Linter Bears

Welcome. This tutorial aims to show you how to use the @linter (ref) decorator in order to integrate linters in your bears.

Note

If you are planning to create a bear that does static code analysis without wrapping a tool, please refer to this link instead.

This tutorial takes you through the process of writing a local linter Bear. If you want to write a global linter Bear, for a tool that does not run once for each file, but only once for the whole project, you can still go through the steps and then read about the differences of global linter Bears at Global Linter Bears.

Why is This Useful?

A lot of programming languages already have linters implemented, so if your project uses a language that does not already have a linter Bear you might need to implement it on your own. Don’t worry, it’s easy!

What do we Need?

First of all, we need the linter executable that we are going to use. In this tutorial we will build the PylintTutorialBear so we need Pylint, a common linter for Python. It can be found here. Since it is a python package we can go ahead and install it with

$ pip3 install pylint

Writing the Bear

To write a linter bear, we need to create a class that interfaces with our linter-bear infrastructure, which is provided via the @linter decorator.

from coalib.bearlib.abstractions.Linter import linter

@linter(executable='pylint')
class PylintTutorialBear:
    pass

As you can see pylint is already provided as an executable name which gets invoked on the files you are going to lint. This is a mandatory argument for the decorator.

The linter class is only capable of processing one file at a time, for this purpose pylint or the external tool needs to be invoked every time with the appropriate parameters. This is done inside create_arguments,

@linter(executable='pylint')
class PylintTutorialBear:
    @staticmethod
    def create_arguments(filename, file, config_file):
        pass

create_arguments accepts three parameters:

  • filename: The absolute path to the file that gets processed.
  • file: The contents of the file to process, given as a list of lines (including the return character).
  • config_file: The absolute path to a config file to use. If no config file is used, this parameter is None. Processing of the config file is left to the Bear’s implementation of the method.

You can use these parameters to construct the command line arguments. The linter expects from you to return an argument sequence here. A tuple is preferred. We will do this soon for PylintTutorialBear.

Note

create_arguments doesn’t have to be a static method. In this case you also need to prepend self to the parameters in the signature. Some functionality of @linter is only available inside an instance, like logging.

def create_arguments(self, filename, file, config_file):
    self.log("Hello world")

So which are the exact command line arguments we need to provide? It depends on the output format of the linter. The @linter decorator is capable of handling different output formats:

  • regex: This parses issue messages yielded by the underlying executable.
  • corrected: Auto-generates results from a fixed/corrected file provided by the tool.
  • unified-diff: This auto-generates results from a unified-diff output provided by the executable.

In this tutorial we are going to use the regex output format. But before we continue with modifying our bear, we need to figure out how exactly output from Pylint looks like so we can parse it accordingly.

We get some promising output when invoking Pylint with

$ pylint --msg-template="L{line}C{column}: {msg_id} - {msg}" --reports=n

Sample output looks like this:

No config file found, using default configuration
************* Module coalib.bearlib.abstractions.Linter
L1C0: C0111 - Missing module docstring
L42C48: E1101 - Class 'Enum' has no 'reverse' member
L77C32: E1101 - Class 'Enum' has no 'reverse' member
L21C0: R0912 - Too many branches (16/12)
L121C28: W0613 - Unused argument 'filename'

This is something we can parse easily with a regex. So let’s implement everything we’ve found out so far:

@linter(executable='pylint',
        output_format='regex',
        output_regex=r'L(?P<line>\d+)C(?P<column>\d+): (?P<message>.*)')
class PylintTutorialBear:
    @staticmethod
    def create_arguments(filename, file, config_file):
        return ('--msg-template="L{line}C{column}: {msg_id} - {msg}"',
                '--reports=n', filename)

As you can see, the output_regex parameter consists of named groups. These are important to construct a meaningful result that contains the information that is printed out.

For the exact list of named groups @linter recognizes, see the API documentation.

For more info generally on regexes, see Python re module.

Let’s brush up our output_regex a bit to use even more information:

@linter(...
        output_regex=r'L(?P<line>\d+)C(?P<column>\d+): '
                     r'(?P<message>(?P<origin>.\d+) - .*)'),
        ...)

Now we use the issue identification as the origin so we are able to deactivate single rules via ignore statements inside code.

This class is already fully functional and allows to parse issues yielded by Pylint!

