topic Re: A macro behaving as it is not supposed to in SAS Programming

A macro behaving as it is not supposed to

RDzh — Thu, 30 Jul 2015 18:54:31 GMT

Hello everyone!

As a practice exercise I've been writing a macro that does this: substitutes the n-th occurrence of a substring within a string with another string. Thus, if I have the string KingRobertKingGeorgeKingAdamKing and I will use the macro to substitute, say, the 2nd occurrence of King with, say, Queen, and haveKingRobertQueenGeorgeKingAdamKing.

Now I've written a fairly successful macro to do this, except that it shows some behaviour it is not supposed to; here is the code (but don't read it line by line; just focus on where the problem arises, as explained in the red text and after the code). I provide the code as well as the data-set I use it with if you want to run it for yourself and see the issue first-hand. So...

%macro auxm(ds, col, str1, str2, n_occurr);

%let ds_id = %sysfunc(open(&ds));

%let colVarnum = %sysfunc(varnum(&ds_id, &col));

%let colVartype = %sysfunc(vartype(&ds_id, &colVarnum));

/* Check */ %put Check: vartype for the column is &colVartype;

%if &colVartype ne C %then %put The column is not character;

%else %do;

%put The column is character;

%let lstr1 = %length(&str1);

/* Check */ %put Length of the string to be changed is &lstr1;

data &ds.2;

set &ds;

temp = &col;

if &n_occurr = 1 then do;

if index(temp, "&str1") = 1 then temp = "&str2"||substr(temp, &lstr1 + 1);

else temp = substr(temp, 1, index(temp, "&str1") - 1) || "&str2" || substr(temp, index(temp, "&str1") + &lstr1);

end;

else do; Problem arises here: the code in green below is not supposed to be hit when n_occurr = 1, but it does get hit! Why?

do i = 1 to (&n_occurr - 1);

if index(temp, "&str1") = 1 then temp = substr(temp, &lstr1 + 1);

else temp = substr(temp, 1, index(temp, "&str1") - 1) || substr(temp, index(temp, "&str1") + &lstr1);

end;

drop i;

pos = index(temp, "&str1") + (&n_occurr - 1)*(&lstr1);

col3 = &col;

col3 = substr(col3, 1, pos - 1)||"&str2"||substr(col3, pos + &lstr1);

drop age temp pos;

end;

run;

%end;

%mend auxm;

In more detail, my problem is this: since I work with two cases, based on whether n_occurr = 1 or n_occurr > 1 (i.e., whether I want to change the first or any subsequent occurrence of a given substring), when I have n_occurr = 1 for some reason my code also hits the else do condition.

I test this code with the dataset

data test;

input Name $25. age;

datalines;

RadoslavRadossRado 12

IvanRadowwRadollRados 5

SimonRadoseRadosseqRadok 31

;

run;

and with

%auxm(test, Name, Rado, XXX, 1), which works as supposed only when I put in comments the else-do part above;

%auxm(test, Name, Rado, XXX, 2), which works fine.

Any idea will be much appreciated!

Re: A macro behaving as it is not supposed to

ballardw — Thu, 30 Jul 2015 19:23:08 GMT

Please explicitly indicate which code is not supposed to run when n_occurr=1. Highlight all the lines.

Re: A macro behaving as it is not supposed to

RDzh — Thu, 30 Jul 2015 19:33:23 GMT

Thanks for your feedback and interest, ballardw!

See, issue is: I work with two cases, when n_occurr = 1 and when it is > 1 (that is, when I want to change the first occurrence of the substring and when I want to change some further occurrence). Consequently, I use if &n_occurr = 1 then do and then else do ; but somehow the statements in the else-do block also get hit when n_occurr = 1. When I put them in /* */ the macro works fine for n_occurr = 1.

Thanks and kind regards

Re: A macro behaving as it is not supposed to

ballardw — Thu, 30 Jul 2015 20:19:13 GMT

Change this:

drop i;

pos = index(temp, "&str1") + (&n_occurr - 1)*(&lstr1);

col3 = &col;

col3 = substr(col3, 1, pos - 1)||"&str2"||substr(col3, pos + &lstr1);

drop age temp pos;

end;

To:

end;

drop i;

pos = index(temp, "&str1") + (&n_occurr - 1)*(&lstr1);

col3 = &col;

col3 = substr(col3, 1, pos - 1)||"&str2"||substr(col3, pos + &lstr1);

drop age temp pos;

The reassignments were only in the "else do" loop. Which was not executing when n_occur =1.

