Keeping Up With the FUN: New Functions in SAS 9

Deb Cassidy, Dublin, OH

ABSTRACT
This presentation will cover some of the
many new data step functions found in SAS
9. Some of these functions will add new
functionality. Other functions will be faster either in the processing time or less typing.
Some new functions are very similar to
functions found in other languages.
(Note: In the actual presentation, this was
not presented in the traditional lecture
format. It was done in a FUN format to fit
the title.)
INTRODUCTION
The first big question when is SAS 9
available? You can get 9.0 now by request.
SAS 9.1 will be available to select
customers in 3rd quarter 2002. Everyone
should have it by 1st quarter 2004. As of the
latest information at the time this paper was
written, it was still being decided if there
would be a mass shipment or some other
method in place to obtain SAS 9.1.
Are you wondering why you should care?
Version 6 HELP lists 19 categories with 236
functions. SAS 9 HELP lists 26 categories
with 452 functions quite a jump. If you are
used to looking up functions by category,
you will want to know that category names
have been modified and some functions
were reclassified.
Why are the new functions? Some offer new
functionality. Others have faster system
performance or a faster way for you to write
and review the code. Still other functions
were added to be similar to other languages.

This paper will cover some of the many

functions. The best place to find out about
all the functions is SAS 9 Help.
Note: is used to represent a blank in the
code samples when it may not be obvious.
PUTTING THINGS TOGETHER
You have a first name and last name and
now you want to create a single field in the
form Doe, Jane. The old way was
OLDWAY=LEFT(TRIM(LAST)) || ,
LEFT(TRIM(FIRST));

||

For some people, the hardest part was

figuring out how to specify the concatenate
symbol because there is not such a symbol
on the keyboard!
Here are four new concatenate functions.
But which one really gives you what you
need?
CAT1=CAT(LAST,, ,FIRST);
CAT2=CATT(LAST,, ,FIRST);
CAT3=CATS(LAST,, ,FIRST);
CAT4=CATX(, ,LAST,FIRST);
At first glance, you might think they all
would work. The first three statements say
to concatenate everything together.
However, CAT does not work in this case
because you have trailing blanks in your
original fields that are kept. CATT trims
trailing blanks. CATS is similar but strips
off both leading and trailing blanks. Too
bad they both also trim the blank after the
comma that you wanted to keep. CATX to
the rescue. In this function, the delimiter is
specified first. Leading and trailing blanks
are removed for the other fields but not from
the delimiter. You will need to remember

than the concatenate functions are like many

of the functions in previous versions - the
default length is $200 and should normally
be specified.
SELECTING SINGLE CHARACTERS
The old method was to use
OLDWAY=SUBSTR(FIELD,5,1);
Now you can use either of the following:
LETTER=CHAR(FIELD,5);
FRSTLTR=FIRST(FIELD);
Some people will say there is not much
difference. These are functions that were
probably added for efficiency or to be
consistent with other languages.
BLANKS
Do you have blanks in your fields that you
want to remove? The old way was
OLDWAY=LEFT(TRIM(FIELD));
While that was not complicated, the
following makes it even easier.

You get a message telling you were the

divison by zero occurred.
NOTE: Division by zero detected at line 79
column 7.
In most cases, it shows up way too many
times in the log.
The old way to get rid of these messages
was to use the following logic:
if B NE 0 then Y=A/B;
else Y=.;
The much simpler way is
Y=DIVIDE(A,B);
The bad news is that DIVIDE is not
available until SAS 9.1 so you may still be
waiting a few months for an easy way to get
rid of those nasty messages.
FINDING A STRING
Do you need to find a string within a field?
The old code was
OLD=INDEX(FIELD,HI);

Some people might want to put those blanks

back in as in the following example.

This returned the column of the first letter of

the first occurrence of the string. The new
function works in much the same way and
corresponds to functions in other languages.

ADDBLANKS=SUBPAD(FIELD);

NEW=FIND(FIELD,HI);

THE DREADED DIVISION BY 0

You code says

The FIND function also has some other

options. You can specify I to ignore case
or T to trim leading and trailing blanks.
Here are a few examples.

NEW=STRIP(FIELD);

Y=A/B;
How could something so simple cause
problems? Easy, if you did not now you had
records with a value of 0 for B.

A=abcedfghijhih;
B= ' xy ';
Statement
Old1=index(a,hi);
Old2=index(a,HI);
Old3=index(upcase(a),HI);
New1=find(a,'hi');
New2=find(a,'HI');
New3=find(a,'HI','i');
New4=find(b, ' xy ','t');

NEW=NOTALPHA(FIELD);
Returns
8
0
8
8
0
8
2

The other parameter you can specify for this

function is the starting point and the
direction. However, you need to make sure
you understand what happens when you go
in a negative direction. Here are a few
examples to show that negative simply
means find the first occurrence starting from
the right but still reading from the left. It
does not mean the field is checked as though
it had been reversed.
Statement
New5=find(a,'hi',11);
New6=find(a,'hi',-11);
New7=find(a,'hi',-13);
New8=find(a,'hi',-15);

