Struct tuple representation codegen #63
Conversation
This will become relevant when we have structs with tuple representation!
| } | ||
|
|
||
| func (g structReprTupleReprGenerator) EmitNodeMethodLength(w io.Writer) { | ||
| // This is fun: it has to count down for any unset optional fields. |
There was a problem hiding this comment.
don't we have special rules for optional fields in tuple representation? like, only the last one can be optional for it to make much sense?
There was a problem hiding this comment.
Yep. There's some comments about that at the top of the file. (There can be several optionals... as long as they're all consecutive, and the matching is greedy.)
I think I was lazy in this snippet: some of the code is very similar to code used in optional-value-presense-counting in structs representations as maps, and I just didn't bother to rewrite it to use the assumptions we're allowed here.
(Optimistically, I was thinking we might be able to make them share code entirely, but I experienced "mental OOM" trying to get there, so this is the non-DRY'd shrapnel. There will be a lot of stuff like this that deserves revisit in the near future: things will either turn out to be very "obvious" candidates for DRYing, or, very much deserving of further specialization -- but I'm frequently giving up trying to tell the difference as I chase feature-completeness right now. Mostly, I'm trying to at least mark them in comments, but I'm sure I often fail.)
Codegen now supports structs with a tuple representation strategy -- that is, a representation that's a list, where the position alone corresponds to each field in the struct, and thus to the expected type and the meaning of the value.
This is a pretty exciting feature to have, because it's often used in protocols where it's desirable to minimize the encoded size (trading down on legibility (and information that could be used for version/feature-detection) to do so -- but that's often a valid trade to make!). (In particular, it's quite heavily used in the Filecoin project; support for this suddenly gets us the ability to model the vast majority of their protocol with our codegen.)