janbessai / ecoop2021artifacts Goto Github PK

M2 is untested and the code coverage is low right now.

Producers in Interpreter need to be covariantly overwritten whenever there is a new method.
Otherwise:

exp.newMethod() // works
exp.simplify().newMethod() // breaks

Might be skipped - there are not tests in the original study at all.

The HTML dom example is interesting, but won't make it into the paper and should be moved from here to a standalone project.

The project should have a toplevel README.md with instructions.

The finalized classes in xml use toplevel Exp<...> for their children, while those in exp use finalized.Exp

This also should be checked for the TAPL and Scala code.

Convert is our invention and not part of the original EP-Trivially approach.
If this breaks producers/binary-methods we need to cast or bail and mark the spot as problematic.

The code should include some comments where the other approaches violate which EP assumption.
Some of this needs to be toplevel (readme), e.g. OO where "no modification" is violated.
Other things are best put into comments.

The TAPL book case study needs to go into this project.

The basic Visitor and Exp interfaces should reside in m0, since Visitor mentions code from m0.

The original paper leaves it unclear how to deal with side-effects in trivially, but truncate can at least be implemented in m4 (and then potentially break in m4+k).

The template for extensible-visitor, trivially, and interpreter test M(n+1) should call MnTest.TestTemplate.super.test();

The generic dom needs to be replaced with xml.

When the changes are done we need to run the benchmarks once and export the results for the paper.

Some simplify methods compare eval() results using == which is problematic since they are Double instead of double