Returns
11
8
11
11

SPOTTING TYPOS
Do you need to check your data to make
sure the values are legitimate. For example,
you need to check that only the letters A-Z
appear in a field.
OLDWAY=VERIFY(FIELD,
ABCDEFGHJKLMNOPQRSTUVWXYZ)
;
While it looks simple, you had to type every
letter from A-Z so it might be easy to miss
one! (Did you notice if a letter was missing
in the above statment?)
The new method is much easier:

This will return the position of the first

character that is not alpha. You can also
easily find the first character that is alpha by
using:
NEW=ANYALPHA(FIELD);
Are you more intersested in numerics than
alpha? How about punctuation characters?
The not/any functions are listed here along
with some examples.
ALNUM - alphanumeric
CNTRL - control character
DIGIT - digit
FIRST first character of a variable name
GRAPH graphical character
LOWER lower case
NAME character in a variable name
PRINT printable character
PUNCT punctuation character
SPACE white-space character
UPPER upper case letter
XDIGIT hex digit character
A= '$_ab12A: 5';
Statement
A1=anyalnum(a);
A2=anycntrl(a);
A3=anydigit(a);
A4=anyfirst(a);
A5=anygraph(a);
A6=anylower(a);
A7=anyname(a);
A8=anyprint(a);
A9=anypunct(a);
A10=anyspace(a);
A11=anyupper(a);
A12=anyxdigit(a);

Result
3
0
5
2
1
3
2
1
1
9
7
3

A= '$_ab12A: 5';
Statement
X1=notalnum(a);
X2=notcntrl(a);
X3=notdigit(a);
X4=notfirst(a);
X5=notgraph(a);
X6=notlower(a);
X7=notname(a);
X8=notprint(a);
X9=notpunct(a);
X10=notspace(a);
X11=notupper(a);
X12=notxdigit(a);

Result
1
1
1
1
9
1
1
0
3
1
1
1

HOW LONG IS IT?

Are you one of the many folks who just
didnt like the fact that
LEN=LENGTH();
returned 1? If there isnt anything there, how
could it possible have a length of one? If
you have
LEN=LENGTH(A);

A=;
B=LONGER;
C=LONG WITH BLANKS
Length D $200;
D=EXTRA LONG;

STATEMENT
A0=LENGTH(A);
A1=LENGTHN(A);
A2=LENGTHC(A);
A3=LENGTHM(A);

RETURNS
1
0
1
1

B="LONG";
B0=LENGTH(B);
B1=LENGTHN(B);
B2=LENGTHC(B);
B3=LENGTHM(B);

4
4
4
4

C0=LENGTH(C);
C1=LENGTHN(C);
C2=LENGTHC(C);
C3=LENGTHM(C);

16
16
24
24

D0=LENGTH(D);
D1=LENGTHN(D);
D2=LENGTHC(D);
D3=LENGTHM(D);

10
10
200
200

you also get a length of 1.

Well now, you have some choices.
LENGTH returns the length excluding
trailing blanks but returns a 1 if there isnt
anything in the string.
LENGTHN returns the length excluding
trailing blanks but unlike LENGTH, it will
return a 0 for blank character string.

SAME OR NOT?
Do you need to check if two fields are the
same? Then COMPARE is for you (do not
confuse it with the PROC which works on
datasets).
X=name;
Y=naem;

LENGTHC returns the length including

trailing blanks. It can never return 0.

COMPARE will return the location of the

first character that is different.

LENGTHM returns the amount of memory

that is allocated for the string.

Z=COMPARE(X,Y);

In the above example, it is pretty obvious

that the 3rd letter differs so COMPARE
returns 3.
However, if you switch the order of your
fields to

hear the collective moan when they were

told the right answer. That was the
motivation behind doing this paper in a
fun format and emphasizing the moral of
the story assuming how a function works
without looking into the details can be
hazardous to your results!

Z=COMPARE(Y,X);
then COMPARE returns -3. Why?
The sign of the result tells you which field
comes first based on your sort sequence. It
is negative if the first field listed would sort
before the second field listed. Otherwise, it
is positive. In this case, on a Windows
system, naem would sort before name
so the result was negative. Remember, sort
order can vary by operating system and you
can overrice the sort order with options.
You have some other options. For example,
you can ignore case in doing your
comparison, remove leading blanks or
quotation marks, and truncate fields to the
shorter length.
Here are a few more examples.
X
Y
Comparison
Name naem (x,y)
Name naem (y,x)
Name naem (x,y,i)
X
X
(x,y)
X
X
(x,y,l)
ab
AB (x,y)
ab
AB (x,y,n)
ABC AB
(x,y)
ABC AB
(x,y,:)

Result
-1
1
3
-1
0
2
0
3
0

SMALLEST & LARGEST

A variation of this presentation was given at
SUGI except the audience was only asked a
couple questions including one about the
new LARGEST function. Most people
guessed the wrong answer and you could

The audience was asked for the results of the

function
LARGE2=LARGEST(2,A,B,C,D);
when the following values were used.
A=1
B=3
C=5
D=7
Some in the SUGI audience replied 7.
Others realized the 2 indicated the second
item and replied 3. Others replied D or
B which is the variable name. The
answer is really 5. Why? Those who
thought the 2 indicated the second item
were right except it isnt the second item in
order but the second largest item. In this
case, 7 is the largest value and 5 is the
second largest.
The related SMALLEST function works in
the same way.
UP OR DOWN
There have always been various rules for
rounding and they vary by region as well as
industry. I was always taught that if the
value ends in 0-4, you round down.
Otherwise, you round up for 5-9. However,
other people were taught to that values
ending in 5 are rounded up or down
depending on whether the previous digit is
odd or even.

