vpax

FWIW, I ran across this topic while doing the script optimization work. I suspect what you frame above is vestigial from back when `enum` types didn't have to be named....

(FYI, #1472 isn't happening any time soon)

Good to hear, as that was in fact exactly the use case (which I ran across in a different context) that made me think "geez we need to find these"!

Thanks, @ckreibich ! I misread that phrase during my search as saying that support _is_ provided. I got down this path, however, due to `testing/btest/language/table-pattern-index.zeek`, whose presence clearly indicates an...

Point taken re the semantics being more clear for `table[subnet] of T`. Regarding avoiding iterating through the table, I don't think under the hood we could (easily) implement anything other...

This started sounding familiar as I thought further about it. Sure enough, the current semantics appear to be a conscious choice - see the discussion in https://github.com/zeek/zeek/pull/251.

I'm fine with revisiting, I just am cautious as my recollection is that it's tricky. That said, if we just used C's semantics, say, seems it would be tried-and-tested.

Alas, I missed the original notification for this and only saw it now when @timwoj flagged it for me. I looked into this and found what my vague memory recalled,...

Main todos: 1. Rewrite templater in C++ 2. Put together PR with main compilation functionality 3. Put together PR with low-level ZAM optimization 4. Decide whether we want to support...

Makes sense to me.