How to identify invalid (corrupted) values stored in Oracle DATE columns

问题

Oracle 10.2.0.5

What is the easiest way to identify rows in a table that have "invalid" values in DATE columns. By "invalid" here what I mean is a binary representation that violates Oracle rules for date values.

I recently had an issue with an invalid date stored in a column.

I was able to use a query predicate to find a particular problematic row:

  WHERE TO_CHAR(date_expr,'YYYYMMDDHH24MISS') = '00000000000000'

In the case I had, the century byte was invalid...

 select dump(h.bid_close_date) from mytable h where h.id = 54321

 Typ=12 Len=7: 220,111,11,2,1,1,1

The century byte should be 100 + two digit century. In this case, there was an extra 100 added, as if the century value was "120", making the year "12011". (The only way I know to get invalid DATE values into the database is using OCI, using native 7-byte DATE representation.)

In this case, the TO_CHAR function returned an identifiable string, which I could use for identifying the wonky DATE value.

My question: is there an more general or easier approach (preferably using a SQL SELECT statement) to identify rows with "invalid" values in DATE columns.

回答1:

This identifies invalid months

SELECT rowid,
       pk_column,
       DUMP(date_column, 1010) AS dump1
FROM   table
WHERE  TO_NUMBER(SUBSTR(DUMP(date_column, 1010), INSTR(DUMP( date_column, 1010),
                                              ',', 1, 2
                                                     ) + 1,
                                  INSTR(DUMP(date_column, 1010), ',', 1, 3) - (
                                  INSTR(DUMP( date_column, 1010), ',', 1, 2) + 1
                                  ))) = 0;

Update using the same where clause, I found the month number was zero in these cases.

回答2:

This is a pretty unusual scenario (although I have come across something similar once before). The more common problem is finding invalid dates which are held as strings in a date column. You could adapt the solution for that to your situation, by building your own date validator.

Something like this:

create or replace function is_a_date 
    ( p_date in date )
    return varchar2
is
    d date;
begin
    d := to_date(to_char(p_date,  'SYYYYMMDDHH24MISS'),  'SYYYYMMDDHH24MISS') ;
    if d != p_date then
        return 'not a proper date';
    else
        return 'good date';
    end if;
exception
    when others  then
        return 'not a date';
end;
/

This converts a date into a string and back again. It catches exceptions thrown by date casting. If the end product is not the same as the input date then presumably something got lost in translation; to be honest I'm not sure whether the 12011 date would cast successfully to a string, so this is a belt'n'braces approach. It's a bit tricky writing this utility without some test data!

This query would identify all the non-valid dates:

 select h.id, dump(h.bid_close_date)
 from mytable h 
 where h.bid_close_date is not null
 and is_a_date(h.bid_close_date) != 'good date';

回答3:

Without adding a function, a simple predicate

TO_CHAR(date_col,'YYYYMMDDHH24MISS') = '000000000000'

appears to be satisfactory to identify corrupted values stored in an Oracle DATE column. The addition of a function appears to be unnecessary. Checking for corrupted dates should be able to be done in a SQL SELECT statement, and not require a user to have CREATE FUNCTION privilege on the database.

回答4:

I'd go for a method that can be implemented as a check constraint, probably through checking that the date is in the range of legal values.

However, ask yourself also whether it is valid to have a date of 1066-10-14? It is a legal value, but you probably do not have invoices printed on that day, for example. So you might like to roll the invalid date check into a larger issue of what you really consider to be valid in the context of your application.

回答5:

I had an SQL Error: ORA-01841: (full) year must be between -4713 and +9999, and not be 0 01841. 00000 - "(full) year must be between -4713 and +9999, and not be 0".

So to identify the rows that had a bad date I did the following.

declare
    cursor mydates is select table_pk, your_date_col from table;
    c_date table.your_date_col%type;
    c_pk table.table_pk%type;
    testrow table.your_date_col%type;
begin
    open mydates;
    loop
    begin
        fetch mydates into c_pk, c_date;
        exit when mydates%notfound;
        testrow := TO_TIMESTAMP(c_date,'YYYY-MM-DD HH24:MI:SS');
    exception when others then
        dbms_output.put_line('bad file: ' || c_pk);
    end;
    end loop;
    close mydates;
end;

So all I did was create a cursor, loop through the elements and tested each one and display the identifier so I could easily find the bad rows.

来源：https://stackoverflow.com/questions/8734540/how-to-identify-invalid-corrupted-values-stored-in-oracle-date-columns

标签

sql

Oracle

date

oracle10g