Re: A macro behaving as it is not supposed to

data_null__ — Thu, 30 Jul 2015 21:47:57 GMT

This example is adapted from the SAS documentation example for POSNEXT

%let o=2; /*Which occurrence to replace.*/
%let t=rado; /*target*/
%let r=-Replacement-; /*String to replace it with*/
data test;
   retain ExpressionId;
   if _n_ eq 1 then ExpressionID = prxparse("/&t/i");
   input text :$25. age;
   retain r "&r"; 
   start = 1; 
   stop  = length(text);
   i = 0; 
   call prxnext(ExpressionID, start, stop, text, position, length);
      do while (position > 0);
         i + 1; 
         found = substr(text, position, length);
         *put found= position= length=;
         if &o eq i then leave;
         call prxnext(ExpressionID, start, stop, text, position, length);
      end; 
   length new $60; 
   if &o eq i then new = cats(substrn(text,1,position-1),r,substrn(text,position+length));
   put; 
   put text;
   put new;
   put; 
   datalines; 
RadoslavRadossRado   12
IvanRadowwRadollRados         5
SimonRadoseRadosseqRadok       31
SimonRados                     33
;;;;
   run; 
proc print; 
   run;

Re: A macro behaving as it is not supposed to

RDzh — Thu, 30 Jul 2015 23:01:22 GMT

No, it seems you got that wrong, ballardw, or I get your explanation wrong

My problem is precisely that the code inside the else-do does execute when n_occurr = 1, and I don't want it to, nor do I think it should. I'm interested also why it does; this seems to me like a bug or something about SAS I don't understand.

If you want, try my code in SAS to see what I mean.

Thanks and kind regards!

Re: A macro behaving as it is not supposed to

RDzh — Thu, 30 Jul 2015 23:03:02 GMT

Thanks for this, data _null_; it will be interesting to study this solution.

But I'm still keen to understand what goes wrong with my own.

Kind regards

Re: A macro behaving as it is not supposed to

Patrick — Thu, 30 Jul 2015 23:42:12 GMT

I can't replicate what you describe. Your code works as you want it to. Run below cut-down version of your macro and you'll see.

%macro auxm(ds, col, str1, str2, n_occurr);

%let ds_id = %sysfunc(open(&ds));

%let colVarnum = %sysfunc(varnum(&ds_id, &col));

%let colVartype = %sysfunc(vartype(&ds_id, &colVarnum));

/* Check */

%put Check: vartype for the column is &colVartype;

%if &colVartype ne C %then

%put The column is not character;

%else

%do;

%put The column is character;

%let lstr1 = %length(&str1);

/* Check */

%put Length of the string to be changed is &lstr1;

data _NULL_;

set &ds;

temp = &col;

if &n_occurr = 1 then

do;

put "XXXXXXX part 1";

end;

else

do;

put "XXXXXXX part 2";

end;

run;

%end;

%mend auxm;

%auxm(sashelp.class, Name, Rado, XXX, 1)

You could of course check the value of "&n_occur" on macro level and then only generate the SAS code block you actually want to execute. This wouldn't change the result though.

Re: A macro behaving as it is not supposed to

Tom — Thu, 30 Jul 2015 23:56:42 GMT

Sounds like you are confused between using macro logic (%IF ...%THEN ... ) to selectively generate code and data step logic (IF ... THEN ...) to selectively execute statements.

Consider this example.

data want ;

set sashelp.class;

if 0=1 then do;

drop NAME ;

end;

run;

Versus this example:

%macro example;

data want ;

set sashelp.class ;

%if 0=1 %then %do;

drop name ;

%end;

run;

%mend example;

%example;

The first data step will drop the variable NAME because a DROP statement is evaluated when the data step is compiled and has no executable component than can be controlled by logic flow.

Re: A macro behaving as it is not supposed to

RDzh — Fri, 31 Jul 2015 00:39:19 GMT

I started getting at something like this, because my code worked fine when I changed it this way:

%if &n_occurr = 1 %then %do;

data &ds.2;

set &ds;

temp = &col;

if index(temp, "&str1") = 1 then temp = "&str2"||substr(temp, &lstr1 + 1);

else temp = substr(temp, 1, index(temp, "&str1") - 1) || "&str2" || substr(temp, index(temp, "&str1") + &lstr1);

run;

%end;

%else %do;

data &ds.2;

set &ds;

temp = &col;

do i = 1 to (&n_occurr - 1);

if index(temp, "&str1") = 1 then temp = substr(temp, &lstr1 + 1);

else temp = substr(temp, 1, index(temp, "&str1") - 1) || substr(temp, index(temp, "&str1") + &lstr1);

end;

drop i;

pos = index(temp, "&str1") + (&n_occurr - 1)*(&lstr1);

col3 = &col;

col3 = substr(col3, 1, pos - 1)||"&str2"||substr(col3, pos + &lstr1);

drop age temp pos;

run;

%end;

But thanks very much for revealing this aspect of SAS to me!

Also, thanks to the rest of you for your participation!

Re: A macro behaving as it is not supposed to

RDzh — Fri, 31 Jul 2015 00:52:00 GMT

What is a good place to read more about this distinction, between what counts as code generation and what can/cannot be controlled by logic flow in a data step?

Thanks and kind regards