EXTRACT substring from both left and right of specific character within variable

[Jumboshrimps]

Have a bunch of records as below:

INJECT 35-40 UNITS INTO THE SKIN DAILY 

INJECT 45-70 UNITS INTO THE SKIN DAILY.

10-12 UNITS AFTER BREAKFAST, LUNCH AND DINNER

VIA INSULIN PUMP AROUND 100-150 UNITS DAILY

Created two additional variables  to place strings to the immediate LEFT ("LOWERNG")

& RIGHT ( "UPPRNG" ) of where "-" is found in the string. 

I've identified where the "-" is in the variable with the index command. Can be any position in the variable, usually near the beginning of the string, but not always.  Using                                                                                                               

IF HYPHEN >0 THEN LOWERNG = substr(INSTRUCTIONS,HYPHEN,HYPHEN-5); - five characters to the LEFT
IF HYPHEN >0 THEN UPPRNG = substr(INSTRUCTIONS,HYPHEN,HYPHEN+5); - five characters to the RIGHT

intent is  to grab five characters to the immediate right ( "UPPRNG" ) of "-" and  five characters to the immediate left ("LOWERNG"). 

Or so I thought.

Here are some results: (HYPHEN is the position of the "-")

INSTRUCTION	HYPHEN	UPPRNG	LOWERNG
INJECT 0-8 UNITS INTO THE SKIN 3 TIMES DAILY.	9	-8 UNIT	-8 U
INJECT 10-20 UNITS INTO THE SKIN DAILY.	10	-20 UNI	-20 U
INJECT 40-55 UNITS INTO THE SKIN DAILY.	10	-55 UNI	-55 U

UPPERNG & LOWERNG are identical (more or less), both variables containing strings from the RIGHT of where the hyphen is. Not what I expected at all - below is what I thought I was going to receive:

INSTRUCTION	HYPHEN	LOWERNG	UPPRNG
INJECT 0-8 UNITS INTO THE SKIN 3 TIMES DAILY.	9	ECT 0	8 UN
INJECT 10-20 UNITS INTO THE SKIN DAILY.	10	ECT 10	20 UN
INJECT 40-55 UNITS INTO THE SKIN DAILY.	10	ECT 40	55 UN

I can scrub out any characters later with                                                                                             

COMPRESS(UPPRNG,'0123456789','k') ;                                                                                             

COMPRESS(LOWERNG,'0123456789','k') ;                                                                                                                

After all the characters are scrubbed out, am going to add LOWERNG & UPPRNG, divide by two, (will need to use a put command in there somewhere) so I get an average of the upper and lower ranges.  I would like to exclude the hyphen in either  LOWERNG or UPPRNG results.

But I'm stuck.

Thanx.

[SASKiwi · Accepted solution]

There are probably more elegant ways to do this but here is one approach. I've kept it fairly simple so you can see what each statement does:

data have;
  input @1 instructions $50.;
  pos = findw(instructions, 'UNITS');
  dose = substr(instructions, 1, pos - 2);
  words = countw(dose);
  lowerng = input(scan(dose, words -1), best12.);
  upperng = input(scan(dose, words), best12.);
  average = (lowerng + upperng)/2;
datalines;
INJECT 35-40 UNITS INTO THE SKIN DAILY 
INJECT 45-70 UNITS INTO THE SKIN DAILY.
10-12 UNITS AFTER BREAKFAST, LUNCH AND DINNER
VIA INSULIN PUMP AROUND 100-150 UNITS DAILY
;
run;

[Jumboshrimps]

That works.
Thanx.
Had a bad experience with scan function on this data and best10. - don't remember what the error was - but I should have persevered.

[mkeintz]

If "UNITS" is always preceded by an expression like  xxx-yyy, and if that is always the first instance of a dash in INSTRUCTIONS, then you could:

1. Get the first "word" of instructions, where the only word separator is a '-':  
     i.e.    scan(instructions,1,'-')      which extracts
          INJECT 35    from INJECT 35-40 UNITS INTO THE SKIN DAILY
   
   
2. From the results of #1, get the last word (where words are now separated by blanks).  Note the -1 below means to get the last word of the expression)
          scan(....,-1,' ')         which extracts   35 from   INJECT 35
   
   
3. The output of #2, when converted from text to numeric, produces the lower range
         lowerng=input(...,best12.)
   
   
4. Do the analogous operation to get the upper range.

data have;
  input instructions $50.;
datalines;
INJECT 35-40 UNITS INTO THE SKIN DAILY 
INJECT 45-70 UNITS INTO THE SKIN DAILY.
10-12 UNITS AFTER BREAKFAST, LUNCH AND DINNER
VIA INSULIN PUMP AROUND 100-150 UNITS DAILY
run;
data want;
  set have;
  lowerng= input(scan(scan(instructions,1,'-'),-1,' '),best12.);
  upperng= input(scan(scan(instructions,2,'-'),+1,' '),best12.);
run;

[ChrisNZ]

Function substr() does not work like what you made up.

Read the documentation.

The third parameter is the length. You always want 5 in your case.