Additional checks for unsigned integers to avoid underflow and overflow

[nekiro]

Additional checks in getNumber(L, arg)

Prints a handy warning if unsigned integer is negative or exceeds the max number for defined type.

[DSpeichert]

Note the potential performance impact as mentioned by Mark.

[MillhioreBT]

This implementation causes performance to be compromised, in fact it drops by about 70%~

You can try the following code and you can compare the time, you will notice that with this implementation increases the time dramatically.

time_t now = OTSYS_TIME();
for (int i = 0; i < 1000000000; i++) {
    getNumber<time_t>(L, 2);
}
std::cout << "Test getNumber of LuaInterface: " << std::setprecision(15) << (double)(OTSYS_TIME() - now) << std::endl;

I know that there are never cases in which getNumber is called with this frequency, I don't know if it is worth it, Someone else should do tests, I don't know if the tests I did were enough or well done

[yamaken93]

@MillhioreBT are you testing with release version all optimizations enabled?

[MillhioreBT]

@MillhioreBT are you testing with release version all optimizations enabled?

Of course, the code is clear to the naked eye, obviously it takes more time, but it wouldn't be a big problem really, since at small scales there doesn't seem to be a much difference.
If calling getNumber(L, 2); 1000000000 times requires 3 seconds, with these modifications it will require approximately 5 seconds, but let's face it, when can this situation occur?

[MillhioreBT]

It would be amazing to have the opinion of @marksamman about this

[nekiro]

Is it dropping performance in general or is it with invalid value only?
Well IMO TFS should have something called "debug mode" and I'm not talking about debug build, but config.lua switchable variable which would turn on/turn off certain code, this is one of the cases where it would turn it off/on, cause this can impact performance, but it's very helpful feature.

[Zbizu]

imo not worth losing 70% of the performance for that. Should be optional in compiler settings.

[DSpeichert]

The test conducted by @MillhioreBT is synthetic, not representative of what datapacks in the wild do.
In the absence of other comments over the time, I'm willing to take it as a feature unless real-world performance problems are reported.

[dex-89]

After that I have this problem...