Question

I have written a parser-generator that has expressions like this:

a = f:"foo" "bar" { /* your code here */ }

Which, after a fashion, generates code that looks about like this:

r0 = this.ReturnHelper<string>(cursor, cursor, () => /* your code here */);

This is perfect, because the users can specify an expression, like f to just return a value, or they can add an extra set of curly braces and write a method. So far so good.

Now, I also allow assertions like this:

a = w:word !{ w == "disallowed" }

Which compiles into something like this, currently:

var w = this.ParseWord(...).Value;

...

if (!( w == "disallowed" ))
{
    r2 = new ParseResult<string>(cursor, cursor, string.Empty);
}

I would LIKE to have the assertions support the same method-body option that the ReturnHelper provides, but it seems inefficient to have a method bool AssertionHelper(Func<bool> assertion) that just calls the predicate.

I was imagining that new Func<bool>(() => /* your code here */)() would work, but I’m not sure that it is the most efficient.

Any ideas?

EDIT

How about ((Func<bool>)(() => /* code */))()? Does that avoid an allocation? Nope, that generates identical IL to the above version.

Answer 1

It’s not clear why you need to mention Func<bool> explicitly at all. Why not just:

this.AssertionHelper<string>(cursor, cursor, () => !( w == "disallowed" ))

? Assuming it’s going to be a single expression with no parameters and it’s of type bool, that should be fine.

If it needs a parameter w, just change it to:

this.AssertionHelper<string>(cursor, cursor, w => !( w == "disallowed" ))

I believe the MS implementation of the C# compiler will avoid creating a new delegate instance on each invocation of this – but really, does it matter? Have you profiled your application to check whether this is actually a significant performance hit at all? If you have, that’s fine – and I’ve certainly seen some interesting data around some cases where micro-optimization for delegates can make a difference – but I wouldn’t worry about it until I’d found it’s an actual problem.