filter_parameters(
                    self._get_process_output_metadata(), kwargs)
                return self.process_output(output, filename, file,
                                           **process_output_kwargs)

        def __repr__(self):
            return '<{} linter object (wrapping {!r}) at {}>'.format(
                type(self).__name__, self.get_executable(), hex(id(self)))

    # Mixin the linter into the user-defined interface, otherwise
    # `create_arguments` and other methods would be overridden by the
    # default version.
    result_klass = type(klass.__name__, (klass, LinterBase), {})
    result_klass.__doc__ = klass.__doc__ or ''
    return result_klass


@enforce_signature
def linter(executable: str,
           use_stdin: bool=False,
           use_stdout: bool=True,
           use_stderr: bool=False,
           config_suffix: str='',
           executable_check_fail_info: str='',
           prerequisite_check_command: tuple=(),
           output_format: (str, None)=None,
           **options):
    """
    Decorator that creates a ``LocalBear`` that is able to process results from
    an external linter tool.

    The main functionality is achieved through the ``create_arguments()``
    function that constructs the command-line-arguments that get parsed to your
    executable.

    >>> @linter("xlint", output_format="regex", output_regex="...")
    ... class XLintBear:
    ...     @staticmethod
    ...     def create_arguments(filename, file, config_file):
    ...         return "--lint", filename

    Requiring settings is possible like in ``Bear.run()`` with supplying
    additional keyword arguments (and if needed with defaults).

    >>> @linter("xlint", output_format="regex", output_regex="...")
    ... class XLintBear:
    ...     @staticmethod
    ...     def create_arguments(filename,
    ...                          file,
    ...                          config_file,
    ...                          lintmode: str,
    ...                          enable_aggressive_lints: bool=False):
    ...         arguments = ("--lint", filename, "--mode=" + lintmode)
    ...         if enable_aggressive_lints:
    ...             arguments += ("--aggressive",)
    ...         return arguments

    Sometimes your tool requires an actual file that contains configuration.
    ``linter`` allows you to just define the contents the configuration shall
    contain via ``generate_config()`` and handles everything else for you.

    >>> @linter("xlint", output_format="regex", output_regex="...")
    ... class XLintBear:
    ...     @staticmethod
    ...     def generate_config(filename,
    ...                         file,
    ...                         lintmode,
    ...                         enable_aggressive_lints):
    ...         modestring = ("aggressive"
    ...                       if enable_aggressive_lints else
    ...                       "non-aggressive")
    ...         contents = ("<xlint>",
    ...                     "    <mode>" + lintmode + "</mode>",
    ...                     "    <aggressive>" + modestring + "</aggressive>",
    ...                     "</xlint>")
    ...         return "\\n".join(contents)
    ...
    ...     @staticmethod
    ...     def create_arguments(filename,
    ...                          file,
    ...                          config_file):
    ...         return "--lint", filename, "--config", config_file

    As you can see you don't need to copy additional keyword-arguments you
    introduced from ``create_arguments()`` to ``generate_config()`` and
    vice-versa. ``linter`` takes care of forwarding the right arguments to the
    right place, so you are able to avoid signature duplication.

    If you override ``process_output``, you have the same feature like above
    (auto-forwarding of the right arguments defined in your function
    signature).

    Note when overriding ``process_output``: Providing a single output stream
    (via ``use_stdout`` or ``use_stderr``) puts the according string attained
    from the stream into parameter ``output``, providing both output streams
    inputs a tuple with ``(stdout, stderr)``. Providing ``use_stdout=False``
    and ``use_stderr=False`` raises a ``ValueError``. By default ``use_stdout``
    is ``True`` and ``use_stderr`` is ``False``.

    Documentation:
    Bear description shall be provided at class level.
    If you document your additional parameters inside ``create_arguments``,
    ``generate_config`` and ``process_output``, beware that conflicting
    documentation between them may be overridden. Document duplicated
    parameters inside ``create_arguments`` first, then in ``generate_config``
    and after that inside ``process_output``.

