Skip to content

Latest commit

 

History

History
706 lines (511 loc) · 27.1 KB

Contributing.md

File metadata and controls

706 lines (511 loc) · 27.1 KB

Guidelines for contributing to EDMC

Work on Issues

If you are not part of the core development team then you should only be performing work that addresses an open issue.

So, if what you think needs doing isn't currently referred to in an open issue, then you should first open an issue. Please use the correct template if applicable.

Check with us first

Whilst we welcome all efforts to improve the program it's best to ensure that you're not duplicating, or worse, wasting effort.

There is sometimes a misconception that Open Source means that the primary project is obliged to accept Pull Requests. That is not so. While you are 100% free to make changes in your own fork, we will only accept changes that are consistent with our vision for EDMC. Fundamental changes in particular need to be agreed in advance.


Text formatting

The project contains an .editorconfig file at its root. Please either ensure your editor is taking note of those settings, or cross-check its contents with the editorconfig documentation , and ensure your editor/IDE's settings match.


General workflow

  1. You will need a GitHub account.
  2. Fork the repository on GitHub into your account there (hereafter referred to as 'your fork').
  3. In your local copy of your fork create an appropriate WIP branch.
  4. Develop the changes, testing as you go (no we don't have any actual tests yet).
    1. Be as sure as you can that the code works as you intend and hasn't introduced any other bugs or regressions.
    2. Test the codebase as a whole against any unit tests that do exist, and add your own as you can.
    3. Check your code against flake8 periodically.
  5. When you're sure the work is final:
    1. Push your WIP branch to your fork (you probably should have been doing this as you worked as a form of backup).
    2. Access the WIP branch on your fork on GitHub and create a Pull Request. Mention any Issue number(s) that it addresses.
  6. Await feedback in the form of comments on the Pull Request.

IMPORTANT: Once you have created the Pull Request any changes you make to that WIP branch and push to your fork will be reflected in the Pull Request. Ensure that only the changes for the issue(s) you are addressing are in the WIP branch. Any other work should occur in its own separate WIP branch. If needs be make one branch to work in and another for the Pull Request, merging or cherry-picking commits as needed.


Git commit conventions

  • Please use the standard Git convention of a short title in the first line and fuller body text in subsequent lines.
  • Please reference issue numbers using the "hashtag" format #123 in your commit message wherever possible. This lets GitHub create two-way hyperlinks between the issue report and the commit. Certain text in a PR that fixes an issue can auto-close the issue when the PR is merged. Note the caveats about the extended forms being necessary in some situations.
  • If in doubt, lean towards many small commits. This makes git bisect much more useful.
  • Please try at all costs to avoid a "mixed-up" commit, i.e. one that addresses more than one issue at once. One thing at a time is best.

Git branch structure and tag conventions

Somewhat based on git-flow, but our particular take on it:

Branches

stable

This will either have HEAD pointing to the latest stable release code or might have extra code merged in for a hotfix that will shortly be in the next stable release. If you want the latest stable release code then use the appropriate Release/A.B.C tag!

beta

If we run any pre-release betas with actual builds released, not just a branch to be run from source, then this branch will contain that code. As per stable above, this branch might be ahead of the latest pre-release due to merging of hotfixes. Use the appropriate tag if you want to be sure of the code you checkout. If there hasn't yet been a new beta version this could be far behind all of: main, develop, stable.

develop

This is the branch where all current development is integrated. No commits should be made directly to this as the work should be done in a separate branch used in a Pull Request before being merged as part of resolving that Pull Request.

main

Yes, we've renamed this from master. See "Using 'main' as the primary branch in Git" for instructions on ensuring you're cleanly using it in any local clone.

This branch should contain anything from develop that is considered well tested and ready for the next stable merge.

master

This is no longer used. If the branch is even present then it's no longer updated. You should be using main instead.

releases

Currently the version of the edmarketconnector.xml 'appcast' file in this branch is what live clients check to be notified of new versions. This can potentially be replaced with the stable branch's version, but some care will be necessary to ensure no users are left behind (their client checking the releases branch which then no longer exists). For the time being this should always be kept in sync with stable as each new release is made.

