tests: replace all "assert" by unittest assertions

If the assert fails because of a comparison, it is useful that the
test suite log contains a comparison of these values.
unittest.assertEqual() and friends do that for us.

* HACKING: Add a section about Python tests.
* tests/sanity/style.test: Forbid the use of "assert" in
Python tests.
* tests/python/298.py, tests/python/341.py, tests/python/471.py,
tests/python/accparse2.py, tests/python/aiger.py,
tests/python/aliases.py, tests/python/alternating.py,
tests/python/bdddict.py, tests/python/bdditer.py,
tests/python/bugdet.py, tests/python/complement_semidet.py,
tests/python/declenv.py, tests/python/decompose_scc.py,
tests/python/det.py, tests/python/dualize.py, tests/python/ecfalse.py,
tests/python/except.py, tests/python/game.py, tests/python/gen.py,
tests/python/genem.py, tests/python/implies.py,
tests/python/intrun.py, tests/python/kripke.py,
tests/python/langmap.py, tests/python/ltl2tgba.py,
tests/python/ltlf.py, tests/python/ltlparse.py,
tests/python/ltlsimple.py, tests/python/mealy.py,
tests/python/merge.py, tests/python/mergedge.py,
tests/python/misc-ec.py, tests/python/optionmap.py,
tests/python/origstate.py, tests/python/otfcrash.py,
tests/python/parity.py, tests/python/parsetgba.py,
tests/python/pdegen.py, tests/python/prodexpt.py,
tests/python/randgen.py, tests/python/relabel.py,
tests/python/remfin.py, tests/python/removeap.py,
tests/python/rs_like.py, tests/python/satmin.py,
tests/python/sbacc.py, tests/python/sccfilter.py,
tests/python/sccinfo.py, tests/python/sccsplit.py,
tests/python/semidet.py, tests/python/setacc.py,
tests/python/setxor.py, tests/python/simplacc.py,
tests/python/simstate.py, tests/python/sonf.py, tests/python/split.py,
tests/python/streett_totgba.py, tests/python/streett_totgba2.py,
tests/python/stutter.py, tests/python/sum.py,
tests/python/synthesis.py, tests/python/toparity.py,
tests/python/toweak.py, tests/python/tra2tba.py,
tests/python/trival.py, tests/python/twagraph.py,
tests/python/zlktree.py: Replace all occurrences of "assert" by calls
to unittest.TestCase methods.
This commit is contained in:
Alexandre Duret-Lutz 2022-03-04 16:59:04 +01:00
parent 7b7e1b254b
commit 530cf7ca47
69 changed files with 1314 additions and 1116 deletions

44
HACKING
View file

@ -290,8 +290,8 @@ would understand with:
make check LOG_DRIVER=$PWD/tools/test-driver-teamcity
Coding conventions
==================
C++ Coding conventions
======================
Here some of the conventions we follow in Spot, so that the code looks
homogeneous. Please follow these strictly. Since this is free
@ -682,3 +682,43 @@ Other style recommandations
* Always code as if the person who ends up maintaining your code is
a violent psychopath who knows where you live.
Coding conventions for Python Tests
===================================
Unless you have some specific reason to write test cases in C++ (for
instance do test some specific C++ constructions, or to use valgrind),
prefer writing test cases in Python. Writing test cases in C++
requires some compilation, which slows down the test suite. Doing the
same test in Python is therefore faster, and it has the added benefit
of ensuring that the Python bindings works.
We have two types of Python tests: Python scripts or jupyter
notebooks. Jupyter notebooks are usually used for a sequence of
examples and comments that can also serve as part of the
documentation. Such jupyter notebooks should be added to the list of
code examples in doc/org/tut.org. Testing a notebook is done by the
tests/python/ipnbdoctest.py scripts, which evaluate each cells, and
checks that the obtainted result is equivalent to the result saved in
the notebook. The process is a bit slow, so plain Python scripts
should be prefered for most tests.
If you do need a notebook to tests Jupyter-specific code but this
notebook should not be shown in the documentation, use a filename
starting with '_'.
Tests written as Python scripts should follow the same convention as
shell scripts: exit 0 for PASS, exit 77 for SKIP, and any other exit
code for FAIL.
Do not use assert() in those scripts, as (1) asserts can be disabled,
and (2) they provide poor insights in case of failures. Instead do
from unittest import TestCase
tc = TestCase()
and then use tc.assertTrue(...), tc.assertEqual(..., ...),
tc.assertIn(..., ...), etc. In case of failures, those will print
useful messages in the trace of the tests. For instance multiline
strings that should have been equal will be presented with a diff.