VERKKOLIB (name under progress)
The group library for general purpose analysis and visualization code. This repository should not contain project-specific code. Most of the code in this repo is written in Python, but also other languages are allowed. Use common sense when combining different programming languages.
Documentation, test results and code coverage: http://becs.aalto.fi/~darstr1/verkko/docs/build/html/index.html
None so far.
STRUCTURE OF THE REPOSITORY
- All the functionality related to plotting and visualization.
- All your random stuff.
- Added as needed for sub-projects.
- Personal general purpose code, without testing or backwards-compatibility requirements. Used for things too specific to be worth refining, but should be somewhere. Contains separate subdirectories for different group members. e.g., u/darst.
About code backwards-compatibility and versioning
The idea of backwards compatibility is that, once written, code should continue to work the same in the future even if more features are added. For a big shared library like this, this is a very important (and difficult) topic.
Code relating to scientific calculations should absolutely remain backwards compatible. For example, changing a default option may affect the integrity of other people's work.
All other code should remain as backwards-compatible as possible, for example plotting related code. Small changes, such as improvements to plot output formats, are generally OK but use your judgment. Other people will be relying on this!
If need for "rewriting" a function exists, the new version number should be added to the end of the name:
At this point, you can take the opportunity to change anything and make general improvements. This can be done at both the function and module level. Alternatively, you could take the opportunity to change the name to something else that is more descriptive. In the docstrings, make sure that you document what the new code replaces, any differences, and advice on which version to use (if it isn't obvious).
- These rules apply to mature code, not things that are too new to be used by many people. It's always best to design things properly in the first place, so take the time to talk to others and adjust things while they are new in order to design things as well as possible at the beginning.
In general, we try to stick to the PEP style and documentation guidelines: http://legacy.python.org/dev/peps/pep-0008/ http://legacy.python.org/dev/peps/pep-0020/ http://legacy.python.org/dev/peps/pep-0257/
For a brief summary, see the file python_style.md of this repository. Most of these styles have good reasons, so consider the styles because they make sense, not just because it's the "way".
CODE TESTING PRACTICES
All scientific code should be tested. However, reality indicates that tests slowly evolve.
When submitting code, also test cases covering (hopefully every line of) the code are required. The test scripts should be located in the same directory as the actual code:
However, for the following directories (only one so far):
no testing is required. (For u/ this condition is relaxed as the code there is personal and/or under development)
With visualizations/plotting routines, it is not required to provide explicit tests. However, an example test/documentation script is required.
PEER REVIEWING CODE
Based on the meeting, systematic peer-reviewing of the code is perhaps too much. However, when adding new stuff that many people are going to use, it is good to ask someone to look over your code. With smaller edits of the code, this is not necessary.
More importantly, try to improve on other people's code when using it! This applies especially to the documentation and testing.
DON'T BE AFRAID
The shared code-library is most useful when many people are using it.
Thus, don't be too afraid of 'breaking things' as everything can be restored!