02-16-2012 01:04 AM
I have a sas7dbat file
|branch_code||Character||7||$CHAR7.||$CHAR7.||BRANCH CODE||All value are of three english character|
|PFS Cust Segment Code||Character||6||All value are of three english character|
|Verdate||Character||204||All value are of 6 numberic character as "201112"|
|New Cust Ind||Character||2||$2.00||$2.00||New Cust Ind||All value are of one english character|
|OccupationCMP||Character||13||The longest value for this field is 10 english characters.|
|cmp||Character||25||The longest value for this field is 10 english characters.|
|NoCount||Numeric||8||??it's length limit was 8, but I do find a value over 100M|
with 3.2M records and 837M disk space.
How to shrink it?
Thanks in advance.
02-16-2012 08:08 AM
This sounds like a quiz question. Since unnecessarily used space still takes up space, you are wasting a lot of space in all of the character fields. As for the numeric one, read the documentation: http://support.sas.com/documentation/cdl/en/hostunx/61879/HTML/default/viewer.htm#a000344718.htm
02-16-2012 09:42 AM
First, you have to be very sure that your "Comments" are correct. For example, you have to be 100% certain that VERDATE is never longer than six characters. Since BRANCH_CODE uses a $CHAR7. format/informat, it is not enough to print some values. The $CHAR formats imply that the variable might contain leading blanks. You have to determine if it does contain leading blanks, whether you would like to left-hand-justify the characters. The length needed might be impacted, but the sorted order might change as well.
After you have completed your analysis, set the option and the lengths appropriately. For example:
length VERDATE $ 6;
Set the length BEFORE the SET statement, for as many variables as you have analyzed to verify the needed length.
02-16-2012 10:36 AM
You will probably get pretty good reduction by just using the COMPRESS=YES option.
If you are correct about the maximum lengths of the actual data then you can make the file smaller by re-creating it with shorter character variables.
Where you would want to avoid this is if this is just one example of a number of datasets that you have with this structure. The other potential datasets might have longer actual values.
There is usually not much value (and some risk) in storing numbers in less than 8 bytes. SAS always stores numbers are floating point.
The easiest way to recreate with shorter lengths is to set the length before referencing the original data (as suggested above).
Watch out for the variables that have permanent formats attached. You should change those (or better just remove them). You will need to put the FORMAT statement after the set statement for it to win out.
data want ;
length branch_code $3 .... ;
set old ;
format _character_ ;