Strip alpha to take average

Question

I have a table that looks like:
sub_dt date,
cd    varchar2(5),
day01 varchar2(5),
day02 varchar2(5),
day03 varchar2(5),
day31 varchar2(5))
sub_dt       cd       day01      day02      day03
01-JAN-06  A10    e5           e4            e5
04-FEB-06  A10    5             5              6
It contains daily measurement values for the given cd. Sometimes the data is entered as an estimate, so it is stored with a preceeding e. Why I don't know. But what I need to do is create two different types of averages:
1. The aveverage measurement by day - pretty easy avg(day01), avg(day02)...
2. The monthly average for the cd.
In either case I am stuck with these non-numeric values. What is the best approach for stripping out the alpha characters so that I can actually perform the math? I was thinking of using regular expresssion somehow, but I don't know if that is best. I was also thinking about creating a function to do a translate/replace of the alpha and return the number, but that is a whole lot of context switching - 31 times for each row.
Second is that the average over the month is a bit tricky. I think that I am stuck adding the data together and dividing by 31 - (day01 + day02 + day03)/31, but I think that creates potential for incorrect results. Image if someone only inputs data for three days out of each week, then really they only should be divided by 12 and not 31. So after I convert the data to number I would need to evaluate the nulls...analytics maybe?
I am guessing that I am making this more complicated than it really is and just need another pair of eyes to help me see forest through the trees.
Thanks, Tony

Answer 1

William -
The ltrim or substr would be a good method if I knew that the only letter was "e". I wasn't very clear about that part. Since it is a varchar2 field and this application has no data integrity this field can have just about anything in it. Someone could enter e or x or any other combination of letters prior to the actual number.
I like the collection idea, I will give that a shot.
<b>Updated:</b>
I used a combination of these two methods with really nice results.
select sub_dt,cd,avg(regexp_replace(column_value,'[^0-9]+','')) month_avg
from t, table(TableType(day01,day02,day03,....))
group by sub_dt,cd;
DT                  CD           MONTH_AVG
08/02/2006 08:49:40 0056                 4
08/03/2006 08:50:21 1001               145Thanks for the help, really great solutions.
Tony
Message was edited by: T