Discussion

Testing

How can you be sure that the library does what I say it does?

To begin with there is a test suite, admittedly fairly small, but which provides high code coverage and is useful for catching regressions.

Really, most of the work winkling out bugs is done by:

  1. Testing against an existing codebase (Open Babel) to look for cases where there is disagreement in whether a particular complete SMILES is accepted/rejected.

  2. Testing the invariant, that for any complete SMILES string, if any prefix is rejected then any longer prefix (including the complete SMILES) should also be rejected. The corrollary is that if a complete SMILES is accepted, then any prefix of that should also be accepted.

These are tested using a large number of SMILES strings which I construct on-the-fly from ChEMBL:

  1. Read two SMILES strings at a time, A and B

  2. Construct new SMILES strings by combining every prefix of A with every suffix of B, and vice versa

So how can you be sure that the library does what I say it does? You can’t. Not until you run the test scripts, which are all included:

>>> python suite.py
...
OK
>>> python test_invariant.py chembl.smi
0
100
...

Alternate tokens

In the Introduction, I mention that one application of the library might be to flag a problem so that an alternate token could be suggested instead. Is this actually feasible? And even if so, should one do it?

…