Identify a table with maximum rows in Oracle

Question

I have a set of tables in Oracle and I would like to identify the table that contains the maximum number of rows.

So if, A has 200 rows, B has 345 rows and C has 120

Answer 1

David Aldridge correctly points out that querying all_tables could give incorrect results due to missing or stale table statistics. But there is also a problem with using user_segments; Deleted blocks beneath the high water mark would still be counted for the size of the table.

Example:

SQL>create table t as select * from all_objects

Table created.

SQL>select blocks, bytes from user_segments where segment_name = 'T';

    BLOCKS      BYTES
---------- ----------
       768    6291456

SQL>delete from t

52676 rows deleted.

SQL>commit;

Commit complete.

SQL>select count(*) from t;

  COUNT(*)
----------
         0

SQL>select blocks, bytes from user_segments where segment_name = 'T';

    BLOCKS      BYTES
---------- ----------
       768    6291456

SQL>truncate table t;

Table truncated.

SQL>select blocks, bytes from user_segments where segment_name = 'T';

    BLOCKS      BYTES
---------- ----------
         8      65536

Answer 2

You could get the same result with one trawl of the data like so:

SELECT     DISTINCT
           FIRST_VALUE ( t.owner )
             OVER ( ORDER BY t.num_rows DESC NULLS LAST )
                                                 owner,
           FIRST_VALUE ( t.table_name )
             OVER ( ORDER BY t.num_rows DESC NULLS LAST )
                                                 table_name,
           FIRST_VALUE ( t.num_rows )
             OVER ( ORDER BY t.num_rows DESC NULLS LAST )
                                                 num_rows
FROM       all_tables                            t

Answer 3

select max(select count(*) from A union select count(*) from B...)

should work.

edit: if you want something dynamic, you can build a string in PL/SQL with each "count(*)" subquery (for example, listing the table names from USER_TABLES), and then execute te main query with:

execute immediate 'select max('||subquery||')'

Answer 4

Here's another method, likely to be much slower than simply getting ALL_TABLES.NUM_ROWS, but it doesn't depend on statistics having been gathered and gives exact, current values -- although how current depends on how long it takes to run!

-- For running in SQLPlus you need this to see the output.
-- If running in Toad or similar tool, output is enabled by default

    SET SERVEROUTPUT ON SIZE 100000

    DECLARE
      l_rows  INTEGER;
      l_max_rows  INTEGER := 0;
      l_table_name  all_tables.table_name%TYPE := NULL;
    BEGIN
      FOR table_record IN (SELECT table_name FROM all_tables) LOOP

        EXECUTE IMMEDIATE 'SELECT COUNT(*) FROM '||table_record.table_name
          INTO l_rows;

        IF l_rows > l_max_rows THEN
          l_max_rows := l_rows;
          l_table_name := table_record.table_name;
        END IF;
      END LOOP;

      IF l_table_name IS NULL THEN
        dbms_output.put_line( 'All tables are empty' );
      ELSE
        dbms_output.put_line( 'Table ' || table_record.table_name || 
                              ' has ' || TO_CHAR(l_max_rows) || ' rows'
                            );
      END IF;
    END;
    /

Answer 5

Given that you said you were using Oracle I would just query the meta-data.

select table_name, max(num_rows) from all_tables where table_name in ('A', 'B', 'C');

Just saw your edit. Just run the above without the where clause and it will return the largest table in the database. Only problem may be that you might get a SYS$ table or something. Alternately if you are just doing this for your own knowledge just do

select table_name, num_rows from all_tables order by num_rows; 

and you'll see what the biggest are.

Answer 6

This is a query to get the maximum numbers of rows there in a database table..

select table_name, num_rows from USER_TABLES
where num_rows = (select  max(num_rows) from
(select table_name, num_rows from USER_TABLES));