Repository navigation
fix: rm the dummpy bad case guard for catalog inputs - #55
Conversation
huskcass
left a comment
huskcass
left a comment
There was a problem hiding this comment.
No blockers. Checked at 5704438: catalog, compose and site tests pass and the go CI job is green. The internal/server tests need the sibling swiss-catalog checkout, so I didn't run them locally.
Design check: this is a net cut (about 60 lines of parse-time validation gone) with no new options or modes. Dropping checkInputs from site.Parse is safe, since a malformed key never matches a declared input, and compose still refuses a bad placeholder or a type/pattern miss at plan time (covered by the new TestASiteAnswerWithAnUnknownPlaceholderRefusesThePlan).
Low: checkInputs in catalog now accepts values.<key> being a map as well as a scalar (at != in.Key). A scalar input answer would then replace that whole subtree in the derived layer. Either keep the old refusal for a map at the key, or leave it, since no real variant does this.
Low: when extra sets the same key, the site's input answer is now silently ignored. That matches the doc change, but a plan-time note would make it less surprising. Not needed if you'd rather keep it simple.
Verdict: good simplification, fine to merge.
— Review Golden
No description provided.