Question: Is the order of Stmt.getUses() guaranteed to match params?

[zlh20040308]

Hi,

While working on A4, I have a question regarding transferCallEdge().

I noticed that in transferCallEdge(), the arguments from Stmt.getUses() seem to be in the same order as parameters from getParams(). Is this ordering guaranteed by the API?

I checked the docs but didn't find clarification. If it's intended, maybe we could document it to help future users. If not, perhaps we could provide a utility method like CallEdge.getParameterMapping() to make the mapping explicit.

My implementation of transferCallEdge() is as follows:

protected CPFact transferCallEdge(CallEdge<Stmt> edge, CPFact callSiteOut) {
    List<RValue> uses = edge.getSource().getUses();
    List<Var> params = edge.getCallee().getIR().getParams();
    CPFact newFact = new CPFact();
    for (int i = 0; i < params.size(); i++) {
        if (uses.get(i) instanceof Var var) {
            newFact.update(params.get(i), callSiteOut.get(var));
        }
    }
    return newFact;
}

This code relies on an implicit assumption: the order of arguments returned by uses corresponds exactly to the order of parameters in params. While I verified this through debugging, this guarantee is not documented in Stmt.getUses() or JMethod.getIR().getParams().

A cleaner design might expose a getParameterMapping() method directly on CallEdge, making the argument-parameter correspondence explicit and freeing callers from depending on index ordering.

Thanks!