CODEWITHSUNDEEP
arraysstringsasdata-manipulationdata-cleaning
J.Q
J.Q

Reputation: 1031

extracting names from messy character string

messy is a SAS character variable containing a list of papers an author cited in their own work.

Here is one observation of messy.

(label:1;name:Azad, Meghan B;pubyear:2008;volume:4;issue:2;pagenum:195;refwork:Autophagy;collkey:2008051953)(label:2;name:Bai, Jing;pubyear:2012;volume:39;issue:3;pagenum:2697;refwork:Mol Biol Rep;collkey:2012197491)

This record includes 2 references - one that begins at "(label:1;" and another that begins at "(label:2;".

I need to create character variables that return part of the content after "name:" for each reference. For this observation, that would look like this:

clean1            clean2
AZAD.MEGHAN       BAI.JING

I attempt to do this with the scan() function in a data step as follows:

data ex2;
length lastname1-lastname10 
       lastname1-lastname10 
       clean1-clean10 $100; /*initializes empty variables for storing up to 10 names*/
set ex;

array lastname  {*} lastname1-lastname10;
array firstname {*} firsttname1-firstname10;
array clean     {*} clean1-clean10;

do i=1 to count(messy, "label:"); /*loop through messy as many times as there are names*/
  lastname{i}  = scan(messy, 1, "name:"); /*pick up first word after name*/
  firstname{i} = scan(messy, 2, "name:"); /*pick up second word after name*/ 
  clean{i}     = cats(upcase(lastname{i}), ".", upcase(firstname{i})); 
end;

run;

There are (at least) two issues here:

  1. My loop will not pick up the nth name in the observation; it will continue to write the results of scan() (the content after the first occurrence of "name:") to the variables in the lastname and firstname arrays.
  2. I am clearly misunderstanding how scan() itself works! I thought the third argument allowed me to specify a delimiter of my choosing, but the results of scan(messy, 1, "name:"); returns "(l" instead of "AZAD" as I expected.

Specific Ask:

How can I pick up all names in the messy variable and store them separately as clean1, clean2, etc?

Upvotes: 0

Views: 79

Answers (1)

Lee
Lee

Reputation: 1427

I would use PRX* for such things

data ex2;
length clean1-clean10 $100;
set ex;
array clean     {*} clean1-clean10;
if _N_=1 then ExpressionID+prxparse('/[(;]name:([^;)]+)[;)]/');
start = 1;
stop = length(messy);
call prxnext(ExpressionID, start, stop, messy, position, length);
i=0;
do while (position > 0);
  i+1;
  clean{i} = prxposn(ExpressionID, 1, messy);
  call prxnext(ExpressionID, start, stop, messy, position, length);
end;

run;

Upvotes: 1

Related Questions