How to parse characters between two different delimiters?

[xyxu]

I have a dataset with two variables. The first variable is a unique ID for each observation, and the second is a character variable. I am trying to use function SCAN to decompose each value of the second variable into multiple separate observations. Below is an example of the original value:

[2011-12-12--2014-04-04] Moudy El Khodr;[2016-10-01--] Laurent van Tuyckom;[2014-04-04--2016-10-01] Ignace de Coene;[2014-04-04--2016-01-01] Alexander Roose;[2016-10-01--] Alexander Roose;[2014-04-04--2015-03-01] Koen Bosquet;

I want to obtain all characters between each delimiter

[

and delimiter

;

as a piece of information, and save it into an observation with the same ID. Thus, for this case, I want to convert the original observation into 6 separate observations. Different IDs might have different numbers of pieces of information, and I am using function

COUNTW(variable2, '[')

to generate this number. 

I read through the guide for function SCAN, but now my problem is that, how can I parse information between two specific and different delimiters?

[rudfaden]

You might want to try using regex.

data have;
variable="[2011-12-12--2014-04-04] Moudy El Khodr;"; output;
variable="[2016-10-01--] Laurent van Tuyckom;"; output;
variable="[2014-04-04--2016-10-01] Ignace de Coene;"; output;
variable="[2014-04-04--2016-01-01] Alexander Roose;"; output;
variable="[2016-10-01--] Alexander Roose;"; output;
variable="[2014-04-04--2015-03-01] Koen Bosquet;"; output;
run;


data want;
set have;
	newString=prxchange('s/\[(.*);/$1/', -1, variable);	
run;
 

[xyxu]

The six pieces of information were stored in as one value of the second variable in original data. If they were already separate as in your dataset "have", I am done.

[rudfaden]

Ahh. Okay. I misunderstood your problem.

Try this: 

data have;
variable="[2011-12-12--2014-04-04] Moudy El Khodr;[2016-10-01--] Laurent van Tuyckom;[2014-04-04--2016-10-01] Ignace de Coene;[2014-04-04--2016-01-01] Alexander Roose;[2016-10-01--] Alexander Roose;[2014-04-04--2015-03-01] Koen Bosquet;";
run;

data want;
set have;
num=1;
start=1;
	do while (1);
		num=findc(variable,';',start); 
		if num = 0 then leave;
		variable2=substr(variable,start,num-start+1); output;
		start=num+1;
	end;
run;
 

[xyxu]

Thanks for your help. I am trying to use your code to process the dataset, which includes observations with missing values for the variable. Here is a sample and my modified version of your code:

data have;
length ID $3 variable $500;
ID = "001"; variable = "[2011-12-12--2014-04-04] David Tien;"; output;
ID = "002"; variable="[2011-12-12--2014-04-04] Moudy El Khodr;[2016-10-01--] Laurent van Tuyckom;[2014-04-04--2016-10-01] Ignace de Coene;[2014-04-04--2016-01-01] Alexander Roose;[2016-10-01--] Alexander Roose;[2014-04-04--2015-03-01] Koen Bosquet;"; output;
ID = "003"; variable = ""; output;
run;

data want;
set have;
	num_person = countw(variable, "[");
	if num_mgr = 0 then do;
		varible2 = ""; output;
	end;

	if num_person > 0 then do;
		num_find = 1;
		start = 1;
		i = 1;
		do while (i <= num_person);
			num_find =findc(variable, ';', start); 
			variable2=substr(variable, start, num_find - start + 1); output;
			start=num_find + 1;
			i+1;
		end;
	end;
run;

It seems to be doing what I need, but the log shows some error:

NOTE: Invalid third argument to function SUBSTR at line 1200 column 23.

ID=003 variable=  num_mgr=1 varible2=  num_find=0 start=1 i=2 variable2=  _ERROR_=1 _N_=3

Did I make any mistake?

[rudfaden]

The problem is that Variable is empty, but you are telling Substr to extract on character, so it throws an error. You should include a check for Variable being empty.

[RW9]

Well you can use perl regex, or datastep:

data want;
  length str s new_wrd $2000;
  str="[2011-12-12--2014-04-04] Moudy El Khodr;[2016-10-01--] Laurent van Tuyckom;[2014-04-04--2016-10-01] Ignace de Coene;[2014-04-04--2016-01-01] Alexander Roose;[2016-10-01--] Alexander Roose;[2014-04-04--2015-03-01] Koen Bosquet;";
  s=str;
  do while(lengthn(s));
    new_wrd=substr(s,1,findc(s,";"));
    s=substr(s,findc(s,";")+1);
    output;
  end;
run;

If you wanted you could replace the output with a further parse of the individual components.