[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Compiler error handling

How are folks handling errors during the compile phase?

I have the system walking the tree and generating code. But when I encounter
an error, I can't throw an exception (as outXXX doesn't allow it).

Do folks just note the errors in a log someplace, and then set flags?

So, say I discover SYMBOL NOT FOUND, should I just set a flag, note the

So, for example, I can reset the error flag as I'm going into a high level
construct (statement, expression, whatever).

Also, from an idiomatic point of view, I wonder if I'm doing things the
"correct" way.

I've got my DFA class, and it's overriding a bunch of outXXX methods.

In each out method, I generate a bunch of node information using
setOut(node, info).

So, something like this would happen (contrived code follows):

public void outAddExpr(AAddExpr node)
    NodeInfo rhs = (NodeInfo)getOut(node.getRHS());
    NodeInfo lhs = (NodeInfo)getOut(node.getLHS());

    if (rhs.getDataType() != lhs.getDataType()) {
        error("can't add different datatypes")
    } else {
        NodeInfo ni = new NodeInfo();
        ni.type = rhs.getDataType();
        setOut(node, ni);

So, in this case, we're bubbling up expression type infomation.

But I also get a lot of these for "fill in" parts of the grammar, where it's
mostly syntactic sugar.

So, I do this:

public void outSugar(ASugar node)
    NodeInfo ni = (NodeInfo)getOut(node.getThing());
    setOut(node, ni);

It's working fine, I'm just curious if this is "the way it's done" or am I
missing something. In the Thesis, there was a section where Etienne talked
about having a bunch of nested AnalysisAdapter anonymous classes. I admit
that it wasn't particularly clear what he was doing here to me.

So, it's mostly an idiomatic question that I'm doing this the "right way".


Will Hartung