1. SAS Essentials
How SAS Thinks

2. “The DATA step is your most powerful programming tool
“The DATA step is your most powerful programming tool. So understand and use it well.”
Socrates

3. Objectives
understand DATA step:
processes
internals
defaults

4. processes
internals
defaults
compilation of DATA step source code
execution of resultant machine code

5. compile and execute phases of: INPUT (non SAS data) SET
processes
internals
defaults
compile and execute phases of:
INPUT (non SAS data)
SET

6. Compile Time Activities
processes
internals
defaults
Compile Time Activities
syntax scan
source code translation to machine language
definition of input and output files

7. Compile Time Activities
processes
internals
defaults
Compile Time Activities
input buffer
LPDV (logical program data vector)
data set descriptor information

8. Variables added in the order seen by the compiler
processes
internals
defaults
Creation of LPDV
Variables added in the order seen by the compiler
during parsing and interpretation of source statements

9. Compile Time Statements
processes
internals
defaults
Compile Time Statements
location critical BY
WHERE
ARRAY
ATTRIB
FORMAT
INFORMAT
LENGTH
location irrelevant
DROP
KEEP
LABEL
RENAME
RETAIN

10. Retained Variables processes internals defaults
all SAS special variables
_N_
_ERROR_
all vars in RETAIN statement
all vars from SET, MERGE, or UPDATE
accumulator vars in SUM statement(s)

11. Variables Not Retained
processes
internals
defaults
Variables Not Retained
Variables from input statement
user defined variables (other than SUM statement)

12. Type and Length of Variables
processes
internals
defaults
Type and Length of Variables
determined at compile time
by first reference to the compiler (in the DATA step)
Numerics:
length is 8 during DATA step processing
length is an output property

13. INPUT statement
reading non-SAS data

14. Compile Loop and LPDV data a ; put _all_ ; *write LPDV to LOG;
input idnum
diagdate: mmddyy8.
sex $
rx_grp $ 10. ;
time = intck (‘year’, diagdate, today() ) ;
put _all_; *write LPDV to LOG;
cards ;
F placebo
M 300 mg.
F 600 mg.
run;

15. logical program data vector
input buffer
logical program data vector
idnum diagdate sex rx_grp time
numeric numeric char char numeric
Building descriptor portion of SAS data set

16. logical program data vector
idnum diagdate sex rx_grp time _N_ _ERROR_
numeric numeric char char numeric
DKR* keep keep keep keep keep drop drop
*Drop/keep/rename

17. Execution of a DATA Step

18. Execution of a DATA Step
Initialization of LPDV
read input file Y
next step
end of file? N
process statements in step
termination
implied output

19. DATA Step Execution processes internals defaults
Implied read/write loop, stopped by:
no more data to read
explicit STOP
no input data
some execution time errors

20. Execution Time Activities
processes
internals
defaults
Execution Time Activities
execute initialize-to-missing (ITM)
read from input source
modify data using user-controlled statements
supply values of variables to LPDV
output observation to SAS data set

21. Initialization processes internals defaults _N_ set to loop count
_ERROR_ set to 0
user variables set to missing

22. Execution Loop - raw data
data a ;
put _all_ ; *write LPDV to LOG;
input idnum
diagdate: mmddyy8.
sex $
rx_grp $ 10. ;
time = intck (‘year’, diagdate, today() ) ;
put _all_; *write LPDV to LOG;
cards ;
F placebo
M 300 mg.
F 600 mg.
run;
proc contents; run;
proc print; run;

23. LPDV
IDNUM DIAGDATE SEX RX_GRP TIME _N_
F placebo
M 300 mg
F 600 mg
(over all executions of DATA step……..)

24. IDNUM=. DIAGDATE=. SEX= RX_GRP= TIME=. _ERROR_=0 _N_=1
2 data a ;
3 put _all_ ; *write LPDV to LOG;
4 input idnum
diagdate: mmddyy8.
sex $
rx_grp $ 10. ;
8 time = intck ('year', diagdate, today() ) ;
9 put _all_; *write LPDV to LOG;
10 cards ;
IDNUM=. DIAGDATE=. SEX= RX_GRP= TIME=. _ERROR_=0 _N_=1
IDNUM=1 DIAGDATE=-2670 SEX=F RX_GRP=placebo TIME=49 _ERROR_=0 _N_=1
IDNUM=. DIAGDATE=. SEX= RX_GRP= TIME=. _ERROR_=0 _N_=2
IDNUM=2 DIAGDATE=1780 SEX=M RX_GRP=300 mg. TIME=37 _ERROR_=0 _N_=2
IDNUM=. DIAGDATE=. SEX= RX_GRP= TIME=. _ERROR_=0 _N_=3
IDNUM=3 DIAGDATE=-4286 SEX=F RX_GRP=600 mg. TIME=53 _ERROR_=0 _N_=3
IDNUM=. DIAGDATE=. SEX= RX_GRP= TIME=. _ERROR_=0 _N_=4
NOTE: The data set WORK.A has 3 observations and 5 variables.
NOTE: The DATA statement used 0.59 seconds.
14 run; 15
16 proc contents; run;
NOTE: The PROCEDURE CONTENTS used 0.39 seconds.

