Is proc fcmp compatible with regular expressions?

js5 — Tue, 06 Aug 2019 13:04:27 GMT

Dear community,

I am trying to adapt the date conversion function [1] to use regular expressions:

proc fcmp outlib=work.functions.conversions; /*[1]*/
	function datecdash2sdtm(indate $) $; /*[2]*/
		length outdate $10; /*[3]*/
		cdash_date_regex = prxparse("/(UN|\d{1,2}) (\w{3}) (\d{4})/");

/*		if indate ne ' ' then*/
		if prxmatch(cdash_date_regex, indate) then
			do;
				yyyy = prxposn(cdash_date_regex, 3, indate);
				mmm = prxposn(cdash_date_regex, 2, indate);
				dd = prxposn(cdash_date_regex, 1, indate);
				/*yyyy = substr(indate, 8, 4); *//*[4]*/
/*				mmm = upcase(substr(indate, 4, 3));*/
/*				dd = substr(indate, 1, 2);*/

				/* if year not missing */
				if notdigit(yyyy) = 0 then
					do;
						/*[5]*/
						mm = put(mmm, $month.); /*[6]*/

						/* if month not missing */
						if mm ne ' ' then
							do;
								/* if day not missing */
								if notdigit(dd) = 0 then
									do;
										/*[5]*/
										outdate = yyyy || '-' || strip(mm) || '-' || dd; /*[7]*/
									end; /* if notdigit(dd) = 0 */
								else outdate = yyyy || '-' || strip(mm);
							end; /* if mm ne ' ' */
						else outdate = yyyy;
					end; /* if notdigit(yyyy) = 0 */
				else outdate = ' ';
			end; /* if indate ne ' ' */
		else outdate = ' ';
		return(outdate); /*[8]*/
	endsub; /*[9]*/
run;

Unfortunately the function is not working and produces empty output. If I try to just parse the dates in plain data step

data mh_crf;
	set ecrf.mh1b ecrf.mh1a;
	MHSTDTC = datecdash2sdtm(MHTRSTDAT);
	MHENDTC = datecdash2sdtm(MHTRENDAT);
	cdash_date_regex = prxparse("/(UN|\d{1,2}) (\w{3}) (\d{4})/");

	if prxmatch(cdash_date_regex, MHTRENDAT) then
		do;
			yyyy = prxposn(cdash_date_regex, 3, MHTRENDAT);
			mmm = prxposn(cdash_date_regex, 2, MHTRENDAT);
			dd = prxposn(cdash_date_regex, 1, MHTRENDAT);
		end;
run;

then yyyy, mmm and dd variables are populated as expected. Are there any particular things one needs to keep in mind when mixing proc fcmp with regular expressions? Thank you!

[1] https://www.lexjansen.com/pharmasug/2017/TT/PharmaSUG-2017-TT07.pdf

Re: Is proc fcmp compatible with regular expressions?

js5 — Tue, 06 Aug 2019 13:44:07 GMT

It turns out that spaces were causing notdigit conditions to no longer be satisfied. Here is the working code:

proc fcmp outlib=work.functions.conversions; /*[1]*/
	function datecdash2sdtm(indate $) $; /*[2]*/
		length outdate $10; /*[3]*/
		cdash_date_regex = prxparse("/(UN|\d{1,2}) (\w{3}) (\d{4})/");

		if prxmatch(cdash_date_regex, indate) then
			do;
				yyyy = prxposn(cdash_date_regex, 3, indate); /*[4]*/
				mmm = prxposn(cdash_date_regex, 2, indate);
				dd = prxposn(cdash_date_regex, 1, indate);

				/* if year not missing */
				if notdigit(strip(yyyy)) = 0 then
					do;
						/*[5]*/
						mm = put(mmm, $month.); /*[6]*/

						/* if month not missing */
						if mm ne ' ' then
							do;
								/* if day not missing */
								if notdigit(strip(dd)) = 0 then
									do;
										dd = put(input(dd, 2.), z2.);
										/*[5]*/
										outdate = catx("-", yyyy, mm, dd); /*[7]*/
									end; /* if notdigit(dd) = 0 */
								else outdate = catx("-", yyyy, mm);
							end; /* if mm ne ' ' */
						else outdate = strip(yyyy);
					end; /* if notdigit(yyyy) = 0 */
				else outdate = ' ';
			end; /* if indate ne ' ' */
		else outdate = ' ';
		return(outdate); /*[8]*/
	endsub; /*[9]*/
run;

topic Is proc fcmp compatible with regular expressions? in SAS Procedures

Is proc fcmp compatible with regular expressions?

Re: Is proc fcmp compatible with regular expressions?