[Go] *array.Binary.ValueOffset doesn't work correctly

[Korzhenevskyi]

Describe the bug, including details regarding any error messages, version, and platform.

https://github.com/apache/arrow/blob/main/go/arrow/array/binary.go#L76 -- this place is incorrect. For example, valueOffsets contain 6 offsets for the array with 5 elements, but calling array.ValueOffset(5) will cause the panic although the element exists.

It must be

func (a *Binary) ValueOffset(i int) int {
	if i < 0 || i > a.array.data.length {
		panic("arrow/array: index out of range")
	}
	return int(a.valueOffsets[a.array.data.offset+i])
}

similar to strings https://github.com/apache/arrow/blob/main/go/arrow/array/string.go#L66.

Component(s)

Go

[zeroshade]

ValueOffset returns the starting offset for the value at index i. ValueOffset(5) is correct to panic for an array of 5 elements, there's no element at index 5 (index 4 is the last element in the array).

[Korzhenevskyi]

@zeroshade I am talking about inconsistency between binary and string.

I had a loop over binary and string arrays of the same length, and it raised panic for binary. However, when I debugged, I could see the element at the requested index, so the if-statement was not wrong. I think the logic for strings is correct, because as I said the element existed in the array but could not be accessed due to panic.

[zeroshade]

Semantically, the logic in for the Binary array is correct, not the String one.

You should never be calling ValueOffset for an index larger than the size of the array itself. The method returns the offset for the start of the value for the index requested. There is no value at the index of array length + 1.

If you need that value you could iterate through the slice provided by calling ValueOffsets() or by doing len(ValueBytes()).