*Copyright @ www.mycsg.in;

Pre-requisite

This lesson covers behaviour of the SET statement when appending datasets that have different variables or different variable attributes
For an introduction to appending using the SET statement with same-attribute datasets, refer to lesson SAS_APPENDING_L101

What happens when appending datasets with different variables?

When the input datasets on the SET statement have different variables, SAS includes all variables from all datasets in the output dataset
For observations from a dataset that does not contain a particular variable, SAS assigns a missing value for that variable
This behaviour is governed by the Program Data Vector (PDV) which is initialised with all variables found across all input datasets

Appending datasets with different variables

Create example datasets with different variables

We create two datasets from sashelp.class, each retaining a different subset of variables
class_age keeps NAME and AGE only, while class_sex keeps NAME and SEX only
Both datasets share the NAME variable but differ in their remaining variable — this is the scenario we want to demonstrate for appending datasets with different variables

data class_age;
   set sashelp.class(keep=name age);
run;
 
data class_sex;
   set sashelp.class(keep=name sex);
run;

class_age contains NAME and AGE only
class_sex contains NAME and SEX only
When appended, the output dataset will contain NAME, AGE, and SEX
Observations from class_age will have missing values for SEX
Observations from class_sex will have missing values for AGE

data class_combined;
   set class_age class_sex;
run;

Variable length differences when appending using SET

When the same variable exists in multiple input datasets but with different lengths, SAS uses the length defined in the PDV
The PDV takes the length from the first dataset in the SET statement that defines that variable
If a subsequent dataset has a longer length for the same variable, values may be truncated to the length established by the first dataset
SAS will issue a warning in the log when truncation occurs

Appending when the first dataset has a shorter variable length

data short_name;
   length name $5;
   set sashelp.class(obs=5 keep=name age);
run;
 
data long_name;
   set sashelp.class(keep=name age);
run;

short_name has NAME with length 5
long_name has NAME with default length 8
Since short_name is listed first on the SET statement, the PDV sets NAME length to 5
Observations from long_name will have NAME values truncated to 5 characters

data combined_truncated;
   set short_name long_name;
run;

Appending when the first dataset has a longer variable length

When the first dataset has the longer length, no truncation occurs because the PDV length is sufficient for all values

data combined_no_truncation;
   set long_name short_name;
run;

Controlling variable length using a LENGTH statement before SET

The recommended approach to avoid truncation is to explicitly define variable lengths using a LENGTH statement before the SET statement
The LENGTH statement establishes the variable in the PDV with the specified length before SAS reads any input dataset
This overrides the length that would otherwise be taken from the first dataset

data combined_controlled;
   length name $8;
   set short_name long_name;
run;

Variable type conflicts when appending using SET

When the same variable exists in multiple input datasets but with different types (numeric vs character), SAS produces an error and stops processing
Unlike length differences, type conflicts cannot be resolved with a LENGTH statement
The variable must be renamed or converted in one of the input datasets before appending

Resolving a type conflict using RENAME and conversion

Creating two datasets where AGE is numeric in one and character in the other

data age_numeric;
   set sashelp.class(obs=5 keep=name age);
run;
 
data age_character;
   set sashelp.class(obs=5 keep=name age);
   age_char=put(age,2.);
   drop age;
   rename age_char=age;
run;

To resolve the conflict, convert the character AGE back to numeric before appending

data age_character_fixed;
   set age_character;
   age_num=input(age,2.);
   drop age;
   rename age_num=age;
run;
 
data combined_type_resolved;
   set age_numeric age_character_fixed;
run;

*Copyright @ www.mycsg.in;

Pre-requisite

What happens when appending datasets with different variables?

Appending datasets with different variables

Create example datasets with different variables

SAS Log

SAS Log

Dataset View

Variable length differences when appending using SET

Appending when the first dataset has a shorter variable length

SAS Log

SAS Log

Dataset View

Appending when the first dataset has a longer variable length

SAS Log

Dataset View

Controlling variable length using a LENGTH statement before SET

SAS Log

Dataset View

Variable type conflicts when appending using SET

Resolving a type conflict using RENAME and conversion

SAS Log

SAS Log

Dataset View