List of functions in

Function Description
Creates a connection information tuple.
This tuple must be passed into connection routines.
Freeing the tuple is not necessary, but is a good idea if you
 create many of them.  You can cache these handles globally.
!!NOTE!! I have seen most people think that this connects to the DB.
  Nowhere does it say this, and in fact it does not.  It only caches
  the connection information, the host/user/pass/etc.

The optional timeout parameter specifies how long connections should wait before
giving up.  If 0, the default (which is undefined) is used.
Frees an SQL handle.
The handle can be to anything (tuple, connection, query, results, etc).
If you free a database connection, it closes the connection as well.
Opens a database connection.
Returns an SQL handle, which must be freed.
Returns Empty_Handle on failure.
Sets the character set of the current connection.
Like SET NAMES .. in mysql, but stays after connection problems.

If a connection tuple is supplied, this should be called before SQL_Connect or SQL_ThreadQuery.
Also note the change will remain until you call this function with another value.

Example: "utf8", "latin1"
Prepares a query.
The query must always be freed.
This does not actually do the query!
Back-quotes characters in a string for database querying.
Note: The buffer's maximum size should be 2*strlen(string) to catch
all scenarios.
Back-quotes characters in a string for database querying.
Note: The buffer's maximum size should be 2*strlen(string) to catch
all scenarios.
Prepares and executes a threaded query.
This will not interrupt gameplay in the event of a poor/lossed
 connection, however, the interface is more complicated and
 asynchronous.  Furthermore, a new connection/disconnection is
 made for each query to simplify driver support.
The handler should look like:
Executes a query.
Returns 1 if the query succeeded.
Returns 0 if the query failed.
NOTE: You can call this multiple times as long as its parent
 connection is kept open.  Each time the result set will be freed
 from the previous call.
Gets information about a failed query error.
Returns the errorcode.
Returns 1 if there are more results to be read,
 0 otherwise.
Tells whether a specific column in the current row
 is NULL or not.
Retrieves the current result.
A successful query starts at the first result,
 so you should not call SQL_NextRow() first.
Passing no extra params - return int
Passing one extra param - return float in 1st extra arg
Passing two extra params - return string in 1st arg, max length in 2nd
 new num = SQL_ReadResult(query, 0)
 new Float:num2
 new str[32]
 SQL_ReadResult(query, 1, num2)
 SQL_ReadResult(query, 2, str, 31)
Advances to the next result (return value should be ignored).
Returns the number of affected rows.
Returns the number of rows total.
Returns the number of columns total.
Returns the name of a column.
Errors on a bad field number.
Returns the number of a named column, or -1 if not found.
Rewinds a result set to the first row.
Returns the insert id of the last INSERT query.
Returns 0 otherwise.
Returns which driver this plugin is currently bound to.
Sets driver affinity.  You can use this to force a particular
 driver implementation.  This will automatically change all SQL
 natives in your plugin to be "bound" to the module in question.
If no such module is found, it will return 0.  This isn't necessarily bad -
 the user might have typed the wrong driver.  Unless your plugin is built
 to handle different driver types at once, you should let this error pass.
Note, that using this while you have open handles to another database
 type will cause problems.  I.e., you cannot open a handle, switch
 affinity, then close the handle with a different driver.
Switching affinity is an O(n*m) operation, where n is the number of
 SQL natives and m is the number of used natives in total.
Intuitive programmers will note that this causes problems for threaded queries.
 You will have to either force your script to work under one affinity, or to
 pack the affinity type into the query data, check it against the current, then
 set the new affinity if necessary.  Then, restore the old for safety.
Returns the original query string that a query handle used.
For queries which return multiple result sets, this advances to the next
result set if one is available.  Otherwise, the current result set is
destroyed and will no longer be accessible.

This function will always return false on SQLite, and when using threaded
queries in MySQL.  Nonetheless, it has the same effect of removing the last
result set.
This function can be used to find out if a table in a Sqlite database exists.
(updated for newer API)
Use this for executing a query where you don't care about the result.
Returns 0 on failure, 1 on success
Use this for executing a query where you don't care about the result.
Returns 0 on failure, 1 on success
Use this for executing a query and not caring about the error.
Returns -1 on error, >=0 on success (with number of affected rows)
This function has no description.