fix(visual-editing-csm): decode all-numeric _key as a key, not an array index

[voidhrithik]

Fixes #3467.

An all-numeric _key was decoding as an array index, so reordering one of those items threw Found no matching array element to replace.

The encode side writes an index and a _key identically:

if (typeof segment === 'number') str += `${segment}`      // array:5
if (segment._key)                str += `${segment._key}` // array:342330179449

which leaves urlStringToPath to break the tie, and RE_SEGMENT_WITH_INDEX matched any run of digits and got tested first. So the key came back a number, decodeSanityString stringified it to sections[342330179449], getArrayItemKeyAndParentPath found no _key, and the reorder ran remove(~~"342330179449"), which int32-truncates to -1267204231.

The fix is the quantifier:

-const RE_SEGMENT_WITH_INDEX = /^([\w-]+):(0|[1-9][0-9]*)$/
+const RE_SEGMENT_WITH_INDEX = /^([\w-]+):(0|[1-9][0-9]{0,8})$/

Nine digits max, so anything longer falls through to the key branch. That covers every randomKey output, since those are 12 characters. Leading-zero keys already worked. A short hand-authored key like 12345 stays ambiguous, and no magnitude rule fixes that one, so this keeps to the randomKey case the issue reports.

I started out with a MAX_ARRAY_INDEX = 2 ** 32 - 2 constant and a second condition in the loop. Same behaviour, four more lines and a second place to look, so I dropped it. Happy to put it back if you would rather have the bound spelled out than the regex short.

Testing

Three cases added to urlStringToPath.test.ts. The existing array:123 assertion still holds.

I also ran the chain from the issue, createDataAttribute(['sections', {_key}]) through decodeSanityNodeData to getArrayItemKeyAndParentPath, against this branch and against main. On main:

_key=342330179449   sections[342330179449]          explicit=false
_key=123456789012   sections[123456789012]          explicit=false
_key=999999999999   sections[999999999999]          explicit=false
_key=000000000001   sections[_key=="000000000001"]  explicit=true
_key=abc123def456   sections[_key=="abc123def456"]  explicit=true

All six pass on this branch, and ['items', 0], ['items', 5], ['items', 999999999] decode as indexes either way.

[changeset-bot]

🦋 Changeset detected

Latest commit: 819c8a7

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 5 packages

Name	Type
@sanity/visual-editing-csm	Patch
@sanity/core-loader	Patch
@sanity/react-loader	Patch
@sanity/visual-editing	Patch
@sanity/svelte-loader	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

@voidhrithik is attempting to deploy a commit to the Sanity Sandbox Team on Vercel.

A member of the Team first needs to authorize it.

[voidhrithik]

@stipsan could you approve the CI run when you get a chance? It's behind the first-time-contributor gate, so nothing has run yet.

I checked it against main too, where the all-digit keys come back with hasExplicitKey false, same as the issue describes. Index paths decode the same either way.

Not chasing a review, just the button.