Eric suggested in a code review for issue #23731 that maybe we should have the possible optimization levels listed somewhere.

Brett, I know you're away this month, so please ignore this,

Eric, I hope the changes I made reflect your original intent in the code review for #23731.


This is my first change to the C code, so I hope it's close.  The tests passed, so it must be correct.  ;-)  [j/k]

(Sorry for taking so long!)

Thanks for doing this, Cheryl!  I'm leaving a review on your PR. :)

Also, in PEP 421 it says that you need a PEP for this. [1]  "Such a PEP need not be long, just long enough."  (I just realized that the requirement isn't obvious so I've opened issue #37284.)


[1] https://www.python.org/dev/peps/pep-0421/#adding-new-required-attributes

There are (solvable) problems with my original recommendation:

1. sys.implementation is by definition not suitable for third-party import hooks to modify
  + it is set during the Python implementation during runtime init
  + it is effectively read-only after that
2. "opt_levels" is too specific to the CPython status quo
  + there are other ways to encode the optimizations of a bytecode file [1]
  + "optimizations" would probably be more correct
  + that opens a whole can of worms (e.g. what does sys.flags.optimize mean)

So we may want to think this over a bit before going any further.  I'm going to collect my thoughts on this and write more later. :)


[1] In PEP 488 it says:

    It is expected that beyond Python's own two optimization levels,
    third-party code will use a hash of optimization names to specify
    the optimization level, e.g. hashlib.sha256(','.join(['no dead code',
    'const folding'])).hexdigest().