Bug #22264: Warning missing from parse.y hash literals

• Status: Closed
• Assignee: -
• Target version: -
• Backport: 3.3: REQUIRED, 3.4: REQUIRED, 4.0: REQUIRED

[kddnewton (Kevin Newton)]

$ ruby --parser=parse.y -cve '{**{}, a: 1, **{a: 2}}'
ruby 4.1.0dev (2026-04-21T02:07:14Z aot be86d5704f) [arm64-darwin23]
Syntax OK
$ ruby --parser=parse.y -cve '{a: 1, **{a: 2}}'
ruby 4.1.0dev (2026-04-21T02:07:14Z aot be86d5704f) [arm64-darwin23]
-e:1: warning: key :a is duplicated and overwritten on line 1
Syntax OK

[jeremyevans0 (Jeremy Evans)]

I'm not sure whether it is desirable to warn for {a: 1, **{a: 2}}. I would guess that internally, this hits the same warning for {a: 1, a: 2} (which should warn), but for implementation details, it doesn't hit the warning for {**{}, a: 1, **{a: 2}}.

I think we should only warn for keywords in the same hash, not keywords in a splatted hash, even if the splatted hash is a literal. For one, warnings are to help Ruby programmers, and {a: 1, **{a: 2}} is not something a typical Ruby programmer would write (while {a: 1, a: 2} is a fairly common mistake). There is no reason for a Ruby programmer to deliberately keyword splat a literal hash inside a another hash.

[nobu (Nobuyoshi Nakada)]

• Backport changed from 3.3: UNKNOWN, 3.4: UNKNOWN, 4.0: UNKNOWN to 3.3: REQUIRED, 3.4: REQUIRED, 4.0: REQUIRED

I agree with Jeremy's concern, but I think it should be discussed separately.

As for this bug, it stems from an oversight when we allowed ** to be placed in locations other than the end of a hash literal; it should be fixed simply by removing the condition “unless the first element is NULL” from assocs : assocs ‘,’ assoc.
GH-18506

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed

Applied in changeset git|e5132b162737829f744b2b23908b9d08479db977.

---

[Bug #22264] Warn for duplicate keys after leading keyword splat

Leading keyword splats must not suppress duplicate literal key
warnings.