ChaiScript script evaluation should have a non-throwing alternative


Continuing the discussion from Best way to protect application from script bugs:

Just a related note:

Basically, that there is only exception handling to catch issues from scripts is becoming VERY problematic
in high-performance applications that allow editing code inside them. In this situation,
throwing and catching an exception have (very) unpredictable performance which is bothersome
if you need to not make the performance fluctuate too much.

My concrete example is with a game: in this game I sometime need the player to be able to (optionally) write scripts.
However player will necessarily make a lot of mistakes and I need this part of the game to behave like
any compiler+vm. So I report errors, but the fluctuation of performance due to exception handling makes
the experience sometime a bit nasty.

It also totally prevent the use of ChaiScript in contexts where throwing an exception is simply forbidden (a lot of C++ shops and projects).

Also, it does not total sense to throw exeption from ChaiScript in all situations: executing a failing script might be just normal behavior of the code using ChaiScript.

Simply said, it should not be ChaiScript that decide how to dispatch errors resulting from wrong script code.


  • I do use exceptions in some places in my game, but only for system failures, not logical ones;
  • ChaiScript should totally throw if it’s a problem which is not about the script being wrong (not sure how to fix the boundaries, but syntax issues and not-found names should not be exceptions);
  • I’m suggesting that it would be very useful to have an alternative way to evaluate scripts, which does not throw on script that is lexically wrong or access symbols that do not work; but it should not be the default.

I don’t know of a good solution to combine both with and without exceptions (maybe “expected” return type?) but I do consider the throw-on-failure-to-evaluate-script a major performance issue in my project.


I appreciate the desire to have a completely exception-free version, but at this point I believe that would take a complete rewrite of the internals of ChaiScript to make that possible. I might be wrong, but exception handling is even used internally in a few cases.

At the time of the original implementation I believed that there was no additional cost to using exceptions, I’ve learned since then that is not true of course.


That is what I feared.
I don’t know enough of ChaiScript code yet (and I lack time) but I might be able to study potential solutions limiting the need to rewrite soon. I’ll notify you if I have an adequate solution.


By the way, would heavily modify ChaisScript for higher perfs, more customization of at least error handling would be a good target goal for a new major version? Or do you have other plans already ongoing?


That’s largely the goal of this branch that I’ve been working on: it has a modular optimization engine the ability to add in callbacks on each script element execution so that you could possibly implement something like a debugger or “break” capability.


Nice! I will take a look at it, thanks for the pointer.