Work in progress conventions

Remember, you should always be working versus a single issue, even if the work is part of a Milestone or Project. There might be cases where issues aren't duplicates, but your work still addresses more than one. In that case pick one for the naming scheme below, but mention all in commit messages and the Pull Request.

In all cases the branch should be named as per the scheme <class>/<issue number>/<title>:

  • <class> - We have several classes of WIP branch:

    • fix - For working on bug fixes, e.g. fix/184/crash-in-startup
    • enhancement - For enhancing an existing feature, e.g. enhancement/192/add-thing-to-wotsit
    • feature - For working on new features, e.g. feature/284/allow-users-to-frob
  • <issue-number> is for easy reference when citing the issue number in commit messages. If you're somehow doing work that's not versus an issue then don't put the <issue number>- part in.

  • <title> is intended to allow anyone to quickly know what the branch is addressing. Try to choose something succinct for <title>, it's just there for easy reference, it doesn't need to be the entire title of the appropriate issue.

The branch you base your work on will depend on which class of WIP it is. If you're fixing a bug in the latest stable then it's best to base your branch on its HEAD. If it's a fix for a beta release then base off of beta's HEAD. If you're working on a new feature then you'd want to base the work on develop's HEAD.

Important: Please under no circumstance merge from the source branch after you have started work in your WIP branch. If there are any non-trivial conflicts when we merge your Pull Request then we might ask you to rebase your WIP branch on the latest version of the source branch. Otherwise, we'll work out how to best merge your changes via comments in the Pull Request.

Tags

Stable Releases

All stable releases MUST have a tag of the form Release/Major.Minor.Patch on the commit that was HEAD when the installer for it was built.

Pre-Releases

Tags for pre-releases should be of one of two forms, following Version Strings conventions.

  • Initial beta releases should have versions of the form:

    Major.Minor.Patch-beta<serial>

    with the <serial> starting with 1 and incrementing with each new beta pre-release.

  • Release candidates should have versions of the form:

    Major.Minor.Patch-rc<serial>

    with the <serial> starting with 1 and incrementing with each new release candidate.

The tag should thus be Release/Major.Minor.Patch-(beta|rc)<serial>.

The Semantic Versioning +<build metadata> should never be a part of the tag.


Version conventions

Please see Version Strings for a description of the currently used version strings.

Historically a A.BC form was used, based on an internal A.B.C.D version string. This was changed to simply A.B.C.D throughout for 4.0.0.0, 4.0.1.0 and 4.0.2.0. It would also continue for any other increment of only the 'C' (Patch) component.

Going forwards we will always use the full Semantic Version and 'folder style' tag names, e.g. Release/Major.Minor.Patch.

Currently, the only file that defines the version code-wise is config/__init__.py. Changelog.md and edmarketconnector.xml are another matter handled as part of the release process.


Python Environment

Whilst you can use whatever IDE/development environment best suits you, much of the setup in this project has only been tested against PyCharm or VSCode, along with 'git bash' command-line.

Use the version denoted by .python-version

We only test, and build with, the python version as defined in the file .python-version. Trying to use any other version might mean things just don't run at all, or don't work as expected.

Use a Python virtual environment

Always use a Python virtual environment specific to working on this project.

An example, when using Python 3.11.x would be:

python -m venv ../edmc-venv-3.11

Note how the 'venv' is placed in a sub-directory of the parent directory of the project. This avoids any issues with scripts working recursively picking up your 'venv' files.

If you have good reason to put the 'venv' inside the project directory then you MUST use either venv or .venv, else you'll run into all sorts of problems with pre-commit checks.

Install the development requirements

Whilst simply running the project only requires pip install -r requirements.txt-provided modules, development work will instead require:

pip install -r requirements-dev.txt

NB: This itself will also take note of requirements.txt.

This will ensure you have all the necessary tools to hand for the pre-commit checks.

