you are probably thinking of b-trees which are unbalanced.  b+tree is o(log(n)) based on depth of tree and you always have to go to bottom to get leaf.

On Mon, Jun 22, 2015 at 7:44 PM, Justin Swanhart <greenlion@gmail.com> wrote:

No, you have to look in the leaf!

On Mon, Jun 22, 2015 at 7:43 PM, Roberto Spadim <roberto@spadim.com.br> wrote:

yeap, i just want know if the unique allow stop the conquer at first result (unique key) or if it continue (non unique key)

2015-06-22 23:42 GMT-03:00 Justin Swanhart <greenlion@gmail.com>:

Hi,

b+tree search is binary search, which is divide and conquer.  Saving scanning a key is like sneezing in a hurricane, you still are o(log(n)) in.

On Mon, Jun 22, 2015 at 7:39 PM, Justin Swanhart <greenlion@gmail.com> wrote:

Hi,

If you do a code inspection you will find that uniqueness is checked on insertion, and the database cares not about it after that.

--Justin

On Mon, Jun 22, 2015 at 7:20 PM, Roberto Spadim <roberto@spadim.com.br> wrote:

and there's optimization with this flag? i see that equal could be optimized cause if we found 1 row we can return to user that row and stop select

 
in other words, if i see a index with primary key or unique columns, could i change the index to unique and get some kind of performace? today i think not ,cause unique key check if it's really unique (for example at insert/update/replace), i'm wrong?

--

Roberto Spadim
SPAEmpresarial - Software ERP

Eng. Automação e Controle