Getting Error - ORA-01858: a non-numeric character was found where a numeric was expected

Question

I am getting the error in the below sql:

ORA-01858: a non-numeric character was found where a numeric was expected

SELEC         


        

Answer 1

You can solve the problem by checking if your date matches a REGEX pattern. If not, then NULL (or something else you prefer).

In my particular case it was necessary because I have >20 DATE columns saved as CHAR, so I don't know from which column the error is coming from.

Returning to your query:

1. Declare a REGEX pattern.
It is usually a very long string which will certainly pollute your code (you may want to reuse it as well).

define REGEX_DATE = "'your regex pattern goes here'"

Don't forget a single quote inside a double quote around your Regex :-)

A comprehensive thread about Regex date validation you'll find here.

2. Use it as the first CASE condition:

To use Regex validation in the SELECT statement, you cannot use REGEXP_LIKE (it's only valid in WHERE. It took me a long time to understand why my code was not working. So it's certainly worth a note.

Instead, use REGEXP_INSTR
For entries not found in the pattern (your case) use REGEXP_INSTR (variable, pattern) = 0 .

    DEFINE REGEX_DATE = "'your regex pattern goes here'"

    SELECT   c.contract_num,
         CASE
            WHEN REGEXP_INSTR(c.event_dt, &REGEX_DATE) = 0 THEN NULL
            WHEN   (  MAX (TO_CHAR (TO_DATE (c.event_dt, 'YYYY-MM-DD'), 'MMDD'))
                    - MIN (TO_CHAR (TO_DATE (c.event_dt, 'YYYY-MM-DD'), 'MMDD')))
                 / COUNT (c.event_occurrence) < 32
            THEN
              'Monthly'
            WHEN       (  MAX (
                            TO_CHAR (TO_DATE (c.event_dt, 'YYYY-MM-DD'), 'MMDD'))
                        - MIN (
                            TO_CHAR (TO_DATE (c.event_dt, 'YYYY-MM-DD'), 'MMDD')))
                     / COUNT (c.event_occurrence) >= 32
                 AND   (  MAX (
                            TO_CHAR (TO_DATE (c.event_dt, 'YYYY-MM-DD'), 'MMDD'))
                        - MIN (
                            TO_CHAR (TO_DATE (c.event_dt, 'YYYY-MM-DD'), 'MMDD')))
                     / COUNT (c.event_occurrence) < 91
            THEN
              'Quarterley'
            ELSE
              'Yearly'
         END
FROM     ps_ca_bp_events c
GROUP BY c.contract_num;

Answer 2

The error you are getting is either because you are doing TO_DATE on a column that's already a date, and you're using a format mask that is different to your nls_date_format parameter[1] or because the event_occurrence column contains data that isn't a number.

You need to a) correct your query so that it's not using TO_DATE on the date column, and b) correct your data, if event_occurrence is supposed to be just numbers.

And fix the datatype of that column to make sure you can only store numbers.

[1] What Oracle does when you do: TO_DATE(date_column, non_default_format_mask) is: TO_DATE(TO_CHAR(date_column, nls_date_format), non_default_format_mask)

Generally, the default nls_date_format parameter is set to dd-MON-yy, so in your query, what is likely to be happening is your date column is converted to a string in the format dd-MON-yy, and you're then turning it back to a date using the format MMDD. The string is not in this format, so you get an error.

Answer 3

This error can come not only because of the Date conversions

This error can come when we try to pass date whereas varchar is expected
or
when we try to pass varchar whereas date is expected.

Use to_char(sysdate,'YYYY-MM-DD') when varchar is expected

Answer 4

I added TO_DATE and it resolved issue.

Before modification - due to below condition i got this error

record_update_dt>='05-May-2017'

After modification - after adding to_date, issue got resolved.

record_update_dt>=to_date('05-May-2017','DD-Mon-YYYY')