[docs] [issue37004] SequenceMatcher.ratio() noncommutativity not well-documented

Dennis Sweeney report at bugs.python.org
Tue May 21 22:09:15 EDT 2019

New submission from Dennis Sweeney <sweeney.dennis650 at gmail.com>:

I understand that the SequenceMatcher's ratio method does not guarantee that SequenceMatcher(None, a, b).ratio() == SequenceMatcher(None, b, a).ratio(). Below is a counterexample:

    # Example from https://mail.python.org/pipermail/python-list/2010-November/593063.html
    >>> SequenceMatcher(None, 'BRADY', 'BYRD').ratio()
    >>> SequenceMatcher(None, 'BYRD', 'BRADY').ratio()

I was recently solving a problem that required a textual similarity ratio function and I wrongly assumed that SequenceMatcher treated both input strings symmetrically, which was an extremely difficult bug to find, especially because for many simple tests, the ratio IS symmetric:

    >>> SequenceMatcher(None, 'apple', 'banana').ratio()
    >>> SequenceMatcher(None, 'banana', 'apple').ratio()

I would like to see a clearer warning of this asymmetry in the documentation for the difflib module. Perhaps something like

      .. note::

         Caution: The result of a :meth:`ratio` call is *NOT* symmetric with 
         respect to the order of the arguments. For instance::
            >>> SequenceMatcher(None, 'brady', 'byrd').ratio()
            >>> SequenceMatcher(None, 'byrd', 'brady').ratio()

Without such a note near the ratio methods' documentations, it is far too easy to google for a Python stdlib functionality for computing text similarity, skip straight to the ratio method, look at the examples given, try some of your own simple examples, and accidentally convince oneself that this symmetry exists.

messages: 343138
nosy: Dennis Sweeney, docs at python
priority: normal
severity: normal
status: open
title: SequenceMatcher.ratio() noncommutativity not well-documented

Python tracker <report at bugs.python.org>

More information about the docs mailing list