handling TimeConstrained[] in SymPy evaluation

[mmatera]

An idea to fix TimeConstrained and related functions, at least for evaluations involving sympy.
Still I can produce a core dump if I ask for a ridiculus task in Sympy, but is a problem of Sympy, not from Mathics

[mmatera]

@davidar, ideas about how to implement this in Pyodide are welcome.

[davidar]

Unfortunately I don't think there is a good solution to that in pyodide yet, at least not without cooperation from the javascript environment hosting the pyodide interpreter (https://pyodide.org/en/stable/usage/keyboard-interrupts.html). Might just need to disable it in pyodide for now.

[rocky]

Unfortunately I don't think there is a good solution to that in pyodide yet, at least not without cooperation from the javascript environment hosting the pyodide interpreter (https://pyodide.org/en/stable/usage/keyboard-interrupts.html). Might just need to disable it in pyodide for now.

@davidar Thanks for the information and link. It confirms what I had meant to look up.

What Mathics3 should do is provide a signal-handling mechanism that causes the current computation to stop after a certain amount of time (SIGALRM).

Mathics3 should also trap other signals like SIGINT. In wolframscript, Ctrl-C (SIGINT) goes into an Interrupt prompt. Later, we should provide a compatible (and possibly much better) interrupt interface.

If I have this correct, the link you have indicates that pyodide does provide a means for sending trapping and sending signals back to Python.

One other thing to note about Python's signal handling - only the main thread can trap signals. I don't think this is a problem, just something to note.

With this we can provide a somewhat coarse mechanism for handling timeouts that is compatible with pyodide.

@mmatera Now that you may have been just informed about signal handling and timeouts, please don't jump ahead and start implementing this in code. (If you want to experiment privately - by all means go ahead.)

If there is a need right now for TimeConstrained, I'd appreciate knowing what the deal is.
Otherwise, let's leave adding signal handlers in Mathics3 and having that work in pyodide for later.

[rocky]

@mmatera Please give me a couple of days for me to pull down this branch to try out. Thanks.

[mmatera]

@mmatera Please give me a couple of days for me to pull down this branch to try out. Thanks.

Sure, no hurry. I just put this because when I looked over the issue with Simplify I realized that this would be a way to support TimeConstrained in systems that support Threads.

[rocky]

I tried

TimeConstrained[Integrate[Sin[x]^1000000,x],1]

I am getting a RecursionError that is not handled. Worse is that calls with_traceback() with the wrong number of parameters. A PR has been created to address this.

[mmatera]

Actually, with Pyston I got a core dump for that test. But it is reasonable: Sympy does not control by itself how much memory/work needs to converge.