Hello all,
In working through the kinks of the example with Paul in the latest milestone, it's
clear that the text translator / file resource adapter functionality is as confusing as
ever for a user. To simplify our approach to text we should at least remove the use of a
separate descriptor file. There are two logical successors to that approach:
1. move the additional metadata onto the model as extension properties (HEADER_ROW,
DELIMITER, etc.)
2. introduce a system function similar to Postgres' copy from
(
http://www.postgresql.org/docs/8.1/static/sql-copy.html) that would essentially reuse the
connector logic, but not require a table based physical model of the files.
Assuming the use of a simple file translator/ resource adapter procedure to get files, the
downside to either approach is that relative file names would now be stored on either the
model or in a function.
I support the second, since it completely removes the need for a heavy-weight text
translator and is feasible before RC1.
Steve