Set up pre-commit

In order to have any submitted PR be in the least-worse shape when first opened you MUST run the checks as specified in .pre-commit-config.yaml.

pre-commit install --install-hooks

Now whenever you git commit the various checks will be run to ensure your code is compliant with our requirements. If you have a temporary need to bypass this (e.g. wanting to commit one change and fix a non-compliant file later) you can add -n to the git commit arguments.

NB: There is a problem with pre-commit if using VSCode. There's a workaround in one of the comments.

Consider running pytest before any git push

The GitHub workflows for PRs and pushes will run pytest and flag an error if the tests don't pass, so it's in your interests to ensure you've not broken any tests.

You could endeavour to remember to run pytest manually, or you could add this git hook:

.git/hooks/pre-push

#!/bin/sh

# If this script exits with a non-zero status nothing will be pushed.
#
# This hook is called with the following parameters:
#
# $1 -- Name of the remote to which the push is being done
# $2 -- URL to which the push is being done
#
# If pushing without using a named remote those arguments will be equal.
#
# Information about the commits which are being pushed is supplied as lines to
# the standard input in the form:
#
#   <local ref> <local sha1> <remote ref> <remote sha1>

remote="$1"
url="$2"

echo "-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-"
echo " Running pytest..."
pytest || exit 1
echo " All tests passed, proceeding..."
echo "-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-"

exit 0

It's probably overkill, and will become painful if enough tests are added, to run pytest in a pre-commit hook.


Linting

We use flake8 for linting all python source.

While working on your changes, please ensure that they pass a check from flake8 using our configuration and plugins. If you installed requirements-dev.txt with pip, you should simply be able to run flake8 your_files_here to lint your files.

Note that if your PR does not cleanly (or mostly cleanly) pass a linting scan, your PR may be put on hold pending fixes.

Unit testing

Where possible please write unit tests for your PRs, especially in the case of bug fixes, having regression tests help ensure that we don't accidentally re-introduce a bug down the line.

We use the pytest for unit testing.

The files for a test should go in a sub-directory of tests/ named after the (principal) file or directory that contains the code they are testing. For example:

  • Tests for journal_lock.py are in tests/journal_lock.py/test_journal_lock.py. The test_ prefix on test_journal_lock.py is necessary in order for pytest to recognise the file as containing tests to be run.
  • Tests for config/ code are located in tests/config/test_config.py, not tests/config.py/test_config.py

The sub-directory avoids having a mess of files in tests, particularly when there might be supporting files, e.g. tests/config/_old_config.py or files containing test data.

Invoking just a bare pytest command will run all tests.

To run only a sub-set of tests you can use, e.g. pytest -k journal_lock. You might want to use pytest -rA -k journal_lock if you have any debug print() statements within the test code itself, so you can see the output even when the tests all succeed.

Adding --trace to a pytest invocation causes it to drop into a pdb prompt for each test, handy if you want to step through the testing code to be sure of anything.

Otherwise, see the pytest documentation.

Test Coverage

As we work towards actually having tests for as much of the code as possible it is useful to monitor the current test coverage.

Running pytest will also produce the overall coverage report, see the configured options in pyproject.toml.

One issue you might run into is where there is code that only runs on one platform. By default pytest-cov/coverage will count this code as not tested when run on a different platform. We utilise the coverage-conditional-plugin module so that #pragma comments can be used to give hints to coverage about this.

The pragmas are defined in the tool.coverage.coverage_conditional_plugin.rules section of pyproject.toml, e.g.

[tool.coverage.coverage_conditional_plugin.rules]
sys-platform-win32 = "sys_platform != 'win32'"
...

And are used as in:

import sys

if sys.platform == 'win32':  # pragma: sys-platform-win32
    ...
else:  # pragma: sys-platform-not-win32
    ...

Note the inverted sense of the pragma definitions, as the comments cause coverage to not consider that code block on this platform.

As of 2022-10-02 and coverage-conditional-plugin==0.7.0 there is no way to signal that an entire file should be excluded from coverage reporting on the current platform. See this GitHub issue comment .


