pqxx column names in multi table request

Question

This question is specific to libpqxx.

Given an SQL statement like the following:

string s = "SELECT a.foo, b.bar FROM tableOne a, tableTwo b WHERE a.X=b.X"

and sending it to a pqxx transaction:

trans.exec(s.c_str(), s.c_str());

What names will the columns have in the result field?

In other words, assuming 1 row is selected:

pqxx::result::const_iterator row = result.begin();
int foo = row->at(FOO_COLUMN).as<int>();
int bar = row->at(BAR_COLUMN).as<int>();

What values should FOO_COLUMN and BAR_COLUMN have? Would they be "a.foo" and "b.bar", respectively?

If the SQL statement renamed the variables using the "as" keyword, then I suppose the column name would be whatever "as" set it to, is that right?

Normally I would try an SQL and print the column values, but as I am developing both the software and the database itself, doing that test is not very easy right now.

Thanks!

Answer 1

The names are going to be foo and bar. If they were aliases in the query, then the aliases would be returned, the original names being lost.

Column names in results never include table names.

If they were named tablename.colname, it would be ambiguous anyway because SELECT 1 as "foo.colname" is valid and produces a column foo.colname despite the fact there is no foo table.

The way to identify the table from which a column originates, when it applies, is to call pqxx::result::column_table(column_number) which returns the oid of the table. The table's name is not available directly but could be queried in pg_class with the oid.

Also note that column names don't have to be unique within a resultset. SELECT 1 AS a, 2 AS a is valid and produces two columns with exactly the same name.