Using Severities

coala uses three types of severities that categorize the importance of a result:

  • INFO
  • NORMAL
  • MAJOR

which are defined in coalib.results.RESULT_SEVERITY. Pylint output contains severity information we can use:

L1C0: C0111 - Missing module docstring

The letter before the error code is the severity. In order to make use of the severity, we need to define it inside the output_regex parameter using the named group severity:

@linter(...
        output_regex=r'L(?P<line>\d+)C(?P<column>\d+): (?P<message>'
                     r'(?P<origin>(?P<severity>[WFECRI])\d+) - .*)',
        ...)

So we want to take up the severities denoted by the letters W, F, E, C, R or I. In order to use this severity value, we will first have to provide a map that takes the matched severity letter and maps it to a severity value of coalib.results.RESULT_SEVERITY so coala understands it. This is possible via the severity_map parameter of @linter:

from coalib.results.RESULT_SEVERITY import RESULT_SEVERITY

@linter(...
        severity_map={'W': RESULT_SEVERITY.NORMAL,
                      'F': RESULT_SEVERITY.MAJOR,
                      'E': RESULT_SEVERITY.MAJOR,
                      'C': RESULT_SEVERITY.NORMAL,
                      'R': RESULT_SEVERITY.NORMAL,
                      'I': RESULT_SEVERITY.INFO},
        ...)

coalib.results.RESULT_SEVERITY contains three different values, Info, Warning and Error you can use.

We can test our bear like this

$ coala --bear-dirs=. --bears=PylintTutorialBear --files=sample.py

Note

In order for the above command to work we should have 2 files in our current dir: PylintTutorialBear.py and our sample.py. Naming is very important in coala. coala will look for bears by their filename and display them based on their classname.

Normally, providing a severity-map is not needed, as coala has a default severity-map which recognizes many common words used for severities. Check out the API documentation for keywords supported!

Suggest Corrections Using the corrected and unified-diff Output Formats

These output formats are very simple to use and don’t require further setup from your side inside the bear:

@linter(...
        output_format='corrected')

or

@linter(...
        output_format='unified-diff')

If your underlying tool generates a corrected file or a unified-diff of the corrections, the class automatically generates patches for the changes made and yields results accordingly.

Adding Settings to our Bear

If we run

$ pylint --help

We can see that there is a --rcfile option which lets us specify a configuration file for Pylint. Let’s add that functionality to our bear.

import os

from coalib.bearlib.abstractions.Linter import linter
from coalib.results.RESULT_SEVERITY import RESULT_SEVERITY

@linter(executable='pylint',
        output_format='regex',
        output_regex=r'L(?P<line>\d+)C(?P<column>\d+): '
                     r'(?P<message>(?P<severity>[WFECRI]).*)',
        severity_map={'W': RESULT_SEVERITY.NORMAL,
                      'F': RESULT_SEVERITY.MAJOR,
                      'E': RESULT_SEVERITY.MAJOR,
                      'C': RESULT_SEVERITY.NORMAL,
                      'R': RESULT_SEVERITY.NORMAL,
                      'I': RESULT_SEVERITY.INFO})
class PylintTutorialBear:
    @staticmethod
    def create_arguments(filename, file, config_file,
                         pylint_rcfile: str=os.devnull):
        return ('--msg-template="L{line}C{column}: {msg_id} - {msg}"',
                '--reports=n', '--rcfile=' + pylint_rcfile, filename)

Just adding the needed parameter to the create_arguments signature suffices, like you would do for other bears inside run! Additional parameters are automatically queried from the coafile. Let’s also add some documentation together with the metadata attributes:

@linter(...)
class PylintTutorialBear:
    """
    Lints your Python files!

    Checks for coding standards (like well-formed variable names), detects
    semantical errors (like true implementation of declared interfaces or
    membership via type inference), duplicated code.

    See http://pylint-messages.wikidot.com/all-messages for a list of all
    checks and error codes.
    """

    @staticmethod
    def create_arguments(filename, file, config_file,
                         pylint_rcfile: str=os.devnull):
        """
        :param pylint_rcfile:
            The configuration file Pylint shall use.
        """
        ...

Note

The documentation of the param is parsed by coala and it will be used as help to the user for that specific setting.

Finished Bear

Well done, you made it this far! Now you should have built a fully functional Python linter Bear. If you followed the code from this tutorial it should look something like this

import os

from coalib.bearlib.abstractions.Linter import linter
from coalib.results.RESULT_SEVERITY import RESULT_SEVERITY

@linter(executable='pylint',
        output_format='regex',
        output_regex=r'L(?P<line>\d+)C(?P<column>\d+): '
                     r'(?P<message>(?P<severity>[WFECRI]).*)',
        severity_map={'W': RESULT_SEVERITY.NORMAL,
                      'F': RESULT_SEVERITY.MAJOR,
                      'E': RESULT_SEVERITY.MAJOR,
                      'C': RESULT_SEVERITY.NORMAL,
                      'R': RESULT_SEVERITY.NORMAL,
                      'I': RESULT_SEVERITY.INFO})
class PylintTutorialBear:
    """
    Lints your Python files!