Imports used only in core plugins

Because the 'core' plugins, as with any EDMarketConnector plugin, are only ever loaded dynamically, not through an explicit import statement, there is no way for py2exe to know about them when building the contents of the dist.win32 directory. See docs/Releasing.md for more information about this build process.

Thus, you MUST check if any imports you add in plugins/*.py files are only referenced in that file (or also only in any other core plugin), and if so YOU MUST ENSURE THAT PERTINENT ADJUSTMENTS ARE MADE IN build.py IN ORDER TO ENSURE THE FILES ARE ACTUALLY PRESENT IN AN END-USER INSTALLATION ON WINDOWS.

An exmaple is that as of 2022-02-01 it was noticed that plugins/eddn.py now uses util/text.py, and is the only code to do so. py2exe does not detect this and thus the resulting dist.win32/library.zip does not contain the util/ directory, let alone the util/text.py file. The fix was to update the appropriate packages definition to:

            'packages': [
                'sqlite3',  # Included for plugins
                'util',  # 2022-02-01 only imported in plugins/eddn.py
            ],

Note that in this case it's in packages because we want the whole directory adding. For a single file an extra item in includes would suffice.

Such additions to build.py should not cause any issues if subsequent project changes cause py2exe to automatically pick up the same file(s).


Debugging network sends

Rather than risk sending bad data to a remote service, even if only through repeatedly sending the same data you can cause such code to instead send through a local web server and thence to a log file.

  1. This utilises the --debug-sender ... command-line argument. The argument to this is free-form, so there's nothing to edit in EDMarketConnector.py in order to support a new target for this.
  2. The debug web server is set up globally in EDMarketConnector.py.
  3. In code where you want to utilise this you will need at least something like this (taken from some plugins/edsm.py code):
from config import debug_senders
from edmc_data import DEBUG_WEBSERVER_HOST, DEBUG_WEBSERVER_PORT

TARGET_URL = 'https://www.edsm.net/api-journal-v1'
if 'edsm' in debug_senders:
  TARGET_URL = f'http://{DEBUG_WEBSERVER_HOST}:{DEBUG_WEBSERVER_PORT}/edsm'

...
r = this.requests_session.post(TARGET_URL, data=data, timeout=_TIMEOUT)

Be sure to set a URL path in the TARGET_URL that denotes where the data would normally be sent to. 4. The output will go into a file in %TEMP%\EDMarketConnector\http_debug whose name is based on the path component of the URL. In the code example above it will come out as edsm.log due to how TARGET_URL is set.


Coding Conventions

In general, please follow PEP8

Adhere to the spelling conventions of the libraries and modules used in the project.

Yes, this means using 'color' rather than 'colour', and in general will mean US, not British, spellings.


Control flow

Never oneline any control flow (if, else, for), as it makes spotting what happens next difficult.

Yes:

if something_true:
    one_thing_we_do()

No:

if something_true: one_thing_we_do()

Yes, some existing code still flouts this rule.

Scope changes

Always use Line breaks after scope changes. It makes seeing where scope has changed far easier on a quick skim

Yes:

  if True:
    do_something()

  else:
    raise UniverseBrokenException()

  return

No:

  if True:
    do_something()
  else:
    raise UniverseBrokenException()
  return

Use Type hints

Please do place type hints on the declarations of your functions, both their arguments and return types.


Use logging not print(), and definitely not sys.stdout.write()

EDMarketConnector.py sets up a logging.Logger for this under the appname, so:

import logging
from config import appname
logger = logging.getLogger(appname)

logger.info(f'Some message with a {variable}')

try:
    something
except Exception as e:  # Try to be more specific
    logger.error(f'Error in ... with ...', exc_info=e)

DO NOT use the following, as you might cause a circular import:

    from EDMarketConnector import logger

Setting up logging in plugins is slightly different.

We have implemented a logging.Filter that adds support for the following in logging.Formatter() strings:

  1. %(qualname)s which gets the full <module>.ClassA(.ClassB...).func of the calling function.
  2. %(class)s which gets just the enclosing class name(s) of the calling function.

If you want to see how we did this, check EDMCLogging.py.

