Does making a field unique make it indexed?

Question

If I make a unique constraint on a field, do I also need to make an index on that field in order to get a scalable insert time? Or is this done for me (even if the index it uses isn't publicly accessible?)

Specifically, I'm working with Apache Derby for prototyping, although I will probably be moving it to MySQL in the semi-near future. I'm also hoping there might be something in the SQL standard that says something about this.

I will never have a need to search by this field, so I would rather not make a useless index. But I'd rather have a useless index than have an O(n) insert time.

Answer 1

PRIMARY KEY >= UNIQUE >= INDEX == KEY

InnoDB data is ordered by the PK. MyISAM PK acts the same as UNIQUE.

INSERT must add a "row" to each and every index (of any kind) that you have. This takes some time. (Usually, not enough time to matter.) Indexes are all stored in BTree format. MyISAM BTree blocks are 1KB; InnoDB uses 16KB.

Inserting into InnoDB updates the PK and the data simultaneously.

Inserting into MyISAM usually "appends" the data to the .MYD. Separately, it adds a row to the PK (if any).

INSERT has to first verify that there is no duplicate key for any PRIMARY or UNIQUE key. This is done by using the index. And, hence, why the UNIQUE and FOREIGN KEY CONSTRAINTs really build indexes. This is O(logN), but usually CPU, not I/O, because of efficient caching.

Answer 2

--EDIT--

My original answer (below) is probably not useful to you at all because it does not address the question of unique constraints. As others have said, these constraints are usually implemented with an implied unique index. In special cases this might not be true (eg disable novalidate for Oracle).

The question could be: Is it possible to enforce uniqueness without an index? Generally speaking the answer is no though in some cases a Clustered Index will mean that the index and the table are the same object.

--END EDIT--

You said "I'd rather have a useless index than have an O(n) insert time.", but in general databases do not have O(n) insert time. There are two cases to consider:

1. A normal table with or without indexes:

   New rows are dumped at the top of the heap. The RDBMS probably only looks at 1 block, so not just O(1) but very small O(1).

   If the table has indexes, a pointer to the row will be added to each. This will usually be an O(log(n)) operation.

2. A table with some sort of clustering going on, eg an Index Organized Table or cluster for Oracle, or a Clustered Index for SQL Server and others:

   New rows are inserted into a particular block, which may cause the block to split or overflow, but whatever happens it is still O(log(n)) or better, caused by the b-tree or similar structure used to find the block.

Answer 3

To answer to question in bold: Yes, making a field unique does index it like s primary key. In fact, I had discussed this in another question with regard to Primary Keys Having Its Own Name to distinguish it from other Unique (Candidate) Keys.

As for constraints, indexes are created for you so that constraint paradigm is setup. You should be able to remove duplicate indexes, even UNIQUE keys, as long as the constraint you made does not reference other UNIQUE keys you personally made apart from the constraint paradigm.

You may never have to search for this field, but MySQL will sure have to as its path to determine validity of keys and determine how to go about performing ON DELETE CASCADE and ON UPDATE CASCADE operations.

The UNIQUE index simply guarantees uniqueness of tuples (singletons, pairs, triplets, ..., n-tuples, etc) in every row in the table.

It is at your discretion to remove such duplicate indexes, provided you do not break the constraint paradigm you wan the table to have.