    Checks for coding standards (like well-formed variable names), detects
    semantical errors (like true implementation of declared interfaces or
    membership via type inference), duplicated code.

    See http://pylint-messages.wikidot.com/all-messages for a list of all
    checks and error codes.

    https://pylint.org/
    """

    @staticmethod
    def create_arguments(filename, file, config_file,
                         pylint_rcfile: str=os.devnull):
        """
        :param pylint_rcfile:
            The configuration file Pylint shall use.
        """
        return ('--msg-template="L{line}C{column}: {msg_id} - {msg}"',
                '--reports=n', '--rcfile=' + pylint_rcfile, filename)

Adding Metadata Attributes

Now we need to add some more precious information to our bear. This helps by giving more information about each bear and also helps some functions gather information by using these values. Our bear now looks like:

import os

from coalib.bearlib.abstractions.Linter import linter
from dependency_management.requirements.PipRequirement import PipRequirement
from coalib.results.RESULT_SEVERITY import RESULT_SEVERITY

@linter(executable='pylint',
        output_format='regex',
        output_regex=r'L(?P<line>\d+)C(?P<column>\d+): '
                     r'(?P<message>(?P<severity>[WFECRI]).*)',
        severity_map={'W': RESULT_SEVERITY.NORMAL,
                      'F': RESULT_SEVERITY.MAJOR,
                      'E': RESULT_SEVERITY.MAJOR,
                      'C': RESULT_SEVERITY.NORMAL,
                      'R': RESULT_SEVERITY.NORMAL,
                      'I': RESULT_SEVERITY.INFO})
class PylintTutorialBear:
    """
    Lints your Python files!

    Checks for coding standards (like well-formed variable names), detects
    semantical errors (like true implementation of declared interfaces or
    membership via type inference), duplicated code.

    See http://pylint-messages.wikidot.com/all-messages for a list of all
    checks and error codes.

    https://pylint.org/
    """

    LANGUAGES = {"Python", "Python 2", "Python 3"}
    REQUIREMENTS = {PipRequirement('pylint', '1.*')}
    AUTHORS = {'The coala developers'}
    AUTHORS_EMAILS = {'[email protected]'}
    LICENSE = 'AGPL-3.0'
    CAN_DETECT = {'Unused Code', 'Formatting', 'Duplication', 'Security',
                  'Syntax'}
    SEE_MORE = 'https://pylint.org/'

    @staticmethod
    def create_arguments(filename, file, config_file,
                         pylint_rcfile: str=os.devnull):
      """
      :param pylint_rcfile:
          The configuration file Pylint shall use.
      """
      return ('--msg-template="L{line}C{column}: {msg_id} - {msg}"',
              '--reports=n', '--rcfile=' + pylint_rcfile, filename)

Running and Testing our Bear

By running

$ coala --bear-dirs=. --bears=PylintTutorialBear -B

We can see that our Bear setting is documented properly. To use coala with our Bear on sample.py we run

$ coala --bear-dirs=. --bears=PylintTutorialBear --files=sample.py

To use our pylint_rcfile setting we can do

$ coala --bear-dirs=. --bears=PythonTutorialBear \
> -S pylint_rcfile=my_rcfile --files=sample.py

You now know how to write a linter Bear and also how to use it in your project.

Congratulations!

Global Linter Bears

Some linting tools do not run on file level, i.e. once for each file, but on project level. They might check some properties of the directory structure or only check one specific file like the setup.py.

For these tools we need a GlobalBear and we can also use @linter to give us one, by passing the parameter global_bear=True:

from coalib.bearlib.abstractions.Linter import linter

@linter(executable='some_tool',
        global_bear=True,
        output_format='regex',
        output_regex=r'<filename>: <message>'')
class SomeToolBear:
    @staticmethod
    def create_arguments(config_file):
        pass

The create_arguments method takes no filename and file in this case since there is no file context. You can still make coala aware of the file an issue was detected in, by using the filename named group in your output_regex if relevant to the wrapped tool.

Where to Find More...

If you need more information about the @linter decorator, refer to the API documentation.