Skip to content

Latest commit

 

History

History
331 lines (268 loc) · 12.6 KB

CONTRIBUTING.md

File metadata and controls

331 lines (268 loc) · 12.6 KB

Contributing to HADDOCK3

Welcome, we made many efforts to facilitate your contribution to this fantastic project. There are several ways to contribute:

  • you can improve tutorials and/or documentation
  • improve the code itself (maybe you even found some bug 🐛?)
  • improve error messages so they become clearer
  • add a new simulation module altogether
  • write more unittests (we dare you to do that :godmode:)

Before attempting any development, please install HADDOCK3 following the instructions in the INSTALL file. Afterwards, follow the instructions in this file.

1. Contributing new code

HADDOCK3 has two main testing workflows. Here, within the repository, we test the HADDOCK3's Python shell, code style, and package building. Our Continuous Integration (CI) pipeline is based on tox and GitHub Actions. We will explain you how to use them.

To contribute to the HADDOCK3's Python shell, follow these steps:

  1. Fork the HADDOCK3 repository

  2. Create a new branch in your fork (git checkout -b <new_branch_name>)

  3. Develop your code and tests:

    1. HADDOCK3 source is in src/haddock. Always implement code in the lowest Python version supported (this case is 3.9).
    2. Tests sit in the tests/ folder. Use pytest for testing.
    3. See more details on how to contribute with code and tests in the subheadings below.
  4. While you are developing (or when you think you have finished), you can (should) use our tox environments to test your code. Use the following commands from the main repository folder:

    1. tox -e test runs tests in current python version. If you tox to report test names and status for every single test (high verbosity) use tox -e test -- -vv.
    2. tox -e lint shows you errors in the code style.
    3. tox -e build simulates building the HADDOCK3 package.
    4. Run the above altogether with the simple tox command
    5. If you want to submit high-quality code, you may wish to run tox -e radon to assess your code cyclomatic complexity.
  5. You can work on these tox tests until they all pass green before submitting your PR.

  6. Check if your contribution fulfills the requeriments proposed in the PR template, these are based on the experience of some developers to ensure the long-term survival of the codebase. Suggestions are always welcomed.

  7. We also have an examples folder with test cases that you can run (should) to ensure the integrity of the python shell as a whole:

    1. Navigate to any of the examples folder and run the -test.cfg file (see USAGE).
    2. *-test.cfg runs are meant for testing purposes only. The examples/ folder also contain *-full.cfg files for production runs. You don't need to run these for testin.
    3. If you have a powerful computer and want to run all tests in a row, navigate to the examples folder and run python run_tests.py -b. The -b flag ensures the run will stop asap an error is found.
    4. if your computer is not powerful enough, you can ask us to run those tests once you submitted the pull request.
    5. For example, the docking-protein-protein-test.cfg run in less than 3 minutes on a Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz laptop using 7 cores. All -test.cfg together take about 45 minutes.
  8. Add a list of your new additions to the CHANGELOG.md file by adding a new sub-header as described bellow. This is mandatory for tox -e build to pass. Note this applies only after we have released the stable 3.0.0 version.

# Changelog

## new_version

* your change one
* your change two

(... the rest of the file ...)
  1. If you have difficulties with tox, let us know. These tox tests are the same that run on the GitHub Actions once you send the PR. So, sending the PR is another way to ensure all tests pass.

  2. Submit your PR if you haven't done so yet 😉

    1. GitHub allows submitting "Draft PRs". You can use this option to let us know you are working on some new good stuff so we can help you from the beginning.

1.1 Contributing with code (additional details)

You should also know that... HADDOCK3 folders, files, and code follow a well defined structure with very specific patterns. Inside each source folder you will (likely) find a README file describing the structure of the folder and presenting guidelines on how to better contribute to that folder and respective files. Summarizing here:

  • New command-line clients go in the clis/ folder. See how the current clients here created and use that as a template. Remember to update also the setup.py file.
  • Add any new functions that you foresee are general and could serve different places in the code to the libs/ folder. Find a lib*.py files that would serve your purpuses, otherwise create a new one.
  • Any plug-in like functionality, for example "check if input is correct", should go into its own python module inside the gear/ folder. Any variables, functions, and classes related only to that implementation should go all inside the same module (Python file).
  • Any general hard definitions, or physical constants, should go in the core/ folder.
  • If you want to implement a new HADDOCK3 module, navigate to the modules/ folder and follow the instructions in the README file there. You will see that all folders and files follow a pattern.
  • Talk with us before developing any CNS related part.

1.2 Contributing with tests (additional details)

Inside the tests/ folder you will find several test_*.py files. Normally, each file has the tests for each *.py file in the source. If you create new *.py files you should create a new test file of the same name, test_new_file_name.py. Aim at 100% test coverage for the code you have created. Write tests according to pytest. You can see examples in our test_*.py files. You can run the tests using the tox -e test commands explained above. Or, if you want to run the tests for a singly file use tox -e test -- tests/test_myfile.py.

1.3 Dependencies

HADDOCK3 is highly interconnected with other projects. Many of them use HADDOCK3 core functionalities. Therefore, we aim to keep HADDOCK3 with the lowest dependency footprint possible. Avoid adding dependencies when developing new functionalities. How?