The traditional ROUND rounds the first

argument to the nearest multiple of the
second argument, or to the nearest integer
when the second argument is omitted. The
new ROUNDE returns and even multiple
when the first argument is halfway between
the two nearest multiples. ROUNDZ is like
ROUND but it does not use fuzzing.

This time you really do get different results

using the fuzz factor and not using it.
NO MORE PUT?
Do you ever need to convert a numeric
value to a character? The old way was:
OLDWAY=PUT(Y,10.2);

(123.45,.1)
(123.55,.1)
(3.156,.003)
(3.157,.003)

ROUND
123.5
123.6
3.156
3.156

ROUNDE
123.4
123.6
3.156
3.156

ROUNDZ
123.4
123.5
3.156
3.156

STILL A LITTLE FUZZY?

The fuzz factor says that if things are close,
SAS will consider them equal. There are
several procs which allow you to specify the
fuzz factor. Normally, the fuzz factor is 1E12. Many functions used the fuzz factor by
default. There are now functions which use
0 as a fuzz factor. These include FLOORZ,
CEILZ, INTZ and MODZ. However, there
is a caution in the HELP stating that you
may be unexpected results!
Here are a few examples for the value
3.55943
Regular
Z
CEIL
4
4
FLOOR
3
3
INT
3
3
MOD,2
1.55943
1.55943
Are you wondering why they are the same? I
kept trying to come up with an example and
then I finally read the HELP a little closer.
Lets say your value is in exponential
notation such as (1.-1.e-13)
CEIL
FLOOR
INT
MOD,2

Regular
1
1
1
1

Z
1
0
0
1

where Y is a numeric value and OLDWAY

becomes character with the length
dependent upon the length you specify in the
format. In this case, the length of the
character field is 10.
The new way uses the VVALUE function
FORMAT Y 10.2;
NEWWAY=VVALUE(Y);
Well, this takes more code to write so the
question is whether it is more efficient. Ill
leave that up to you to determine. I will tell
you there is a negative side to this method.
The default length for the new variable is
$200!
However, there is a related function which
could prove to be quite useful. Here is an
example:
DATA MYVAL;
FIELD1=17.5;
FIELD2=21.2;
FORMAT FIELD1 5.2 FIELD2 Z6.2;
MYFIELD=FIELD1;
MYVALUE=VVALUEX(MYFIELD);
When MYFIELD is set to FIELD1, the
result in MYVALUE is 17.50. When
MYFIELD is set to FIELD2, then the result
in 021.20. The value of MYVALUE will be
determined by MYFIELD. I see great use

when something in your code or data

specifies what MYFIELD should be for each
record.
If you arent familiar with the Z. format, it
keeps the leading zeroes. At first glance, I
expected 2 leading zeroes since I specified a
length of 6 and 21.20 only accounts for 4
digits. Then I remembered the decimal is
counted as part of the length. One other
point this is another function which
defaults to length $200 so dont forget to
specify a length statement.
MORE!
This is only the tip of the iceberg as far as
new functions. Some are more useful in
some industries that others. The examples I
picked are all related to papers I have
written in the past - and now need to update
to reflect SAS 9!

HOW TO LEARN MORE

The Online Documentation contains lots of
help on these new functions. Depending
upon your site, you may be able to access it
locally. If not or you just can not wait until
they install SAS 9 at your site, then check
out SUPPORT.SAS.COM. You need to
register but access is free.
Need some customized help in figuring out a
function and what it is doing? Then try SASL, an electronic user group that covers the
entire world! There are several ways to get
access to SAS-L. There are numerous SUGI
and SESUG papers which explain the
options. Check out Joe Kelleys paper at this
SESUG to learn how SAS-L functions
behind the scenes. Joe is not only an active
SESUG member but the man behind the
screen for the University of Georgia which
is one of the host sites for SAS-L. My
preferred method of reading SAS-L is to use

http://www.listserv.uga.edu/archives/sas-l.html

but some people prefer newsgroups or

getting e-mail.

CONCLUSION
There are lots of new functions in SAS 9.
Unless you want to be considered out-ofdate, youll want to review the functions and
see which ones can improve your work. As
I pointed out in some of the examples,
though, make sure you understand the
function, what it does and how to use the
parameters correctly or you could end up
with wrong results.

CONTACT INFORMATION
Deb Cassidy
[email protected]
614-757-7136