Develop better tooling for static and dynamic analysis

[gvanrossum]

When we have an idea for rewriting some byte code (e.g. ADD_INT) we should be able to assess the relevance to a large body of code quickly. For example, static analysis could pull down the 1000 most popular Python packages from PyPI or GitHub and count opcodes for all .py files found there.

Dynamic analysis requires running code, which makes things more complicated, and moreover running it with a modified CPython interpreter. But we should at least have some standard way to do this, rather dan doing it ad-hoc each time a question turns up.

[gvanrossum]

A list of popular PyPI packages is published monthly here: https://hugovk.github.io/top-pypi-packages/

There are docs for accessing PyPI via API here: https://warehouse.pypa.io/api-reference/json.html

Is it worth it to separately also download popular GitHub projects? If they contain Python code they should be on PyPI, right? (Assuming they are popular.)

[gvanrossum]

I'm done with this (for the time being), the results are posted in https://github.com/faster-cpython/tools/