Had a vague memory of seeing this before, but can't find any jq issue about... but turns i reported it for gojq itchyny/gojq#194 some time ago :)

I guess jq should behave the same as gojq?

I don't know what it should do but I'd expect it to either fail or be a lawful lens.

014b45b added the check of whether the index is an integer, and if not return null. We should apply the same logic in jv_set(). This was #239. The intent seems to have been to require indices to be in the INT_MIN..INT_MAX range, and the integer check seems to have been glossed over.

• This is fine: .[1.5] == null and .[1.5] = "x" -> error
• But so is this: .[1.5] == .[1] and .[1.5] = "x" -> set .[1] = "x"
• And so is this: .[1.5] -> error and .[1.5] = "x" -> error

What's not fine is the current inconsistency.

Which behavior is best?

I think it should behave the same as out-of-bounds negative indices (so the first option). This also reduces the extent of the breaking change.

We're also inconsistent w.r.t. the indices in slices:

: ; ./jq -cn '"foobar"|.[1.5:3.5]'
"oob"
: ; ./jq -cn '"foobar"|.[1.5:3.5] = "xyz"'
jq: error (at <unknown>): Cannot update field at object index of string
: ; ./jq -cn '[range(10)]|.[1.5:3.5] = "xyz"'
jq: error (at <unknown>): A slice of an array can only be assigned another array
: ; ./jq -cn '[range(10)]|.[1.5:3.5] = ["xyz"]'
[0,"xyz",4,5,6,7,8,9]

Fun!

Another inconsistency:

: ; jq -n '[1,2,3] | .[1.5]'
null
: ; jq -cn '"foobar"|.[1.5]'
jq: error (at <unknown>): Cannot index string with number

Got curious about .[{start:1,end:2}] slicing syntax. Seems to be a difference code path ending up here i think https://github.com/jqlang/jq/blob/master/src/jv_aux.c#L16?

Ah sorry for the noise, now i see how it works 👍