Convert pylint output to simplified CodeClimate JSON suitable for Gitlab Code Quality checks

pylint_to_gitlab_codeclimate.py

import hashlib
import html
import json
import sys
from typing import Optional

from pylint.reporters import JSONReporter
from pylint.lint import Run

# map pylint categories to CodeClimate severity
PYLINT_CATEGORY_TO_SEVERITY = {
    "fatal": "blocker",
    "error": "critical",
    "warning": "major",
    "refactor": "minor",
    "convention": "minor",
}


class GitlabCodeClimateReporter(JSONReporter):
    """
    Custom pylint reporter to convert pylint messages into a reduced CodeClimate JSON report
    suitable for Gitlab's Code Quality. Only reports `description`, `fingerprint`, `severity`,
    `location`.

    See:
    1. https://docs.gitlab.com/ee/user/project/merge_requests/code_quality.html#implementing-a-custom-tool.
    2. https://github.com/PyCQA/pylint/blob/master/pylint/reporters/json_reporter.py
    """

    name = "gitlabcodeclimate"

    def display_messages(self, layout: Optional["Section"]) -> None:
        """
        Convert the pylint messages into a reduced CodeClimate report dictionary and dump as JSON.
        """
        codeclimate_dict = [
            {
                "description": html.escape(f"{msg.msg_id}: {msg.msg or ''}", quote=False),
                "severity": PYLINT_CATEGORY_TO_SEVERITY[msg.category],
                "location": {"path": msg.path, "lines": {"begin": msg.line}},
                "fingerprint": hashlib.sha1(
                    (msg.symbol + msg.path + str(msg.line)).encode()
                ).hexdigest(),
            }
            for msg in self.messages
        ]

        print(json.dumps(codeclimate_dict, indent=4), file=self.out)


if __name__ == "__main__":
    Run(sys.argv[1:], reporter=GitlabCodeClimateReporter())

Gitlab's Code Quality documentation is all over the place and using the CodeClimate docker images seems to make things more complicated than it needs to be. The Code Climate Pylint Engine repository (which is what the image uses) gave me a useful starting point but it puts out list of JSON dicts with each issue null-terminated separated (I'm not sure why) and doesn't give a fingerprint.

As specified in the Implementing a custom tool section of the docs Gitlab only needs a small subset of the full specification of the codeclimate fields: description, fingerprint, location.path, location.lines.begin.

After poking around it seems we can get a very lightweight pylint to Gitlab code quality JSON format by subclassing the pylint JSONReporter class and then overriding only the handle_message method to convert the pylint message.

You can then just dump the report to file in your Gitlab CI with python pylint-to-codeclimate.py my_python_package > gl-code-quality-report.json and add the file as a artifacts:reports:codequality:

static-analysis:
  script:
    - python pylint-to-codeclimate.py my_python_package > gl-code-quality-report.json
  artifacts:
    reports:
      codequality: gl-code-quality-report.json

I'd be delighted if you can use it @ahogen and I don't mind at all. Thanks for the heads-up about the severity field. I actually moved off Gitlab for the past six months because of work reasons but I'm back to Gitlab now and I'll have a chance to add the severity in the next couple weeks.

Added severity field information with an arbitrary mapping of pylint categories to CodeClimate severity.

Hi, thanks for the code. I am trying it out. Python 3.7.7 and PyLint 2.11.1, I get the following errors

(dev_venv_py3) C:\Users\nyue\projects\gitlab_python_cicd>python pylint_to_gitlab_codeclimate.py src > gl-code-quality-report.json
Traceback (most recent call last):
  File "pylint_to_gitlab_codeclimate.py", line 67, in <module>
    Run(sys.argv[1:], reporter=GitlabCodeClimateReporter())
  File "C:\Users\nyue\projects\gitlab_python_cicd\dev_venv_py3\lib\site-packages\pylint\lint\run.py", line 375, in __init__
    score_value = linter.generate_reports()
  File "C:\Users\nyue\projects\gitlab_python_cicd\dev_venv_py3\lib\site-packages\pylint\lint\pylinter.py", line 1251, in generate_reports
    self.reporter.display_messages(report_nodes.Section())
  File "C:\Users\nyue\projects\gitlab_python_cicd\dev_venv_py3\lib\site-packages\pylint\reporters\json_reporter.py", line 47, in display_messages
    for msg in self.messages
  File "C:\Users\nyue\projects\gitlab_python_cicd\dev_venv_py3\lib\site-packages\pylint\reporters\json_reporter.py", line 47, in <listcomp>
    for msg in self.messages
AttributeError: 'dict' object has no attribute 'category'

@nyue I had the same issue with Python 3.6.8 and pylint 2.12.2. I found I had to override the handle_message method from the base class, since it was looking for the original JSON output fields as attributes, and not the new, differently-named ones in a dictionary. My version is here: https://gist.github.com/quanterium/53a277cab4136963374cf703dfe399bd

Personally I ended up using pylint-gitlab. It basically worked out of the box, for me. Their README explains usage well.

Even though it stated Python >=3.7, it seems to work fine in 3.6.8.

Thanks for the reports and suggestions @nyue, @quanterium and @ahogen. pylint-gitlab indeed looks like the easiest option but for a lightweight option I've updated the gist so that it now works with Python 3.10 and pylint==2.12.2. As @nyue noted it's easiest now to just override display_messages because this commit makes it now assume messages contains a list of Message.