Detect broken string templates that will raise ValueError when interpolated

[Jamu Kakar]

For example, code like this will break:

  >>> "%(foo)" % {"foo": "bar"}
  Traceback (most recent call last):
    File "<stdin>", line 1, in <module>
  ValueError: incomplete format

It'd be awesome if Pyflakes told me about that kind of issue.

[Jamu Kakar (jkakar)]

I poked through the code and determined that implementing this kind of
check will be a pretty big pain in the neck. We'd need to parse the
string to determine what kind of interpolation is required and then
verify that with the arguments to be interpolated. ie, we'd need to
implement our own string interpolation engine. Ew.

[Jean-Paul Calderone (exarkun)]

A patch to implement this was sent to the mailing list, https://lists.launchpad.net/divmod-dev/msg00193.html

[Jakub Wilk (jwilk)]

Implementing your own string interpolation engine may sound scary at first, but in fact it should be quite easy. For example, pylint has a template parser that is just ~60 lines of code:
https://bitbucket.org/logilab/pylint/src/36355f1fe1fd/checkers/utils.py?at=default#cl-277

I took a different approach in lintian4python, which resulted in even shorted code:
https://bitbucket.org/jwilk/lintian4python/src/5e552f08d8f5/helpers/python/code-analysis?at=default#cl-292

[asmeurer (asmeurer)]

Too bad they're both GPL, as you otherwise could literally just reuse that code.

[Jakub Wilk (jwilk)]

The relevant part of lintian4python is under the same license as pyflakes.

[asmeurer (asmeurer)]

Rather than rewriting Python's string interpolation, would it be possible to just take the string and interpolate it with some dummy objects that have sufficient __int__, __float__, __str__, etc. defined on them, and see if any errors are raised (basically, reuse Python's own engine)?

[Jakub Wilk (jwilk)]

Absolutely. That's what I did in lintian4python.