Find all tables updated on a specific date - sql

Question

I'm using an Oracle DB, and I'm trying to find all tables that were updated on a certain date. All of the tables that track updates have a column called DT_UPDATE. I've been trying this:
SELECT * FROM
(SELECT TABLE_NAME FROM ALL_TAB_COLUMNS WHERE COLUMN_NAME = 'DT_UPDATE')
WHERE DT_UPDATE = <date>
But get this error:
ORA-00904: "DT_UPDATE": invalid identifier
00904. 00000 - "%s: invalid identifier"
*Cause:
*Action:
Error at Line: 3 Column: 7
I've also tried aliasing the nested Select clause.

Answer 1

As #zaratustra said, you have to use dynamic SQL. You can do something like this:
set serveroutput on
declare
counter number;
begin
for r in (
select owner, table_name
from all_tab_columns
where column_name = 'DT_UPDATE'
) loop
execute immediate 'select count(*) from "'
|| r.owner || '"."' || r.table_name
|| '" where dt_update = :dt and rownum = 1'
into counter
using date '2014-07-07';
if counter = 1 then
dbms_output.put_line(r.table_name);
end if;
end loop;
end;
/
For each table_name (and owner, for completeness) identified in all_tab_columns as having a column called dt_update, a new dynamic select is generated, in the form:
select count(*) from "<owner>"."<table_name>"
where dt_update = date '2014-07-07'
and rownum = 1;
The rownum = 1 filter lets the query execution stop as soon as a matching row is found; since you said you want to know which tables were updated, not how many rows or exactly which rows, if one row matches then that is all you really need to know. So for every table the dynamic query gets either 0 or 1.
For any tables that have at least one row matching the date, this printd the table name using dbms_output, so you have to have that enabled - with set serveroutput on, or with the DBMS_OUTPUT panel in SQL Developer, or your favourite client's equivalent.
If I create some tables with that column, but only populate one with the date I'm looking for:
create table tab1 (dt_update date);
create table tab2 (dt_update date);
create table tab3 (dt_update date);
insert into tab1 values (trunc(sysdate) - 1);
insert into tab2 values (trunc(sysdate));
... then running my anonymous block produces:
anonymous block completed
TAB1
Use your own target date, obviously. This assumes your date field doesn't contain a time component. If it does then you'd need to turn that into a range to cover the whole day.
You could also turn this into a pipelined function that takes a date as an argument; this also handles date fields with time elements:
create or replace function get_updated_tables(p_date date)
return sys.odcivarchar2list pipelined as
counter number;
begin
for r in (
select owner, table_name
from all_tab_columns
where column_name = 'DT_UPDATE'
) loop
execute immediate 'select count(*) from "'
|| r.owner || '"."' || r.table_name
|| '" where dt_update >= :dt1 and dt_update < :dt2'
|| ' and rownum = 1'
into counter
using p_date, p_date + interval '1' day;
if counter = 1 then
pipe row (r.table_name);
end if;
end loop;
end;
/
Then you can query it with:
select column_value from table(get_updated_tables(date '2014-07-07'));
COLUMN_VALUE
------------------------------
TAB1
Dynamic SQL is interesting, as you said in a comment, but should only be used when necessary. The generated statement can't be parsed until it's executed, so you might not spot syntax or other errors until run-time. Also make sure you use bind variables for values (but not object names) to avoid SQL injection.

Answer 2

Let's assume we have three tables with the field dt_update, and each of them has one record (doesn't matter if more):
create table tt1 (
dt_update date
);
insert into tt1 values (sysdate);
create table tt2 (
dt_update date
);
insert into tt2 values (sysdate - 1);
create table tt3 (
dt_update date
);
insert into tt3 values (sysdate - 2);
This PL/SQL anonym block prints only tables' names that have record with the value of the column dt_update more than or equals today:
declare
type table_names_tp is table of user_tables.table_name%type index by binary_integer;
table_names table_names_tp;
l_res number(1);
l_deadline date := to_date('2014-07-08', 'YYYY-MM-DD');
begin
select table_name
BULK COLLECT INTO table_names
from user_tab_columns
where lower(column_name) = 'dt_update'
;
for i in table_names.first..table_names.last
loop
execute immediate 'select count(*) from dual where exists (select null from ' || table_names(i) || ' where dt_update >= :dead_line)'
into l_res
using l_deadline;
if l_res = 1
then
DBMS_OUTPUT.put_line('Table ' || table_names(i) || ' was updated after ' || l_deadline);
end if;
end loop;
end;
You can use this code as an example to start writing your code. Pay carefully attention to protect yourself from SQL injections, DO NOT(!) use concatenation of your values, always use bind variables instead. It also helps you to store a cached query plan in SGA, the application will read data from the SGA area and perform soft parsing.