SUMMARY OF CODE REVIEW -- 08/15/1996: Mediator Architecture Overview

Presenter: Yannis Papakonstantinou
Note-taker: Ramana Yerneni

Items of discussion:

General:

We had a code review session on the Mediators, presented by Yannis. In this session, we discussed the high level overview of the mediator architecture. We will have follow up sessions on the actual mediator code reviews.

There were no documents or code sent out for reading before we met for the review session. This was more of a design review and no code was actually read or reviewed. No design documents were read before we met for the review or during the review session. Instead, Yannis presented the high level design on the board.

Action Items:

  1. The reading of the mediator specification should be part of TM_Init instead of the main function.
  2. The parsing and normalization of the mediator specification should be part of TM_Init instead of TM_Query. Yannis thinks there are complex problems to be solved before this can be done. However, he feels that it is a matter of "debugging" a bit and being more careful with reuse of generated variable numbers.
  3. queryResult must be released (its memory freed) at the beginning of TM_Query. Otherwise, we will have memory leak.
  4. The execution engine should keep track of the TM_Objects that it gets from executing the query nodes. It should release all these objects when initializing the engine at the beginning of executing a new query plan.

We will have more code review sessions to go into the details of the mediator code. First, we will have a session on the main/driver code of the mediator. This is scheduled for Wednesday, the 21st of August. It will be similar to the session on 8/14/96 for the wrapper driver code. Then, we will decide on which modules should be reviewed along with a schedule for the reviews.


Ramana Yerneni