    For the tutorial see:
    http://api.coala.io/en/latest/Developers/Writing_Linter_Bears.html

    :param executable:
        The linter tool.
    :param use_stdin:
        Whether the input file is sent via stdin instead of passing it over the
        command-line-interface.
    :param use_stdout:
        Whether to use the stdout output stream.
    :param use_stderr:
        Whether to use the stderr output stream.
    :param config_suffix:
        The suffix-string to append to the filename of the configuration file
        created when ``generate_config`` is supplied. Useful if your executable
        expects getting a specific file-type with specific file-ending for the
        configuration file.
    :param executable_check_fail_info:
        Information that is provided together with the fail message from the
        normal executable check. By default no additional info is printed.
    :param prerequisite_check_command:
        A custom command to check for when ``check_prerequisites`` gets
        invoked (via ``subprocess.check_call()``). Must be an ``Iterable``.
    :param prerequisite_check_fail_message:
        A custom message that gets displayed when ``check_prerequisites``
        fails while invoking ``prerequisite_check_command``. Can only be
        provided together with ``prerequisite_check_command``.
    :param output_format:
        The output format of the underlying executable. Valid values are

        - ``None``: Define your own format by overriding ``process_output``.
          Overriding ``process_output`` is then mandatory, not specifying it
          raises a ``ValueError``.
        - ``'regex'``: Parse output using a regex. See parameter
          ``output_regex``.
        - ``'corrected'``: The output is the corrected of the given file. Diffs
          are then generated to supply patches for results.

        Passing something else raises a ``ValueError``.
    :param output_regex:
        The regex expression as a string that is used to parse the output
        generated by the underlying executable. It should use as many of the
        following named groups (via ``(?P<name>...)``) to provide a good
        result:

        - line - The line where the issue starts.
        - column - The column where the issue starts.
        - end_line - The line where the issue ends.
        - end_column - The column where the issue ends.
        - severity - The severity of the issue.
        - message - The message of the result.
        - origin - The origin of the issue.
        - additional_info - Additional info provided by the issue.

        The groups ``line``, ``column``, ``end_line`` and ``end_column`` don't
        have to match numbers only, they can also match nothing, the generated
        ``Result`` is filled automatically with ``None`` then for the
        appropriate properties.

        Needs to be provided if ``output_format`` is ``'regex'``.
    :param severity_map:
        A dict used to map a severity string (captured from the
        ``output_regex`` with the named group ``severity``) to an actual
        ``coalib.results.RESULT_SEVERITY`` for a result. Severity strings are
        mapped **case-insensitive**!

        - ``RESULT_SEVERITY.MAJOR``: Mapped by ``critical``, ``c``,
          ``fatal``, ``fail``, ``f``, ``error``, ``err`` or ``e``.
        - ``RESULT_SEVERITY.NORMAL``: Mapped by ``warning``, ``warn`` or ``w``.
        - ``RESULT_SEVERITY.INFO``: Mapped by ``information``, ``info``, ``i``,
          ``note`` or ``suggestion``.

        A ``ValueError`` is raised when the named group ``severity`` is not
        used inside ``output_regex`` and this parameter is given.
    :param diff_severity:
        The severity to use for all results if ``output_format`` is
        ``'corrected'``. By default this value is
        ``coalib.results.RESULT_SEVERITY.NORMAL``. The given value needs to be
        defined inside ``coalib.results.RESULT_SEVERITY``.
    :param result_message:
        The message-string to use for all results. Can be used only together
        with ``corrected`` or ``regex`` output format. When using
        ``corrected``, the default value is ``"Inconsistency found."``, while
        for ``regex`` this static message is disabled and the message matched
        by ``output_regex`` is used instead.
    :param diff_distance:
        Number of unchanged lines that are allowed in between two changed lines
        so they get yielded as one diff if ``corrected`` output-format is
        given. If a negative distance is given, every change will be yielded as
        an own diff, even if they are right beneath each other. By default this
        value is ``1``.
    :raises ValueError:
        Raised when invalid options are supplied.
    :raises TypeError:
        Raised when incompatible types are supplied.
        See parameter documentations for allowed types.
    :return:
        A ``LocalBear`` derivation that lints code using an external tool.
    """
    options['executable'] = executable
    options['output_format'] = output_format
    options['use_stdin'] = use_stdin
    options['use_stdout'] = use_stdout
    options['use_stderr'] = use_stderr
    options['config_suffix'] = config_suffix
    options['executable_check_fail_info'] = executable_check_fail_info
    options['prerequisite_check_command'] = prerequisite_check_command

    _prepare_options(options)

    return partial(_create_linter, options=options)