By SPD, I assume he means that he is using the Scalable Performance Data Engine, or Server (SPDE/SPDS).  These are both SAS products.  SPDE being a storage engine avaialble with SAS/BASE and SPDS being a individual product offering.  In both cases proc contents will work for your needs.  Additionally proc contents also works on all of the database products I have worked with: Oracle, SQL/Server, Terradata, Netezza, so I assume it would work with any SAS/ACCESS works with, but couldn't be certain. Next point.  A table view does not really require space to store, so by limiting the variables will help save some cpu and other resources just like reading directly from the dataset it will not affect the disk space to store the view.  The SPDE in my experience requires slightly more space to store than the v9 engine.  Not a substantial difference though.  Typically I utilize SPDE libraries for extrememly large tables (exceeding several hundred GB).  These tables are usually accessed by many processes simultaneously by production operations and need the partitioned structure for better read access and I/O.  The structure of the reads in these situations becomes more random than sequential which being able to spread the dataset across multiple I/O channels becomes highly beneficial to performance. I do not know of any simple way to gather stats on all columns in a dataset as far as their content in concerned.  If you need to look through a dataset looking for variables with no population you will need to actually read through the entire file and analyize the contents yourself.  You could use any of many, many methods to accomplish this.  I would probably use proc freq with special 'catch all' formats and look in the output sets for varaibles where only my no value has count>0, but there are many ways to accomplish this manually. proc format value $charallfmt '' = 'no' other = 'yes'; value numallfmt .,0 = 'no' other = 'yes'; run; proc freq data=spdlib.spdset noprint; tables myvar1 /out=myvar1; tables myvar2 /out=myvar2; /* etc... */ format _character_ $charallfmt. _numeric_ numallfmt.; run; /* etc... */ ... View more

I have never experienced a behavior like you described with PROC FORMAT, do you have any links to provide me with some refernece to read on the issue? ... View more

proc format; invalue myfmt (default=2)   0  = 88   99 = .U   other = _same_; run; data want; input mynum :??myfmt. @@; /* nice way to suppress the unnecessary notes from my previous example, thanks Tom */ if mynum=. then mynum=.U; putlog mynum=; cards4; 0 1 2 Y T D ! . ( 99 0 ; ;;;; run; ... View more

Yes, Peter example of using the multiple execute statements provides the proper output for the datalines statement so it is another way of accomplishing the task.  The inclusion of the 'run' statment, while providing a properly closed block but should not effect the resulting output. ... View more

proc format; invalue numfmt (default=2)   0  = 88   99 = .U   .  = .U   other = _same_; run; data input; input num numfmt.; if _ERROR_=1 then do; num=.U; _ERROR_=0; end; datalines; 0 1 2 3 99 . - ! 3 T Y ; run; You will still get the note about invalid data, but not an error and the value will be equal to .U when an invalid value is encountered. ... View more