Return-Path: pg_adm@postgres.berkeley.edu Received: by postgres.Berkeley.EDU (5.61/1.29) id AA10897; Fri, 13 Nov 92 09:59:32 -0800 Message-Id: <9211131759.AA10897@postgres.Berkeley.EDU> From: Kent Archie Subject: no subject (file transmission) To: postgres@postgres.berkeley.edu Sender: pg_adm@postgres.berkeley.edu To: postgres@postgres.berkeley.edu Date: Fri, 13 Nov 92 11:55:45 CST From: Kent Archie X-Mailer: ELM [version 2.3 PL0] A month or so ago I posted a question similar to this one. I got a response that was helpful but the problem is not quite solved so I will try a more detailed question. Imagine I have two tables, the first is a table of source code (SRC) deltas and has three fields: filename text delta text delta_num int4 The second table is dataflow graph deltas (DFG) and has the same fields. The DFG table is dependent of the SRC table in the sense that if tuples are added to the SRC table, corresponding ones should be added to the DFG table. Assume that the highest SRC delta number is 4 and the highest DFG delta number is 2. What I want to happen is this: If a query is run against DFG for delta number 4, the system (Postgres, I hope) detects this and issues commands to generate the missing DFG deltas from the corresponding SRC deltas. What I have tried on the real system that this approximates is to set up a query rule that is triggered on retrieve for the DFG table. This rule executes C code to query the SRC table for the tuples between 3 and 4 and calculate the corresponding DFG tuples. These are then appended to the DFG table and the query completed. The problem I saw with this is that the append generated as a result of the query could not be completed, apparently because the table was in use by the original DFG query. It is as if Postgres put a read-lock on the DFG table when the original query was executed and this prevents the appends from happening. This kind of activity is crucial to my application. I have a way around it but it relegates Postgres to being a static repository of data. I was hoping to get it more actively involved in the processing of the data. Any ideas? **kent Kent Archie kca@iwtqg.att.com