There is something special about INCR. Why do we provide such an operation if we can do it ourself with a bit of code? After all it is as simple as:
x = GET count
x = x + 1
SET count x
The problem is that doing the increment in this way will only work as long as there is a single client using the key. See what happens if two clients are accessing this key at the same time:
- Client A reads count as 10.
- Client B reads count as 10.
- Client A increments 10 and sets count to 11.
- Client B increments 10 and sets count to 11.
We wanted the value to be 12, but instead it is 11! This is because incrementing the value in this way is not an atomic operation. Calling the INCR command in Redis will prevent this from happening, because it is an atomic operation.
All the Redis operations implemented by single commands are atomic, including the ones operating on more complex data structures. So when you use a Redis command that modifies some value, you don't have to think about concurrent access.