Require that generator params/args be a superset of typegen params

[hofstee]

This is a fix for issue #27.

[rdaly525]

This change looks good.

Two fixes:

1. can you update the checkParams to a name that reflects the operation better?

2. Can you change this to pull request into the 'dev' branch? I need to do a lot more continuous integration work in order to update master (Because coreIR is a dependency on another larger Stanford project)

[hofstee]

Retargeted onto dev.

[hofstee]

I think checkArgs might be a better name than checkParams, but I'm not sure if that's the best name for this.

[rdaly525]

@THofstee Super sorry about this... There are a bunch of changes that are merged that probably are inconsistent with some of the things you did. Let me know if I could help fix anything.

Also I think I need two functions for the checkParams thing. One that checks if they are exactly equal and one if the params are a superset of the args.

Maybe the names areArgsParamsMatching, and areArgsParamsSubset. Or something along those lines. I am terrible at coming up with names. Ideally the name is some sort of question that describes the boolean operation explicitly.

[hofstee]

Out of curiosity, what would be the reasoning for having another function test whether or not they are exactly equal?

[rdaly525]

I do the matching check on the Args being passed into the generator. Is there a good situation where you would want to pass in more args to a generator instance than the generator needed? Perhaps?
I do totally agree that the typegen should be a subset of the generator args.

[hofstee]

I'm not sure. Previously in the other branch I was working in I used it out of convenience because I didn't need a type to pass into runGenerator, but I would get a segfault if I passed in a nullptr so I just grabbed the TypeGen for the Generator and ran it with the arguments for the generator.

I think it's probably better to only have the equality check instead of subset when you provide arguments to the parameters. I can't think of any programming languages that allow you to pass in more arguments to a function than the function has parameters, and I can imagine a lot of nasty bugs that might result from this.

I'll need to double check to see if the nullptr passed to runGenerator still causes a segfault in the dev branch. I'll clean this up a bit tomorrow.

[rdaly525]

Just as a warning, I did clean up the infrastructure for running the typegen and generator functions a bit. I made them into an actual method of the class rather than a random function (although I did keep that ability to pass in a random function). If the API is not clear, definitely let me know and add it to Issue #33. Also if you are getting segfaults, then definitely report an issue. Ideally you should not be able to segfault.

[rdaly525]

Thanks again for looking into this stuff! I know it is annoying that things keeps changing, but that is the definition of Agile!

[hofstee]

Seeing as how there is only one function now to check that args exactly match params, I've just left it renamed checkArgs similar to other compilers.

[rdaly525]

Looks good!