What is the duration of a variable and how is it different to the length?
And why when you use the input function to transform a char variable with duration=1 to numeric, the duration of the new numeric variable is 8?
Ex. A string variable that contains numbers from 0 to 9 (only 1 digit), so that it has duration=1. When converted to numeric, the numeric variable contains numbers from 0 to 9, but duration=8
Not sure where you are seeing duration used in connection with variable definitions. Duration is a measure of time. Perhaps you meant width?
When you talk about a width for a variable you are talking about how many characters does it take to display the variable as a character string. When you specify a format or an informat you include the width you want to use after the format name and before the period. If you are reading a single digit number from a text file then you would use an informat with a width of 1. Or to write an integer between 0 and 9 you can use a format with a width of 1. But the width used in a format or an informat is independent of the length of the variable.
The length of a variable is the number of bytes that SAS will use to store the variable in a dataset. SAS only has two types of variables, floating point numbers and fixed length character strings.
For numbers SAS uses 64 bit floating point numbers so they take 8 bytes. So you cannot define a number with a length larger than 8. If you set the length for a numeric variable to less than 8 then SAS will store truncated values by discarding some of the bits from the mantissa so you lose some of the precision of the value.
For character variables the length is the number of bytes it will store. With single byte encodings (like WLATIN1) each character will take only one byte. But if you use UTF-8 encoding then each individual character could take between 1 and 4 bytes of storage.
For example the DATE9. format was a width of 9 and is used to print date values using 9 characters. But since dates are numbers the length needed to store the variable will be 8, not 9.
Or take your example of a character variable of length one that contains a single digit. You could convert it to a number using an informat like F1. that has a width of just one. But it will still take 8 bytes to represent the number as a floating point value. And SAS will force you use a length of at least 3 to store it into a dataset. (Note on IBM mainframes the minimum length for numeric variables is 2 instead of 3 because they use a different floating point representation.)
