How to find foreign-key dependencies pointing to one record in Oracle?

I have a very large Oracle database, with many many tables and millions of rows. I need to delete one of them, but want to make sure that dropping it will not br开发者_运维百科eak any other dependent rows that point to it as a foreign key record. Is there a way to get a list of all the other records, or at least table schemas, that point to this row? I know that I could just try to delete it myself, and catch the exception, but I won't be running the script myself and need it to run clean the first time through.

I have the tools SQL Developer from Oracle, and PL/SQL Developer from AllRoundAutomations at my disposal.

Thanks in advance!

Here is my solution to list all references to a table:

select
  src_cc.owner as src_owner,
  src_cc.table_name as src_table,
  src_cc.column_name as src_column,
  dest_cc.owner as dest_owner,
  dest_cc.table_name as dest_table,
  dest_cc.column_name as dest_column,
  c.constraint_name
from
  all_constraints c
inner join all_cons_columns dest_cc on
  c.r_constraint_name = dest_cc.constraint_name
  and c.r_owner = dest_cc.owner
inner join all_cons_columns src_cc on
  c.constraint_name = src_cc.constraint_name
  and c.owner = src_cc.owner
where
  c.constraint_type = 'R'
  and dest_cc.owner = 'MY_TARGET_SCHEMA'
  and dest_cc.table_name = 'MY_TARGET_TABLE'
  --and dest_cc.column_name = 'MY_OPTIONNAL_TARGET_COLUMN'
;

With this solution you also have the information of which column of which table is referencing which column of your target table (and you can filter on it).

I always look at the Foreign keys for the starting table and work my way back. The DB tools usually have a dependencies or constraints node. I know PL/SQL Developer has a way to see FK's, but it's been a while since I have used it, so I can't explain it...

just replace XXXXXXXXXXXX with a table name...

/* The following query lists all relationships */ 

select
 a.owner||'.'||a.table_name "Referenced Table"
,b.owner||'.'||b.table_name "Referenced by"
,b.constraint_name "Foreign Key"
from all_constraints a, all_constraints b 
where 
b.constraint_type = 'R'
and a.constraint_name = b.r_constraint_name 
and b.table_name='XXXXXXXXXXXX' -- Table name 
order by a.owner||'.'||a.table_name

I had a similar problem recently, but experienced soon, that finding the direct dependencies is not enough. So I wrote a query to show a tree of multilevel foreign key dependencies:

SELECT LPAD(' ',4*(LEVEL-1)) || table1 || ' <-- ' || table2 tables, table2_fkey
FROM
  (SELECT a.table_name table1, b.table_name table2, b.constraint_name table2_fkey
  FROM user_constraints a, user_constraints b 
  WHERE a.constraint_type IN('P', 'U') 
  AND b.constraint_type = 'R' 
  AND a.constraint_name = b.r_constraint_name 
  AND a.table_name != b.table_name
  AND b.table_name <> 'MYTABLE')
CONNECT BY PRIOR  table2 = table1 AND LEVEL <= 5
START WITH table1 = 'MYTABLE';

It gives a result like this, when using SHIPMENT as MYTABLE in my database:

SHIPMENT <-- ADDRESS
SHIPMENT <-- PACKING_LIST
    PACKING_LIST <-- PACKING_LIST_DETAILS
    PACKING_LIST <-- PACKING_UNIT
        PACKING_UNIT <-- PACKING_LIST_ITEM
    PACKING_LIST <-- PO_PACKING_LIST
...

We can use the data dictionary to identify the tables which reference the primary key of the table in question. From that we can generate some dynamic SQL to query those tables for the value we want to zap:

SQL> declare
  2      n pls_integer;
  3      tot pls_integer := 0;
  4  begin
  5      for lrec in ( select table_name from user_constraints
  6                    where r_constraint_name = 'T23_PK' )
  7      loop
  8          execute immediate 'select count(*) from '||lrec.table_name
  9                              ||' where col2 = :1' into n using &&target_val;
 10          if n = 0 then
 11              dbms_output.put_line('No impact on '||lrec.table_name);
 12          else
 13              dbms_output.put_line('Uh oh! '||lrec.table_name||' has '||n||' hits!');
 14          end if;
 15          tot := tot + n;
 16      end loop;
 17      if tot = 0
 18      then
 19          delete from t23 where col2 = &&target_val;
 20          dbms_output.put_line('row deleted!');
 21      else
 22          dbms_output.put_line('delete aborted!');
 23      end if;
 24  end;
 25  /
Enter value for target_val: 6
No impact on T34
Uh oh! T42 has 2 hits!
No impact on T69
delete aborted!

PL/SQL procedure successfully completed.

