Roger Peppe

Here's a graph illustrating allocated bytes vs n for json and yaml encoders:  Here it is with a log y-axis, which seems to...

FWIW another possibility I've had in mind in the past is to make these schemas available in the standard library. Then they'd be available to regular CUE code, not just...

> Note that currently schema.cue specifies a "strict" variant. It is my understanding that we only use the strict variant, and thus this is what I would expect to be...

Is that perhaps because a map in Go already has nil as a possibility so having a pointer to a map is arguably redundant?

Would you think it's OK if the error message didn't explicitly mention the word "suffix"; something like this? ``` cannot form module version: major version of "mod.example/blah@v0" does not match...

> Presumably we don't just drop what the actual declared module path is. In many code paths, we are working with just dependencies and their versions in canonicalized form, and...

I've been investigating this, and I'm not entirely sure that the "expected" error message is good. A path with no major version suffix does not _always_ imply `@v0` (for example,...

Just to throw out one other thought I had in passing when encountering this issue: IME this issue is most commonly seen when unifying numeric results when there's a cycle...

Here's another example directly derived from [an example found in the wild](https://github.com/SchemaStore/schemastore/blob/9f36b9cca86745a19ed5a74f4b660af5b6684fd6/src/schemas/json/airlock-microgateway-3.0.json#L402-L407), which seems related: ``` exec cue def schema.json -- schema.json -- { "$schema": "http://json-schema.org/draft-04/schema#", "type": [ "string", "null"...

I wonder if someone accidentally pushed a git tag `v0.4.78` and then deleted it. I'm not sure there's anything we can do now other than publish v0.4.79 containing a retraction...