mysql client wishlist

This is my ideal erlang MySQL client.

High Level
 * Groups are process based and multiple group leaders can be spawned and named
 * Supports many connections
 * Connections (and groups of connections) can be named
 * Supports SQL generation
 * Uses records to represent statements and responses

Module layout
 * emysql
   -- Core application module and base functionality
 * emysql_connection
   -- Handles raw connection/socket function to/from mysql server
 * emysql_statement
   -- Handles statement building and parsing.
 * emysql_pool
   -- Handles group manipulation

An example use case:

    {ok, IPWGroup} = emysql:start_group('iplaywow_master'), %% IPWGroup = pid()
    ok = emysql:set_options(IPWGroup, [{reconnect, true}, {log_level, warning}]),
    Connection = {"127.0.0.1", "TheDatabase", "TheUser", "ThePassword", "TheEncoding"},
    ok = emysql:add_connection(IPWGroup, Connection),
    ...
    ok = emysql:prepare(IPWGroup, <<"characters_by_owner">>, <<"SELECT * FROM ...">>),
    ok = emysql:prepare(IPWGroup, <<"characters_all">>, <<"SELECT * FROM ...">>, [{'post-process', fun package_character/1}]),
    Results = emysql:execute(IPWGroup, <<"characters_by_owner">>, [], [{timeout, 5000}, {'post-process', fun package_character/1}]), %% pid(), binary(), [term()], [term()]
    Results2 = emysql:do(IPWGroup, <<"UPDATE atable ...">>, [{timeout, 5000}]), %% pid(), binary(), [term()]

... a little more detail:
 * The actual `emysql` module doesn't manage a specific process or maintain any kind of state. It's just a module that is used to interact with groups and connections.
 * When a group is started, it is given an explicit name and a pid that handles the internal functionally of connection group management is spawned and the name registered to that process.
 * Groups can have options like the log level and if it should reconnect connections that are killed or die.
 * Prepared statements can be given optional defaults like 'post-process'. The post-process function is applied to each row of the data-set returned by that function.

Things that would be nice to get back in `Results`
 * The execution time of the request
 * The connection that it was executed on
 * The noonce of the execution, aka the query sequence id
 * The number of rows affected (if applicable)
 * The id created (if applicable)