SQL>

This example cheats a bit. The name of the target primary key is hardcoded, and the referencing column has the same name on all the dependent tables. Fixing these issues is left as an exercise for the reader ;)

Had a similar situation. In my case I had a couple of records which had ended up with the same ID differing only by case. Wanted to check what dependent records exists for each to know which was easiest to delete/update

The following prints out all child records pointing to the given record, per child table with a count for each table/master record combination

declare
  --
  -- Finds and prints out how many children there are per table and value for each value of a given field
  --

  -- Name of the table to base the query on  
  cTable      constant varchar2(20) := 'FOO';
  -- Name of the column to base the query on
  cCol        constant varchar2(10) := 'ID';
  -- Cursor to find interesting values (e.g. duplicates) in master table
  cursor cVals is
    select id 
    from foo f
    where exists (  select 1 from foo f2 
                    where upper(f.id) = upper(f2.id)
                    and f.rowid != f2.rowid );

  -- Everything below here should just work
  vNum        number(18,0);
  vSql        varchar2(4000);

  cOutColSize   number(2,0) := 30;

  cursor cReferencingTables is
    select
      consChild.table_name,
      consChild.constraint_name,
      colChild.column_name     
    from user_constraints consMast
    inner join user_constraints consChild on consMast.constraint_name = consChild.r_constraint_name  
    inner join USER_CONS_COLUMNS colChild on consChild.CONSTRAINT_NAME = colChild.CONSTRAINT_NAME
    inner join USER_CONS_COLUMNS colMast on colMast.CONSTRAINT_NAME = consMast.CONSTRAINT_NAME
    where consChild.constraint_type = 'R'      
      and consMast.table_name = cTable
      and colMast.column_name = cCol
    order by consMast.table_name, consChild.table_name;
begin

  dbms_output.put_line(
    rpad('Table', cOutColSize) || 
    rpad('Column', cOutColSize) || 
    rpad('Value', cOutColSize) || 
    rpad('Number', cOutColSize)
  );
  for rRef in cReferencingTables loop
    for rVals in cVals loop
      vSql := 'select count(1) from ' || rRef.table_name || ' where ' || rRef.column_name || ' = ''' || rVals.id || '''';
      execute immediate vSql into vNum;
      if vNum > 0 then
        dbms_output.put_line(
          rpad(rRef.table_name, cOutColSize) || 
          rpad(rRef.column_name, cOutColSize) || 
          rpad(rVals.id, cOutColSize) || 
          rpad(vNum, cOutColSize) );
      end if;
    end loop;
  end loop;
end;

I was surprised at how difficult it was to find the dependency order of tables based on foreign key relationships. I needed it because I wanted to delete the data from all tables and import it again. Here is the query I wrote to list the tables in dependency order. I was able to script the deletes using the query below, and import again using the results of the query in reverse order.

   SELECT referenced_table
         ,MAX(lvl) for_deleting
         ,MIN(lvl) for_inserting
   FROM
         ( -- Hierarchy of dependencies
         SELECT LEVEL lvl
               ,t.table_name referenced_table
               ,b.table_name referenced_by
         FROM user_constraints A
         JOIN user_constraints b 
               ON  A.constraint_name = b.r_constraint_name
               and b.constraint_type = 'R'
         RIGHT JOIN user_tables t
               ON  t.table_name = A.table_name
         START WITH b.table_name IS NULL
         CONNECT BY b.table_name = PRIOR t.table_name
         )
   GROUP BY referenced_table
   ORDER BY for_deleting, for_inserting;

Oracle constraints uses Table Indexes to reference data.
To find out what tables are referencing one table, just look for index in reverse order.

/* Toggle ENABLED and DISABLE status for any referencing constraint: */ 

select 'ALTER TABLE '||b.owner||'.'||b.table_name||' '||
        decode(b.status, 'ENABLED', 'DISABLE ', 'ENABLE ')||
       'CONSTRAINT '||b.constraint_name||';' 
  from all_indexes a,
       all_constraints b
 where a.table_name='XXXXXXXXXXXX' -- Table name 
   and a.index_name = b.r_constraint_name;

Obs.: Disabling references improves considerably the time of DML commands (update, delete and insert).

This can help a lot in bulk operations, where you know that all data is consistent.

/* List which columns are referenced in each constraint */

select ' TABLE "'||b.owner||'.'||b.table_name||'"'||
        '('||listagg (c.column_name, ',') within group (order by c.column_name)||')'|| 
        ' FK "'||b.constraint_name||'" -> '||a.table_name||
        ' INDEX "'||a.index_name||'"'
        "REFERENCES"
  from all_indexes a,
       all_constraints b,
       all_cons_columns c
 where rtrim(a.table_name) like 'XXXXXXXXXXXX' -- Table name 
   and a.index_name = b.r_constraint_name
   and c.constraint_name = b.constraint_name
 group by b.owner, b.table_name, b.constraint_name, a.table_name, a.index_name
 order by 1;

 select c.owner, a.table_name, a.column_name, a.constraint_name, 
       c.r_owner as ref_owner, cpk.table_name as ref_table, 
       cpk.constraint_name as ref_pk
from all_cons_columns a 
join all_constraints c on a.owner = c.owner
    and a.constraint_name = c.constraint_name
 join all_constraints cpk on c.r_owner = cpk.owner
    and c.r_constraint_name = cpk.constraint_name
where c.constraint_type = 'r' and c.table_name= 'table_name';