Agreed!  Non-integer array indices is a dangerous approach to programming. ... View more

Doesn't look like anything is wrong. Don't unerstand why you're showing for RESTORING K not being 627. What's posted looks like an excerpt from a much larger program, perhaps the culprit is elsewhere and got edited out. By the way, display format, like 10.5, have nothing to do with internal precision. 1 DATA TEST; 2 HMWID = 62.7; 3 LIN_FT = 1; 4 RUN; NOTE: The data set WORK.TEST has 1 observations and 2 variables. NOTE: DATA statement used (Total process time): real time 0.01 seconds cpu time 0.01 seconds 　 5 6 DATA AVAL (KEEP = K FT); 7 ARRAY LAR{240:799} _TEMPORARY_; 8 SET TEST END=END; 9 PUTLOG HMWID=10.5 +3 HMWID=HEX16.; 10 11 HMWID = HMWID * 10; 12 PUTLOG 'STORING VALUE ' HMWID=10.5 +3 HMWID=HEX16.; 13 LAR{HMWID} + LIN_FT; 14 15 IF END; 16 17 DO J = 240 TO (799-40); 18 K = J; 19 FT = LAR{J}; 20 IF FT THEN DO; 21 OUTPUT AVAL; 22 PUTLOG 'RESTORING VALUE ' K=10.5 +3 HMWID=HEX16.; 23 END; 24 END; 25 RUN; HMWID=62.70000 HMWID=404F59999999999A STORING VALUE HMWID=627.00000 HMWID=4083980000000000 RESTORING VALUE K=627.00000 HMWID=4083980000000000 NOTE: There were 1 observations read from the data set WORK.TEST. NOTE: The data set WORK.AVAL has 1 observations and 2 variables. NOTE: DATA statement used (Total process time): real time 0.03 seconds cpu time 0.01 seconds ... View more

Building off of MikeZdeb's solution, just wanted to see the performance trade-off between PROC TRANSPOSE and the good old DATA STEP.  This was run on an old Win XP machine with PC SAS 9.1.3 (stop laughing!) with 10,000 records and 800 variables... 21   options missing=''; 22   proc transpose data=x out=new (where=(missing(col1)) keep=_name_ col1); 23       by id; 24       var _all_; 25   run; NOTE: Numeric variables in the input data set will be converted to character in the output data set. NOTE: There were 10000 observations read from the data set WORK.X. NOTE: The data set WORK.NEW has 399281 observations and 2 variables. NOTE: PROCEDURE TRANSPOSE used (Total process time):       real time           2.42 seconds       cpu time            2.37 seconds 26 27   data new1(keep=__var); 28       set x; 29       array __num(*) _numeric_; 30       array __chr(*) _character_; 31 32       length __var $ 32; 33 34       do i=1 to hbound(__num,1); 35           if missing(__num(i)) then do; 36               __var = vname(__num(i)); 37               output; 38           end; 39       end; 40 41       do i=1 to hbound(__chr,1); 42           if missing(__chr(i)) then do; 43               __var = vname(__chr(i)); 44               output; 45           end; 46       end; 47   run; NOTE: There were 10000 observations read from the data set WORK.X. NOTE: The data set WORK.NEW1 has 399281 observations and 1 variables. NOTE: DATA statement used (Total process time):       real time           0.38 seconds       cpu time            0.39 seconds DATA step runs considerably faster, but took more time to code.  Is it clearer or more confusing than PROC TRANSPOSE is a matter of personal taste. ... View more

Just to add to data_null_'s answer, 1) s/<[^>]*>// can also be expressed as s/<.*?>//   these two expressions are identical. The "*?" is the lazy repetition operator that stops as soon as possible, whereas the normal "*" repetition operator is greedy and will match as much as possible.  Difference only show up when there are unbalanced delimiters, e.g., <aaa>bbb>   s/<.*>// would eliminate the whole string, s/<.*?>// would eliminate up to the first '>' only and leave bbb> untouched. 2) prxchange( rx, -1, string )    the "-1" would perform rx until the end of string, however many times that happens to be. ... View more

WIth SAS, there are always multiple ways to get things done, which is a good thing.  Calculating period from dates is more work, but is more resistant to changes to input data such as if the sorting order isn't preserved, or if records went missing in the middle.  Using counter is definitely fast and correct if no data issues crop up.  Pick whatever suits your data environment.  Having been bitten multiple times with sloppy input data streams, one picks up defensive programming habits quickly.  ... View more

Thanks, Art.  I didn't know about the sorting options, and it is very good to know.  Sadly, I'm still stuck in the old days runnning SAS 9.1.3 as IT is very slow in upgrading. ... View more

Hard to say.  In practice, poor performing models is usually not a methodology issue as long as you've been relatively careful in approach.  Problem is far more likely with the data, or lack of good data pertaining to the issue at hand. No amount of methodology research/experimentation/wizardry can overcome poorly gathered and/or irrelevant data.  Just my 2 cents worth. ... View more

Something like this should suite your needs. - The duration from Launch to Weekend is determined by intck() function, modify to fit your needs as intck() is very flexible with date handling. - Personally, I'd keep the period column instead of Weekno because sorting on on Prod-Period would give correct sequential results, but sorting on Prod-Weekno won't.  This is because you'd get   'Week1', 'Week10', 'Week11', before 'Week2' in the standard alphabet sorting sequence. data sales;     input Launch:yymmdd8. Weekend:date9. Prod Sales;     format Launch yymmdd10. weekend yymmdd10. sales dollar10. ;     Period = intck('week',launch,weekend);     Weekno = cats('Week',put(period,4.)); cards; 20101102 13Nov2010 100 300.00  20101102 20Nov2010 100 600.00  20101102 27Nov2010 100 1200.00  20101102 04Dec2010 100 1200.00  20101102 11Dec2010 100 1500.00  20101102 18Dec2010 100 1500.00  ; ... View more

Our outsourced UNIX server admins also removed cron from us, but gave us the 'at' command.  So we use at to re-schedule itself.  Here's an example from the 'at' command man pages: To have a job reschedule itself,  at  can  be  invoked  from within  the  at-job.  For  example,  this "daily-processing" script named my.daily runs every day (although crontab is a more appropriate vehicle for such work).  Content of the my.daily file:      # my.daily runs every day      at now tomorrow < my.daily      daily-processing So, if you have access to at,      change second line to 'at now + 10 minute < ...'      change daily-processing to 'sas -batch -noterminal .........' These operating system facilities is usually preferred than having one SAS long session that sleeps for 10 minutes then wakes up to do some stuff and the goes back to sleep. ... View more

Thanks, Wayne. ... View more

Here's a method to loop a macro over every value in a dataset.  The macro itself sequentially reads a SAS dataset, so it doesn't use a macro array, and since it's done in a macro, there's no call execute.  All the methods above should work, this is another one of many approaches possible. %macro loop( data );     %let dsid=%sysfunc(open(&data));     %if (&dsid = 0) %then %do;            %put %sysfunc(sysmsg());         %return;     %end;     %syscall set(dsid);          * links sas dataset variables to macro variables, watch out for name collision ;     %do %while( %sysfunc(fetch(&dsid)) = 0 );         %work(&msgid)               * replace with anything you need like a macro call with arguments ;     %end;     %let dsid=%sysfunc(close(&dsid)); %mend; %macro work( arg );      %put &arg; %mend; %loop( sashelp.adomsg ) This version lists the "msgid" column from sashelp.adomsg dataset. It illustrates the idea of using a macro to loop over SAS dataset and generate code.  And because it's all inside a macro, the looping and code generation is easier without needing to maintain arrays... etc. Modify to suite your needs ... View more