Tests & documentation¶
The scientific community broadly discusses the capabilities and shortcomings of hydrological models from a theoretical point of view. For example, many sensitivity studies address the negative impacts of low data quality. By contrast, we unaware of any studies estimating the adverse effects of bugs and misleading documentation of hydrological computer models. With little attention paid to these issues during evaluation processes (e.g. during peer-review), there is a risk of publishing impaired model results, possibly compromising the drawn conclusions. See, for example, the commentary of Hutton et al., addressing this topic from the scientific perspective.
This section describes strategies to limit the risk of serious errors and outdated documentation. We try to keep the code and the documentation in sync by connecting them as strong as possible, using Python’s “docstring” and “doctest” features.
The first “connection” is writing each piece of the documentation close to the related source code. Therefore, we prefer to write all explanations addressing specific HydPy features as docstrings. Docstrings are documentation strings attached to the Python objects they explain. When extending HydPy, it is a strict requirement that each newly implemented public member (including the sub-members, e.g. the user-available methods of a class) comes with its own docstring. Our Travis CI based continuous integration workflow recognises any missing docstrings using Pylint and reports them as errors.
The second “connection” is to use and extend the functionalities of Sphinx,
which collects the source code, the docstrings, and the usual documentation
files to generate the online documentation. Sphinx relies on the
reStructuredText format, hence follow this format when writing docstrings
and regular documentation files. However, instead of using its regular
referencing style, make use of “substitutions” as defined by class Substituter
of module autodoctools
. Write, for example, the class name “Substituter”
within vertical bars to reference the corresponding class properly. This short
syntax allows making frequent use of substitutions. A helpful side effect is
that, during the generation of the HTML pages, wrong substitutions result in
warnings, interpreted as errors by our Travis CI based continuous
integration workflow (see section Continuous Integration). This
mechanism increases the chances that, when documentation adjustments do not
accompany future code changes, the Travis CI based workflow breaks, forcing
the responsible programmer to adapt the documentation.
The third “connection” is to define a sufficient number of doctests.
Doctests are documentation sections containing valid Python code followed by
the expected results. For developers, doctest
is not always as convenient as
other unit testing frameworks (e.g. unittest
), but it offers the great
advantage to defining tests that are understandable for non-programmers. At
best, each (sub)member should specify its own doctests, in conjunction with
some “normal” explanations about its purpose and usage. Non-programmers should
be enabled to learn using HydPy by repeating the doctests. Besides their
intuitiveness, doctests (like substitutions) offer the significant advantage of
keeping source code and documentation in sync. For example, as long as the
following three-line doctest remains in the documentation, one can be sure that
the current core package contains a module named “objecttools”:
>>> from hydpy.core import objecttools
>>> objecttools.classname(objecttools)
'module'
The Python script run_doctests.py collects the doctests of all modules and executes them.
We measure the “code coverage” (the number of executed code lines divided by the total number of code lines), determined by the coverage library. The current code coverage of HydPy is 100 %. Travis CI reports all future changes introducing uncovered lines as failures to make sure future changes do not undo this favourable situation.