The index_read() function is used to retrieve a row based on a key:

int ha_foo::index_read(byte * buf, const byte * key, uint key_len, enum ha_rkey_function find_flag)

The *buf parameter is a byte array that the storage engine populates with the row that matches the index key specified in *key. The key_len parameter indicates the prefix length when matching by prefix, and the find_flag parameter is an enumerator that dictates the search behavior to be used.

The index to be used is previously defined in the index_init() call and is stored in the active_index handler variable.

The following values are allowed for find_flag:

HA_READ_KEY_EXACT
HA_READ_KEY_OR_NEXT
HA_READ_PREFIX_LAST          
HA_READ_PREFIX_LAST_OR_PREV
HA_READ_BEFORE_KEY
HA_READ_AFTER_KEY
HA_READ_KEY_OR_NEXT
HA_READ_KEY_OR_PREV    

Storage engines must convert the *key parameter to a storage engine-specific format, use it to find the matching row according to the find_flag, and then populate *buf with the matching row in the MySQL internal row format. For more information on the internal row format, see Section 15.9.6, “Implementing the rnd_next() Function”.

In addition to returning a matching row, the storage engine must also set a cursor to support sequential index reads.

If the *key parameter is null the storage engine should read the first key in the index.