Sometimes it's required to read a lot of records from db server as fast as possible and collect it to dictionary etc. Even tough Unidac (at least with PostegreSQL) is fast, having tdataset, tfield in middle will slow down things a littlebit.
I'm looking to solutions like this pseudo level code
q.sql.text := 'select Barcode,price,name from table where ...';
q.open;
q.useOnlyeLowLevel := true; //disables mapping to fields.
while not q.eof do
begin
lean:= q.lowlevel.readString;
lPrice := q.lowlevel.readFloat;
lname := q.lowlevel.readString;
// do something with lean, lprice, lname
q.next;
end;
Of cource this could implemented with other type of solution.

I'm not sure if this would improve execution time in large dataset.
but if it would the it would be a resource saver for virtualserver running multiple RDP servers which run multiple users which execute that kind of query frequently.

I assume this could be possibile because for example postgeSQLs psql.exe can dump query to csv faster than I can step trough all records with UniDac (without writing anything to csv file)