Iterator exceptions

[sydow]

Changes iterators so that they throw exception StopIteration at termination (rather than returning None).

As a consequence, the Iterator class in builtin.act his now defined by

class Iterator[A] (object):
next : () -> A

(previously the return type was ?A)

[plajjan]

This is failing CI testing because of the changes to max() I think. It's the same problem as for pop etc in #1986. I said I would work on that to split the change in two parts. I think I can do the same for this PR. As things currently look, I will have some spare cycles early next week.

[plajjan]

This PR has some bug, not sure where. After I have fixed max / max_def, the functionality of Acton programs should be identical before and after this branch, yet, when I do make gen in the respnet repo, it generates entirely different code, indicating that somewhere we have an actual difference in behavior. I suppose it's best if I can find a smaller reproduction.. easier said than done.

[sydow]

I don’t know what to do about this. If I do git pull in the main branch and then build the system and do make test, all tests pass. I also don’t know what the respnet repo is. Could you explain more? Björn

…

On 15 Dec 2024, at 23:07, Kristian Larsson ***@***.***> wrote: This PR has some bug, not sure where. After I have fixed max / max_def, the functionality of Acton programs should be identical before and after this branch, yet, when I do make gen in the respnet repo, it generates entirely different code, indicating that somewhere we have an actual difference in behavior. I suppose it's best if I can find a smaller reproduction.. easier said than done. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[plajjan]

@sydow respnet is https://github.com/orchestron-orchestrator/respnet/ - in that repo is a code generation step that can be run with make gen. The currently generated code (that is checked in) was run on acton main - and I expect the same code to be generated using this branch. You agree with this, right? This PR is not supposed to alter control flow in programs, right?

However, the generated code does change and so I can only conclude that this PR does alter control flow somehow.

[plajjan]

I have created #2023 which effectively supersedes this PR.

It includes your code @sydow but:

• is rebased on latest main branch, since this PR has conflicts
  • the conflicts are because I added min_def & max_def to the main branch, so we can do a soft-transition in other projects
• I changed back the min() and max() functions so they still take a default argument - this way, we avoid changing too many things in - instead we just focus on the StopIteration part and leave the change to min / max for a second PR
• I did squash in my changes - maybe this was a mistake, but I thought it was easier to see the changes like this

[plajjan]

Closing this, see #2023 instead!