Elasticsearch version: master @ 65dbc17

ODBC specifies the order in which to return the supported data source types back to the application: the sort order [...] can be generalized as DATA_TYPE first, followed by TYPE_NAME, both ascending.

For the current data types returned, ordering by DATA_TYPE will be sufficient (the result is then also second degree ordered by TYPE_NAME).

FYI, the reason for ordering being that [m]ore than one SQL data type can map to a single type identifier, which can make it difficult (for the application) to determine which data type to use., so the ordering is supposed to also be done by how closely the data type maps to the corresponding ODBC SQL data type. This second criteria isn't necessarily applicable to our current set though (the example given is the definition of two integral types, one auto-incrementable, which would need to be ordered as second, since auto-incrementing property is not specific to an ODBC integral type).