Roberto is a collection of configurable development work flows. Its goal is to facilitate the development and quality assurance of some packages in the theochem organization on Github.

With a relatively simple configuration file (.roberto.yaml), the command rob will take the following steps:

1. Optionally install miniconda (and a MacOSX SDK on OSX).
2. Make a conda or venv environment for development and testing
3. Install dependencies (for the package being developed and for all development tools).
4. Build the software in-place, i.e. without installing it.
5. Run Linters (optionally showing only messages related to your changes).
6. Run unit and other tests
7. Build the documentation

When you run rob robot, a few additional steps will be performed, which are not done by default because they are slow and have a low risk of failing:

8. Upload the documentation. (disabled by default)
9. Make source and binary release packages.
10. Deploy the releases. (disabled by default)

(A few minor steps were omitted for clarity.) These steps should work on your local computer in the same way as on a continuous integration system like Github Actions, making it easy to prepare a pull request locally. It is also possible to just run a subset of these tasks, which is often needed when working on the code. Several steps will also reuse previous results (e.g. conda or venv environment) if these are already present, to speed up Roberto.

The preparation tasks (1-3) are somewhat hard-coded but they are clever enough to install a decent development environment with the correct requirements for the remaining tasks. These remaining tasks (4-10) are configurable and can be changed to work for Python and/or CMake projects.

Installation

Python 3 (>=3.7) must be installed. Other dependencies will be pulled in with the instructions below.

Roberto can be installed with conda:

conda install theochem::roberto

It can also be installed with pip. One of the following is fine, whichever you prefer:

pip install roberto
pip install roberto --user
python3 -m pip install roberto
python3 -m pip install roberto --user

On some platforms, you may have to adapt your ${PATH} variable before you can run rob.

Usage

By default, Roberto will use venv, unless it is configured to use conda. While conda is more powerful than venv, it is also a lot slower and requires more storage than venv.

When using conda, be aware that Roberto will install miniconda, by default in ~/miniconda3, if not present yet. You can modify this directory by setting the environment variable ROBERTO_CONDA_BASE_PATH or by putting the following in your global Roberto configuration file ~/.roberto.yaml:

conda:
  base_path: <your/preferred/location>

E.g. you can use this to avoid interference with an existing miniconda install.

Roberto will also make new conda or venv environments for the development of every package, with relatively long names. For example, when Roberto is executed in its own source tree, the conda environment would be roberto-dev-python-3.7.

To use Roberto, just run rob in the root of the source tree, where also the project's .roberto.yaml is located. Use rob --help to get a list of tasks if you are interested in replicating just a part of the CI process. If your ${PATH} variable is not set correctly, you can also run Roberto as python3 -m roberto instead of rob.

It is a good practice to run rob before every git commit to make sure the committed code is clean and working.

When using the cardboardlint tool and when you are working in a development branch, cardboardlint will only show linter messages for lines of code that you have changed. If you would like to see all messages, run Roberto as ROBERTO_ABSOLUTE=1 rob.

More details, e.g. on how to configure Roberto, can be found in the documentation: https://theochem.github.com/roberto

Development

If you have questions or ideas, just open an issue on Github. Practical information on how to contribute can be found in CONTRIBUTING.rst.

Roberto is intentionally a small code base, so one can easily understand how it works by reading the source code. Roberto makes extensive use of invoke to avoid having to write a lot of boiler-plate code.