Gas cost difference between uint[] and mapping(uint => uint)

Question

When mapping for example a large integer to some integer,

uint[] index;
index.length = 1000000;
index[600001] = 4;

is it cheaper to use

mapping(uint => uint) index;
index[600001] = 4;

? one benefit of using uint[] is that it can be deleted, happens to be useful in a dApp I work on.

Answer 1

They would be the same, with one modification.

You would go

uint[1000000] index;
// index.length = 1000000; not needed

The dynamic index you used is only slightly more expensive than the fixed-length array suggested because it needs another 32-byte word for the array length and the extra step of setting it.

Both methods use a 32-byte key to point to a 32-byte word. Both methods will avoid writing unnecessarily and values default to 0 in indexes that haven't been written to.

Neither method benefits from reducing the precision of the values pointed to (to, say, uint8) because a 32-byte word is the smallest space they can address.

Hope it helps.