Less dishonest COME_FROMs in 3.6+ code

[rocky]

Addresses all of problems seen in 3.7 datetime.py

See also Issue #235

[rocky]

This change is causing many some failures to the 3.7.3 library. But I wonder if still shouldn't go forward and fix up the grammar more. What do you think?

[x0ret]

Thank you, i think we can fix related issues in a reasonable time.
however i started reading python-control-flow. Maybe we can have a schedule to embed into this project.

I will check again the library and continue.

[rocky]

Oops - this isn't needed (or used) anymore. The decision (for now) is not to add COME_FROMS to backward (in current Python code generation that means looping) jumps.

[rocky]

I have thought a lot about using python-control-flow. It would require a very serious effort to get code as good as we have now - Python is complicated in its control flow. However it would revolutionize this project and put it a much better place for Python 3.7, 3.8 and beyond. And we could get rid of a lot of hacky code and things would be more intelligible.

A strategy for rolling this in is to do it in a branch and just try one of the more recent Pythons on it. After that works expand to other Pythons.

Another approach is to create an entirely new project that just has code for one Python that uses this. And after that works well expand it.

Your thoughts?

[x0ret]

A strategy for rolling this in is to do it in a branch and just try one of the more recent Pythons on it. After that works expand to other Pythons.

To be honest, I do prefer to have this progress in this project in the future : ) , since you do a lot of work on it. However i think isolating was always a good option for me. Maybe we can start new project for as you said (python 3.8?) and see the progress to make final decision.

I'm trying to adjust myself to python-control-flow as soon as possible.

I really appreciate your efforts in this project. This is so promising.

[x0ret]

@rocky, i checked half of 3.7 lib issues, and it seems most of them are related to chained comparison, So do we continue in this PR?

EDIT: Ok, i get it;

[rocky]

@x0ret I am going to stop working on this for today. I have too much other work. If you could get this to pass the test, I think we could have most of the 3.7 issues behind us.

After that, maybe a release. And then focus more on control flow better way.

[rocky]

Thanks! Let's do it and move on..