Set cue environment within source #3754
-
In using 0.12 i have had trouble with "incDependant already closed" errors. Im trying to distill a reproduction, but may not have time to chase this further. This can be avoided in my case by using CUE_EXPERIMENT=evalv3=0 or v0.11 Stepping back, cuelang is not 1.0 software yet, so i can expect that things will change and the road as a user will have its bumps. Ideally i will isolate a bug, create an issue, and it will be fixed . But there are other competing priorities. I do want to use cue in production, and the fast way to do it is to just turn evalv3 off as i need to. The idea is to allow setting environment variable in cue files. This allows for a quick workaround for users like myself. If cue encounters in any source file a environment setting that disables evalv3 then it will be disabled. This would save me manually putting env vars in my Makefiles and other places that invoke cue for specific cue source files The syntax may look like "environment (ENV=value)" or similar |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 4 replies
-
Thanks for raising this for discussion, @WaywardWizard. Stepping back for one second, we are working flat out towards the goal of getting Returning to your suggestion.
Setting the evaluator at a file/package level is not something that would work, based on the current structure of the evaluator: we never mix evaluator versions for a given configuration. But I think I understand where you are coming from. One thing we have toyed with is the idea of enabling the new evaluator based on the
For example, we could enable the new evaluator if all modules that participate in an evaluation (we need to consider dependencies too) have Perhaps you could help us understand the problems you are experiencing first of all - we can ensure they are being worked on, but also those problems and the situation might help guide us on the right approach for flipping the default to be |
Beta Was this translation helpful? Give feedback.
Thanks for raising this for discussion, @WaywardWizard.
Stepping back for one second, we are working flat out towards the goal of getting
evalv3=1
to be the default, by eliminating bugs in the new evaluator so that people do not have to revert to evalv2. Please let us know if there are bugs/things getting in the way that are preventing you from making the flip entirely.Returning to your suggestion.
Setting the evaluator at a file/package level is not something that would work, based on the current structure of the evaluator: we never mix evaluator versions for a given configuration.
But I think I understand where you are c…