将具有多种日期格式的字符列转换为日期

Question

I'm trying to update a date column with from a varchar column

update tbl set columnA = columnB

Here, columnA is a varchar data type and columnB is a date data type. columnA has various types of date formats, for instance, 09302012, 9/30/2012, 2012-09-30 and more different types

How can I write a single query to update the column with the various types of date formats in single query.

EDITED:::

sorry about the mess up.. i just Realized that these are individual(pieces) updates in ssis package...

we have diff types(6) of makes and for each make they send different files with different date formats

for type 1 contains date format like 09/22/2011 9/22/2011 and 2012-09-22

and the rest all types follows same format..its 09222012

so now i need to write query for individual types...(straightly speaking only two logics one for type 1 and the other for rest all types)

the first query logic contains a case statement for three formats and the second query logic contain logic for other format...

the end result should show up like 2012-09-22 00:00:00 (ie yyyy-dd-mm hh:mm:ss) can u help me out

am a T-sql guy and dont know any thng aby pl-sql(if its in t-sql i would directly do i with convert and substring)

Answer 1

You don't.

Firstly you do it properly next time and store dates in a DATE data type; if this is supplied data then you yell ¹ at your suppliers.

The simplest way to clean your data would be to create a function that tests if a date is in a certain format:

create or replace function is_date ( 
      P_String in varchar2
    , P_Date_Format in varchar2
      ) return number is

   l_date date;

begin

   l_date := to_date(P_String, P_Date_Format);

   return 1;
exception when others then
   return 0;
end;

Next you pick a format model and update just that one.

update my_table
   set date_column = to_date(char_column, 'yyyy-mm-dd')
 where is_date(char_column, 'yyyy-mm-dd') = 1

You then have to pick a different format model and do it all over again until you don't have any records in your date column that are NULL.

_{1. Yelling may be a bit much, but make sure you're heard.}

---

This could be distilled into a single query with a large CASE statement:

update my_table
   set date_column = case when is_date(char_column, 'yyyy-mm-dd') = 1
                               then to_date(char_column, 'yyyy-mm-dd')
                          when is_date(char_column, 'yyyymmdd') = 1
                               then to_date(char_column, 'yyyymmdd')
                          ...
                          snip
                          ...
                     end

Answer 2

If I had a problem like this then my first question would be how to programatically detect what date format columnA is for each row. Assuming that it's doable within a reasoanble LOE (I don't know the complete spectrum of your date formats) I'd then see how I could use the CASE expression to detect the format and then format the date accordingly for each case.

Answer 3

If I understand your question I think you should be able to update your table in one statement. It's not the most elegant solution, but the following should work.

UPDATE tbl
   SET columnA = DECODE(type, '1', DECODE(INSTR(columnB, '-'), 5, TO_DATE(columnB, 'YYYY-MM-DD'),
                                                                  TO_DATE(columnB, 'MM/DD/YYYY')),
                                   TO_DATE(columnB, 'MMDDYYYY'));

In the above statement I'm assuming you have a column named that indicates which date format column B is in. The above statement uses the DECODE to determine if the type is 1. Since the type 1 has 3 possible formats , the statement will then try to determine which format columnB is in. To make our job easier we only need to test for the YYYY-MM-DD format as we can use the MM/DD/YYYY format with both 09/22/2011 and 9/22/2011. So we use the INSTR function to determine the location of the first '-' character. If that position is 5, then we know the column is in the format and can use the appropriate date mask. If the position is not 5, then we know columnB is in the format. Lastly, if the type isn't 1, then we know the date mask should be .