About ChrisNZ

ChrisNZ · ‎05-27-2009

You can use Peter's test and the fact that sas can do character loops (in data steps at least, if not in macros): [pre] data sample; input cpt1 $ cpt2 $ cpt3 $ cpt4 $ cpt5$; do i='D9110','D9310','D9610'; ags + ^^( FIND( catx( '#', of cpt1-cpt5 ), i ) ); end; put ags=; cards; D9110 1100 122 2000 D9110 200 D9610 140 D9110 D9310 3000 D9110 D9230 300 D9110 D9610 3000 7777 7889 D9310 D9110 2333 3333 666 D9610 run; [/pre] This way you have all in one place: - a single list for all test values (in the loop) - a single list for all test variables (in the concatenation list) and you can massage the sums or derive ratios any way you want in the data step.

ChrisNZ · ‎05-27-2009

Thank you very much for your comprehensive and clear replies, Cynthia. I'll stick to writing HTML directly it seems as all the calculations do need to happen in one data step. No big deal, it is clean and it works. 🙂 A few rough things that could be better about the technical format of this forum: - One has to type the < code to type <. Not good! - A 'Code' button inserting [ pre ] tags would be nice - The 'Quote' button's action is crude compared to normal forum behaviour - Copying and pasting your code to sas (XP) pastes a single line. Going thru wordpad restores the line breaks, but loses the indentation. Just my 0.02 Thanks again.

ChrisNZ · ‎05-26-2009

Yes, you need formats, like: [pre] proc rank data=SASHELP.CITIDAY out=ALL1(keep=DFXWUK90 RSCORE) groups=10 ; var DFXWUK90; ranks RSCORE; proc means data=ALL1 noprint nway; var DFXWUK90; class RSCORE ; output out=ALL2 min=MIN max=MAX; data FORMAT; retain FMTNAME 'decile' TYPE 'N'; set ALL2(rename=(MIN=START MAX=END RSCORE=LABEL)); proc format cntlin=FORMAT; data _null_; set SASHELP.CITIDAY(where=( '01mar1990'd <= DATE <= '31mar1990'd)); V2=put(SNYSECM,decile2.); put V2=; run;[pre] You might want to enhance the FORMAT table to take care of values outside the deciles range ( variable HLO). Message was edited by: C. Graffeuille Message was edited by: C. Graffeuille

ChrisNZ · ‎05-26-2009

Unsure I fully understand you goal, you might need to provide more details. Linus' suggestion is good: in general, transposing the dataset (for example to have one record per figure), makes it easy to summarise it in funny ways. [pre] TEST# CAT# RESULT A x 3 B x 6 .... [/pre]

ChrisNZ · ‎05-26-2009

This is compact and easy to read. What out for 2-pass sql performance on large volumes. [pre] data t; input name $ a b; cards; Jim 3 4 Jim 4 5 Jim 6 7 Jim 6 8 Jim 3 2 Suzy 3 1 Suzy 2 1 run; [/pre][pre] proc sql; create table t1(drop=c) as select *, count(name) as c from t group by name having c >4 ; quit; [/pre]

ChrisNZ · ‎05-26-2009

Thanks a lot again, Cynthia, the dynamic color assignment logic is interesting, I have to look at it. Regarding the new table, I would be after something like: [pre] ods html file='twotable.html'; data step ; file print ods statement for table x; ...logic to populate table1 (probably involving loop, array, hash table) file print ods statement for table x+1; ...logic to populate table2 ods html close; [/pre] where 2 new tables are written out at each iteration, for example

ChrisNZ · ‎05-21-2009

If you used a sas session rather than a script file, you could use MPCONNECT to drive and control your parallel processes, and run your final routine upon completion of all 3 parallel processes. If you want to use a script, you probably have to launch the 4th process at the same time as the first ones, and have the 4th process wait and monitor the completion of the others before starting its own processing. If you are unsure how to detect completion, maybe create a dummy completion table/file/environment variable in the other processes that you can delete from the last one?

ChrisNZ · ‎05-21-2009

You might need to convert the blob to a string (if oracle can do this) as substr probably does not take blobs as argument.

ChrisNZ · ‎05-21-2009

Oh, one last tip please: I need to output several tables from my data step. How do I tell ODS that a new table starts? I tried a number of solutions, like providing a new object name in successive file print ods statements, but I didnt succeed.

ChrisNZ · ‎05-21-2009

oooh! Beautiful! Thanks Cynthia. Stage 2 solved, but stage 1 not quite: using a template before hand would require that the red values are known. The red values are determined in the data step, so I need a way to format on the fly. I do it by inserting <span style=></span> around the value at the moment, which is a good-enough workaround for html output. Any better way? Message was edited by: C. Graffeuille Message was edited by: C. Graffeuille

ChrisNZ · ‎05-21-2009

Maybe storing the blob value in a macro variable using select ... into would work, since your blobs are less than 64k. Then you can work on it using the macro language. But if sas supports larger blobs, then there might indeed be a better solution and tech support should be able to tell you. Also, a macro variable might not handle binary values perfectly, but it doesn't hurt to try... Another way would be to read the blob in chuncks smaller than 32k, using oracle's substring function. That would be much better than dumping into a text file! Message was edited by: C. Graffeuille

ChrisNZ · ‎05-21-2009

Bizarre error messages are best addressed by tech support. 🙂

ChrisNZ · ‎05-20-2009

This is a 2-stage question. I think stage 1 might be easier, but I could be wrong. 1- Is there a way to apply traffic-lighting (color) in a data step that writes to an (html) ODS destination (no, I cannot use a procedure)? eg I want Alice in red: ods html file='f:\ddd.htm' style=seaside; data _null_; set sashelp.class; file print ods=(variables=( _N_ age name)); put _ods_ ; run; ods html close; 2-Also, is there a way to reuse the stylesheet's style so the first column _N_ uses the header formatting (bold on yellow background here) ? Thank you for your help.

ChrisNZ · ‎05-20-2009

One entry per question please.

ChrisNZ · ‎05-20-2009

Do you mean "\" is a line continuation character ? The data is difficult to decipher. What would the records look like if they were stored one per line?

Online Status	Offline
Date Last Visited	Thursday