TIP: VARSTOCASES using MATRIX (KRONEKER and RESHAPE)

[David Marso]

/* Simulate 10000 rows of 1000 columns */.
 MATRIX.
 SAVE UNIFORM(10000,1000)/OUTFILE */VARIABLES x0001 TO x1000.
 END MATRIX.
 COMPUTE ID=$CASENUM.
 EXECUTE.

 /* Wish to restructure each original row into 20 rows of 50 columns */.
 /* This would require 50 MAKE subcommands in VARSTOCASES.
 /* (ie. Explicitely typing 1000 variable names if structured as
 /* PiQj and wanting P as rows with Q as columns.
 /* Alternatively one could use XSAVE within a LOOP with supporting vectors.
 /* Downside need to write to disk and indexing can be a little confusing .

/* Downside to my !V2C is that data have to fit in memory
(or call multiple times with partitions of the data).
/* vars have to be all numeric (or modify the SAVE to be string aware).
 DEFINE !V2C (
    vars !CHAREND ("/")
     / id !CHAREND ("/")
   / nrow !CHAREND ("/")
   / pindex !CHAREND ("/")
  / colnames !CHAREND ("/") )
 MATRIX.
 GET ID /FILE * /VARIABLES !id.
 GET qdata /FILE * /VARIABLES !vars.
 SAVE {KRONEKER(ID,MAKE(!nrow,1,1)),KRONEKER(MAKE(NROW(qdata),1,1),T({1:!nrow})),
   RESHAPE(qdata,NROW(Qdata)*!nrow,NCOL(Qdata)/!nrow)}
   /OUTFILE *
   /VARIABLES !id !pindex !colnames.
 END MATRIX.
 !ENDDEFINE.
 !V2C vars x0001 TO x1000/id id /nrow 20 /pindex index/colnames q01 TO q50.