Error handling

[hfossli]

I see quite a few asserts here and there in the code. Does that mean that if for example the solver is overconstrained the program will crash and exit? Or if I try to remove a constraint which hasn't been added it will crash and exit?

[hfossli]

You see I'm considering writing a Swift wrapper and this seems to be my biggest concern so far.

[starwing]

the assert() means never raise error, if assert() occurs, it's the coding bug and not a programming runtime error. any assertion failure should not occurs in production scene.

I have added auto test and coverage test, it's not 100%, that's the way I will work. for now in coverage amoeba should never crash on any input data.

[hfossli]

Great!

[hfossli]

Overconstraining a variable fails with

Assertion failed: (exit.id != 0), function am_remove, file ./amoeba.h, line 963.

Example code

static void test_twice(void) {
    am_Solver *s7 = am_newsolver(debug_allocf, NULL);
    am_Constraint *c15 = am_newconstraint(s7, 1000000000.0);
    am_setstrength(c15, 1000000000.0);
    am_setrelation(c15, 2);
    am_Variable *v14 = am_newvariable(s7);
    am_addterm(c15, v14, 1.0);
    am_addconstant(c15, -10.0);
    am_add(c15);
    am_Constraint *c16 = am_newconstraint(s7, 1000000000.0);
    am_setstrength(c16, 1000000000.0);
    am_setrelation(c16, 2);
    am_addterm(c16, v14, 1.0);
    am_addconstant(c16, -10.0);
    am_add(c16);
}

It seems better if it instead

• returns a proper error code e.g. AM_FAILED
• leaves the solver as if am_add(c16) didn't happen at all

[starwing]

I will test it.

[starwing]

to add a new constraint, we have several ways to try (via Cassowary algorithm):

1. add it directly, if we find a subject.
2. use a artificial variable to add.

if we cannot find a subject and all terms are dummy variable, and constant is not zero (means some constraint like x >= 10 when x <= 0), it's unsatisfied.

then, if we can not add it with a artificial variable, the constraint may unbound.

but, all other implements just throw errors in all these situations, I just try to restore the state of solver and return the proper error code. So I must remove the added constraint if needed, I need may test case on adding, because all other implement not do this.

If I'm not real add the constraint into solver and I try to remove it, this assertion failure will trigger, this is the source of this failure. I will improve it, but I need more test case to fix every the one case.

[hfossli]

I just try to restore the state of solver and return the proper error code

Music to my ears!

I will improve it

Fantastic!

but I need more test case to fix every the one case

I can create more test cases. The test_twice test was created using my swift wrapper. I can enable logging of all actions to the c-api of amoeba so that it is easy to reproduce any errors I get. That's why the solver is named s7 :)

Your library is very easy to use. Your style of writing c code is very clean. I'm very impressed with your skills!

[starwing]

Very very thank you for your support :-)

[starwing]

I have fixed this assertion failure, you could try the latest version.

[hfossli]

Perfect!