From: | Bartosz Dmytrak <bdmytrak(at)eranet(dot)pl> |
---|---|
To: | Falk Grossmann <fgrossmann(at)gmail(dot)com> |
Cc: | pgsql-novice(at)postgresql(dot)org |
Subject: | Re: timestamp in log table after update in another table |
Date: | 2012-02-10 15:07:03 |
Message-ID: | CAD8_UcZc6OKZmkbnxHuwS43AHSfRXB3YZwoX=tnjTFDEGZHQxw@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-novice |
Hi,
You can use NEW record in trigger function e.g
CREATE OR REPLACE FUNCTION public."update_log_tbl"()
RETURNS TRIGGER AS
$$
BEGIN
UPDATE data.log
SET last_update = now()
WHERE unit_id = NEW.unit_id;
RETURN NEW;
end
$$
LANGUAGE 'plpgsql'
VOLATILE
CALLED ON NULL INPUT
SECURITY INVOKER
COST 100;
My assumptions are:
1. trigger will be fired after INSERT and/or UPDATE, because for DELETE
there is no NEW record, for DELETE should be modified;
2. This is row level trigger (not statement);
3. Trigger(s) will be defined on tables db_raw_data;
4. You need log transaction start time not an time when record was
modified during query execution. This is the difference between now() and
clock_timestamp() functions;
5. SECURITY INVOKER is intended - user who inserts/modify data
in db_raw_data has to have privilages to modify data in data.log table (is
is secure?).
NEW record represents inserted or modified record (new version) - is NULL
for DELETE
OLD record represents previous (existing) version of modified record - is
NULL for INSERT.
You can find it useful:
http://www.postgresql.org/docs/9.1/static/plpgsql-trigger.html
Regards,
Bartek
2012/2/10 Falk Grossmann <fgrossmann(at)gmail(dot)com>
> Hi,
>
> I have three tables which are frequently updated with new data. If any of
> the tables are updated I need an automated way to create a timestamp
> (indicating which unit_id has been updated/when) in a separate log table.
> The tables have the following rows
>
> Data tables (db_raw_data1,2,3):
>
> aquis_data | aquis_time | unit_id | ... + a number of attributes specific
> to each table.
>
> Log table:
>
> unit_id | timestamp| Unit_id is a primary key and = to unit_id in the
> raw data table
>
> My approach was to use a trigger in the data tables which would feed a
> timestamp to the corresponding unit_id entry in the log table. I have been
> thinking of a function along the following lines (which obviously hasn't
> been successful):
>
> I have googled this issue but haven't found a reference to updating a
> timestamp in another table. Any suggestions?
>
>
> CREATE OR REPLACE FUNCTION public."update_log_tbl"()
> RETURNS TRIGGER AS
> $$
> BEGIN
>
> UPDATE data.log
> SET last_update = timestamp
> WHERE db_raw_data.unit_id = log.unit_id;
>
> end
> $$
> LANGUAGE 'plpgsql'
> VOLATILE
> CALLED ON NULL INPUT
> SECURITY INVOKER
> COST 100;
>
> ALTER FUNCTION public."log_tbl"()
> OWNER TO postgres;
>
From | Date | Subject | |
---|---|---|---|
Next Message | Andy Halsall | 2012-02-10 17:07:21 | Functions and query analysis - EXPLAIN |
Previous Message | Falk Grossmann | 2012-02-10 09:19:48 | timestamp in log table after update in another table |