Refactor traces in the compiler #1429
Conversation
We were only using shadow variables for this purpose, and it was a lot of additional complexity.
|
Not sure what is going on with the MacOS CI setup. Will try again later in the day, perhaps is some instability. |
There was a problem hiding this comment.
LGTM! Added a comment about the remaining FIXME.
This one seems pretty subtle, so I'm not sure I have a lot to add due to my lacking context, but I didn't spot anything that seemed obviously strange to me.
The MacOS test seems to be consistently failing, I also retried it; not sure what is going on there, the log is pretty sparse, but it seems some part of the tests passes, and it fails during a setup step?
| @@ -224,6 +224,7 @@ const otherOperators = [ | |||
| { name: 'fail', effect: propagateComponents(['read', 'update'])(1) }, | |||
| { name: 'assert', effect: propagateComponents(['read'])(1) }, | |||
| { name: 'q::debug', effect: propagateComponents(['read'])(2) }, | |||
| { name: 'q::lastTrace', effect: parseAndQuantify('Pure') }, // FIXME: Should be in run mode | |||
There was a problem hiding this comment.
just making sure, is this FIXME something that should be fixed in this PR?
There was a problem hiding this comment.
Nope, we don't really have a run mode yet. I didn't find a good issue to link, so I just kept it as a reminder.
I just changed the problematic jq setup action to a different one. There are plenty available, and this one worked :) |
Hello
Towards #1379
Many ideas involving debugging and mbt-oriented features require updates in the compiler. Currently, the
compilerImpl.tsfile is 1833 lines long, and I guess modern LSP solutions are not prepared for files this big, because my editor is struggling a lot!So I'm trying to improve this file a bit. I start by removing the concept of shadow variables. I think we used to have many shadow variables, and therefore we introduced this abstraction, but right now we only have the trace, and I think it is quite unecessary to have it instead of simply keeping the trace in an attribute.
I also extracted trace manipulation logic to a new file.
Behavior shouldn't change, this is simply a refactor.