ronaldooeee / avajava Goto Github PK

working on it...

The optimizer needs to implemented across all entities. Also needs a good number of tests.
Current Milestones (until we can think of more):
-constant folding (in progress)
-unreachable code elimination
-dead code elimination
-branch elimination
-tail recursion (unfolding?)

We have decided that we want type inference, but what about changing the type of the VarDecl later on?
Do we want to make the types of VarDecl's unchangeable, or should they be changeable if a new type is assigned to them?

var x = 10;
x = true; // should this be valid? Or should it return an assignment error?

Should we implement classes? This issue is tied to dot operators as well.

If we had classes, then we could call functions using the dot operator:

Option 1

var x = [ 1, 2, 3, 4, 5 ];
x.length;

Otherwise, we would have to pass in the thing we are performing the function on.

Option 2

var x = [ 1, 2, 3, 4, 5 ];
length(x);

Option 2 is essentially already implemented for us. We just need to add two lines of code in one file.
Option 1 would require us to implement a whole new feature of having Classes, which wouldn't be that much, but do we have time for that?

Make the scanner work and implement tests

On the readme

• You need a logo (that was required for homework 1)
• I don't see the list of example programs

Right now that last statement of every function block requires a semicolon. Then the functionExp terminates with an 'end' keyword.

Do we still want this to be our syntax? The coder would have to know that the very last statement of every Function must have a semicolon.

We also need to implement some built-in functions.

From what we learned in class, it seems like we will need to put it somewhere in the Analyzer.js file and the initialContext. They need to be stored in the symbolTable of the top-level parent context.

I don't understand how a term project with a fair number of committers had only one issue so far and zero pull requests.

Need to figure out how to test all of the code being generated.
Right now we are generating just fine and manually checking each set of generated code by copying and pasting into the node interpreter and seeing if it runs and ends up with the results we were looking for.

Need a way to programmatically check generated code. But how do we do that without hardcoding each end result?

Anyone have any ideas?