So don't worry about adding anything about the class or function you're logging from, it's taken care of.

Do use a pertinent message, even when using exc_info=... to log an exception. e.g. Logging will know you were in your get_foo() function but you should still tell it what actually (failed to have) happened in there.

Use the appropriate logging level

You must ensure necessary information is always in the log files, but not so much that it becomes more difficult to discern important information when diagnosing an issue.

logging, and thus our logger instances provide functions of the following names:

  • info - For general messages that don't occur too often outside of startup and shutdown.
  • warning - An error has been detected, but it doesn't impact continuing functionality. In particular use this when logging errors from external services. This would include where we detected a known issue with Frontier-supplied data. A currently unknown issue may end up triggering logging at error level or above.
  • error - An error in our code has occurred. The application might be able to continue, but we want to make it obvious there's a bug that we need to fix.
  • critical - An error has occurred in our code that impacts the continuation of the current process.
  • debug - Information about code flow and data that is occurs too often to be at info level. Keep in mind our default logging level is DEBUG, but users can change it for the plain log file, but the debug log giles are always at least at DEBUG level.

In addition to that we utilise one of the user-defined levels as:

  • trace - This is a custom log level intended for debug messages which occur even more often and would cause too much log output for even 'normal' debug use. In general only developers will set this log level, but we do supply a command-line argument and .bat file for users to enable it. It cannot be selected from Settings in the UI.

    Do not use a bare logger.trace(...) call unless you're 100% certain it's only temporary and will be removed before any code merge. In that case you would utilise EDMarketConnector.py --trace to see the output.

    Instead, you should gate any TRACE logging using the trace_if() helper method provided on logger:

      logger.trace_if('journal.event.scan', 'my-log-message')

    The string used to identify this tracing should be related to the function of the code, not the particular file, or class, that it is in. This is so that the same string can be used to trace code that spans more than one file, class, or other scope.

    This would then be triggered by running EDMarketConnector with the appropriate command-line arguments:

    EDMarketConnector.py --trace-on journal.event.scan
    

    Note that you do not also need to specify --trace, that's implied.

    This way you can set up TRACE logging that won't spam just because --trace is used.


Use fstrings, not modulo-formatting or .format

fstrings are new in python 3.6, and allow for string interpolation rather than more opaque formatting calls.

As part of our flake8 linting setup we have included a linter that warns when you use % on string literals.

.format() won't throw flake8 errors, but only because it's still the best way to handle untranslated words in otherwise translated phrases. Thus, we allow this, and only this, use of .format() for strings.


Docstrings

Doc strings are preferred on all new modules, functions, classes, and methods, as they help others understand your code. We use the sphinx formatting style, which for pycharm users is the default.

Lack of docstrings, or them not passing some checks, will cause a flake8 failure in our setup.


Comments

LANG comments for translations

When adding translations you must add a LANG comment.

Mark hacks and workarounds with a specific comment

We often write hacks or workarounds to make EDMC work on a given version or around a specific bug. Please mark all hacks, workarounds, magic with one of the following comments, where applicable:

# HACK $elite-version-number | $date: $description
# MAGIC $elite-version-number | $date: $description
# WORKAROUND $elite-version-number | $date: $description

The description should cover exactly why the hack is needed, what it does, what is required / expected for it to be removed. Please be verbose here, more info about weird choices is always prefered over magic that we struggle to understand in six months.

Additionally, if your hack is over around 5 lines, please include a # HACK END or similar comment to indicate the end of the hack.

Use sys.platform for platform guards

mypy (and pylance) understand platform guards and will show unreachable code / resolve imports correctly for platform specific things. However, this only works if you directly reference sys.platform, importantly the following does not work:

from sys import platform
if platform == 'win32':
  ...

It MUST be if sys.platform.


Build process

See Releasing.md for the environment and procedure necessary for building the application into a .exe and Windows installer file.


Translations

See Translations.md for how to ensure any new phrases your code adds can be easily translated.


Acknowledgement

The overall structure, and some of the contents, of this document were taken from the EDDI Contributing.md.