Hello all,
I recently integrated pyflakes into
Leo and am awed by its speed and elegance. Its simplicity is pure genius. A few days ago I set up the
pyflakes-study google group and corresponding
github repo and have been happily playing around with the pyflakes code ever since. To properly see what I have done you will have to look at the .leo file with Leo.
To paraphrase Richard Feynman, that which I cannot alter I do not understand. By this criterion I now understand almost every line of pyflakes. Hehe.
It's possible to speed up pyflakes by at least 10% by using a less elegant traversal scheme, as describe
here. Not sure whether you would be interested--there are some drawbacks.
It's also possible to speed up pyflakes by an additional 5% merely by removing the scope property, as described in
today's post. This post also describes a buglet in pyflakes and its fix.
Most of today's post is geared toward a new tool I am considering. This would compare name conventions with actual usage. The first phase might be a scope resolver, using a simplified
version of the pyflakes code. The second phase will have to do a whole lot more
with Ast.Attribute :-)
You might call the new tool a follow up to my
make-stub-files project, with corresponding
github repo. I announced make-stub-files in my largely unsuccessful
python-static-typechecking group. Boy, am I happy
mypy is doing the heavy lifting.
I would appreciate any comments, suggestions and corrections you might have.
Edward
------------------------------------------------------------------------------------------
Edward K. Ream:
edreamleo@gmail.com Leo:
http://leoeditor.com/------------------------------------------------------------------------------------------