25. -----Alphabetic List of Variables and Attributes-----
Data Set Name: WORK.A Observations:
Member Type: DATA Variables:
Engine: V Indexes:
Created: :18 Saturday, January 20, Observation Length: 42
Last Modified: 11:18 Saturday, January 20, Deleted Observations: 0
Protection: Compressed: NO
Data Set Type: Sorted: NO
Label:
-----Engine/Host Dependent Information-----
Data Set Page Size:
Number of Data Set Pages: 1
File Format:
First Data Page:
Max Obs per Page:
Obs in First Data Page: 3
-----Alphabetic List of Variables and Attributes-----
# Variable Type Len Pos
ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ
5 TIME Num
2 DIAGDATE Num
1 IDNUM Num
4 RX_GRP Char
3 SEX Char

26. PROC PRINT IDNUM DIAGDATE SEX RX_GRP TIME 1 -2670 F placebo 48
M mg
F mg

27. reading existing SAS data
SET statement
reading existing SAS data

28. DATA Step Compile no input buffer
compiler reads descriptor portion of input SAS data set to build the LPDV
returns same variables/attributes, including new variables

29. SET processes internals defaults
determine which SAS data set to be read
identify next observation to be read
copy variable values to LPDV

30. Execution Loop - SAS data
data sas_a ;
put _all_ ;
set a ;
tot_rec + 1 ;
run;

31. logical program data vector
Building LPDV from descriptor portion of old SAS data set
logical program data vector
idnum diagdate sex rx_grp time tot_rec
numeric numeric char char numeric numeric
Building descriptor portion of new SAS data set

32. LPDV 1 -2670 F placebo 48 1 1 1 -2670 F placebo 48 1 2
IDNUM DIAGDATE SEX RX_GRP TIME TOT_REC _N_
F placebo
F placebo
M mg
M mg
F mg
F mg
(over all executions of DATA step……..)

33. LOG idnum=. diagdate=. sex= rx_grp= time=. tot_rec=0 _ERROR_=0 _N_=1
idnum=1 diagdate=-2670 sex=F rx_grp=placebo time=48 tot_rec=1 _ERROR_=0 _N_=1
idnum=1 diagdate=-2670 sex=F rx_grp=placebo time=48 tot_rec=1 _ERROR_=0 _N_=2
idnum=2 diagdate=1780 sex=M rx_grp=300 mg. time=36 tot_rec=2 _ERROR_=0 _N_=2
idnum=2 diagdate=1780 sex=M rx_grp=300 mg. time=36 tot_rec=2 _ERROR_=0 _N_=3
idnum=3 diagdate=-4286 sex=F rx_grp=600 mg. time=52 tot_rec=3 _ERROR_=0 _N_=3
idnum=3 diagdate=-4286 sex=F rx_grp=600 mg. time=52 tot_rec=3 _ERROR_=0 _N_=4

34. PROC PRINT 1 -2670 F placebo 48 1 2 1780 M 300 mg. 36 2
IDNUM DIAGDATE SEX RX_GRP TIME TOT_REC
F placebo
M mg
F mg

35. Logic of a MERGE
compile
execute

36. data left;
input ID X Y ;
cards; ;
data right;
input ID A $ B $ ;
cards;
1 A14 B32
3 A53 B11 ;

37. merge left (in=inleft) right (in=inright); by ID ; run;
proc sort data=left; by ID; run;
proc sort data=right; by ID; run;
data both;
merge left (in=inleft)
right (in=inright);
by ID ;
run;

38. logical program data vector first iteration: MATCH
ID X Y A B INLEFT INRIGHT _N_ _ERROR_
A14 B

39. logical program data vector second iteration: NO MATCH
ID X Y A B INLEFT INRIGHT _N_ _ERROR_

40. logical program data vector third iteration: MATCH
ID X Y A B INLEFT INRIGHT _N_ _ERROR_
A53 B

41. Let’s try this again…………………
data left;
input ID X Y ;
cards; ;
data right;
input ID A $ B $ ;
cards;
1 A14 B32
3 A53 B11 ;

42. merge left (in=inleft) right (in=inright);
proc sort data=left; by ID; run;
proc sort data=right; by ID; run;
data both;
merge left (in=inleft)
right (in=inright);
***** by ID (one-on-one merge);
run;

43. logical program data vector first iteration: 1:1 “MATCH”
ID X Y A B _N_ _ERROR_
A14 B 1
OVERWRITTEN – value came from data set “right”

44. logical program data vector second iteration: 1:1 “MATCH”
ID X Y A B _N_ _ERROR_
A53 B 3
OVERWRITTEN – value came from data set “right”

45. logical program data vector third iteration: 1:1 “NO MATCH”
ID X Y A B _N_ _ERROR_
MISSING – no values from “right”

46. Output SAS data set
ID X Y A B
A14 B32
A53 B11

47. DATA Step Conclusions
Understanding internals and default activities allows you to:
make informed coding decisions
write flexible and efficient code
debug and test effectively
interpret results readily

48. Remember We have discussed DEFAULTS
As soon as you add options, statements, features, etc., the default actions change; TEST them!
You can use these same tools to track what’s happening.