  1. Use the Python standard library as much as possible.
  2. Numpy is always allowed.
  3. If you need to implement heavy calculations, it is best you use Numba instead of C libraries. Talk with us before.
  4. You need a small function from a large library. Try to reimplement it yourself with the Python standard library.
    1. If you can't, talk first with us by opening an issue.
  5. You need a big and complex function or maybe a whole python file from another project. If licenses are compatible, copy their code to the HADDOCK3 repository writing all necessary headers to grant authorship and comply with license requirements.
    1. If licenses aren't compatible, talk is us. We may have an alternative.
  6. Your new module largely depends on a library and reimplementing or copying is not an option. Then, consider if we can use that dependency as a runtime dependency (like gdock or lightdock) instead of an installation dependency.
  7. Nothing of the above is possible. You really need an install dependency. Talk with us before.

1.4 Code style

HADDOCK3 follows nice code style rules. These are almost hard rules, but there is some room for exceptions - common sense prevails. When developing code, run tox -e lint to inspect if your code follows our conventions. We use flake8 with the following rules and docstrings follow numpydoc style. You can hold on our code for style. Nevertheless, here's a dummy code snippet for your reference. Line have a soft max of 80 chars.

"""Module's docstring."""
# any comments. You can add here licensing stuff
# first import standard library
import os
from path import Pathlib

# import third library
import numpy as np

# import from haddock3
from haddock import log
# multiline import
from haddock.libs.libio import (
    lib1,
    lib2,
    lib3,
    lib4,
    )


# all this are possible
GLOBAL_VARIBLE = None
_hidden_global_variable = None

avoid_lower_case_globals = None


def my_nice_func(arg1, arg2, arg3):
    """Docstring here."""
    # do stuf

# if the function has many arguments, separate them by new lines.
# mind the identations
def my_nice_func_with_big_name(
        var1,
        var2,
        var3,
        named_var1="something",
        named_var2="else",
        **kwargs,
        ):
    """Docstring here."""
    # do stuff
    return


multiline_list = [
    var1,
    var2,
    var3,
    var4,
    ]

multiline_dict = {
    "key1": 1,
    "key2": 2,
    "key3": 3,
    "key4": 4,
    "key5": 5,
    }

# separate comprehensions by sections if line don't fit 80 chars
# same for dictionaries
cool_list_comprehension = [
    dome_some_long_stuff(i)
    for i in some_iterable
    if validate(i)
    ]

double_quotes_allowed = "string"
single_quotes_allowed = 'string'
# don't try to homogenize quotes all around the code, respect the
# original input.

# define first use later
generator_with_a_large_name = zip(
    iter1,
    iter2,
    iter3,
    iter4,
    iter5,
    )

for a, b, c, d, e in generator_with_a_large_name:
    # do stuff

# use noqa: E501 for slightly long lines, in case splitting into
# multiple lines reduces readability
this_is_a_long_line = that_makes_no_much_sense(to_separate, in_multiple_lines, abcd)  # noqa: E501

1.5 Code structure

Write code in the form of small functions, because small functions are easier to test. Combine small functions to compose larger functions. If you need to use a global variable in a function, pass it as a default value of a parameter. Avoid using complex classes, or avoid using classes at all, unless you need to maintain a state or you really know what you are doing. Flat is better than nested (though it's harder to write, but it's easier to maintain and read). Use long variable names if needed. Write comments that explain why you do stuff, and not how you do stuff. Use the TODO: flag in your comments to note something for the future. If needed, raise an issue.

1.6 Creating a new module

To develop a new HADDOCK3 module follow our guidelines and templates under src/haddock/modules/_template_cat/_template_mod/.

2. Contributing with documentation

You may contribute to HADDOCK3 documentation by improving parts where documentation is lacking or writing the documentation for the new code you propose. HADDOCK3 documentation is hosted online at https://bonvinlab.org/haddock3.

HADDOCK3 documentation is rendered with Sphinx combining markdown files, restructured text files, and extracting the docstrings in the source code.

The docs/ folder contains all the files used by Sphinx to compile the documentation to HTML files. To incorporate new documentation pages or update the existing ones, navigate around the docs/ folder to learn how we have structured it and add/edit the files you find relevant. You will see that the structure of folders and files follows the design of the documentation website.

You can render the documentation locally to inspect the end result before creating a pull request. To compile the documentation locally: activate the haddock3 python environment inside the haddock3 github folder, run tox -e docs and then use your favourite browser to open the haddock3-docs/index.html file. We invite you to read through Sphinx-doc webpage if you want to exploit any advanced feature of Sphinx, but we already provide examples for virtually any use you may need.

If you need to install any additional library, talk to us first. The documentation requirements are in the devtools folder.

Finally, if you find the need to generate new pages during the HTML compilation part, you may follow the devtools/build_defaults_rst.py as an example. See also the docs/conf.py file setup(app) line.

Troubleshooting:

  1. If you add any new dependency (import statement) in the code, you need to add that library name to the mock list in the docs/conf.py file.