PROC SCAPROC Doesn't read Views created using PROC SQL - sas

Question

I'm building a tool to automate lineage between SAS jobs using SCAPROC and during my testing I found out that if a view is created using proc sql, SCAPROC doesn't capture the input for it and I end up with Jobs that doesn't have inputs.
Does anyone have any idea how to get around this?
Thanks

Answer 1

Views don't involve processing input data, so it makes sense SCAPROC would delay resolving it until it's actually used. While it's unfortunate that it's inconsistent with the data step views, it's not insurmountable.
With this code:
proc scaproc;
record 'h:\record.txt' attr;
run;
data test;
set sashelp.class;
run;
data testv/view=testv;
set sashelp.class;
run;
data result;
set testv;
run;
proc sql;
create view testv_sql as
select * from sashelp.class;
quit;
data result_sql;
set testv_sql;
run;
proc scaproc;
write;
run;
You see this output, around the data step where we use the view:
/* JOBSPLIT: TASKSTARTTIME 16JUL2021:09:46:59.00 */
/* JOBSPLIT: DATASET INPUT MULTI WORK.TESTV_SQL.VIEW */
/* JOBSPLIT: LIBNAME WORK V9 '...' */
/* JOBSPLIT: DATASET INPUT SEQ #C00001.CLASS.DATA */
/* JOBSPLIT: LIBNAME #C00001 V9 '...' */
/* JOBSPLIT: CONCATMEM #C00001 SASHELP */
/* JOBSPLIT: LIBNAME SASHELP V9 '( ... )' */
/* JOBSPLIT: DATASET INPUT SEQ SASHELP.CLASS.DATA */
/* JOBSPLIT: LIBNAME SASHELP V9 '( ... )' */
/* JOBSPLIT: DATASET OUTPUT SEQ WORK.RESULT_SQL.DATA */
/* JOBSPLIT: LIBNAME WORK V9 '...' */
/* JOBSPLIT: FILE OUTPUT h:\record.txt */
/* JOBSPLIT: ATTR WORK.TESTV_SQL.VIEW INPUT VARIABLE:Name TYPE:CHARACTER LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.TESTV_SQL.VIEW INPUT VARIABLE:Sex TYPE:CHARACTER LENGTH:1 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.TESTV_SQL.VIEW INPUT VARIABLE:Age TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.TESTV_SQL.VIEW INPUT VARIABLE:Height TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.TESTV_SQL.VIEW INPUT VARIABLE:Weight TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR #C00001.CLASS.DATA INPUT VARIABLE:Name TYPE:CHARACTER LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR #C00001.CLASS.DATA INPUT VARIABLE:Sex TYPE:CHARACTER LENGTH:1 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR #C00001.CLASS.DATA INPUT VARIABLE:Age TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR #C00001.CLASS.DATA INPUT VARIABLE:Height TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR #C00001.CLASS.DATA INPUT VARIABLE:Weight TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR SASHELP.CLASS.DATA INPUT VARIABLE:Name TYPE:CHARACTER LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR SASHELP.CLASS.DATA INPUT VARIABLE:Sex TYPE:CHARACTER LENGTH:1 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR SASHELP.CLASS.DATA INPUT VARIABLE:Age TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR SASHELP.CLASS.DATA INPUT VARIABLE:Height TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR SASHELP.CLASS.DATA INPUT VARIABLE:Weight TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.RESULT_SQL.DATA OUTPUT VARIABLE:Name TYPE:CHARACTER LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.RESULT_SQL.DATA OUTPUT VARIABLE:Sex TYPE:CHARACTER LENGTH:1 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.RESULT_SQL.DATA OUTPUT VARIABLE:Age TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.RESULT_SQL.DATA OUTPUT VARIABLE:Height TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: ATTR WORK.RESULT_SQL.DATA OUTPUT VARIABLE:Weight TYPE:NUMERIC LENGTH:8 LABEL: FORMAT: INFORMAT: */
/* JOBSPLIT: SYMBOL GET SYS_SQL_UNPUT_TRACE */
/* JOBSPLIT: SYMBOL GET SQLMAGIC */
/* JOBSPLIT: ELAPSED 19 */
/* JOBSPLIT: PROCNAME DATASTEP */
/* JOBSPLIT: STEP SOURCE FOLLOWS */
That gives you the input for that data step - the dataset passed through the view - which would be the proper input for that data step.
If this is a big problem for you, you could create a full pass-through view in the data step just before the PROC SQL view - then treat it in your lineage as the true input (and ignore the SQL created one).
data result_sql;
set testv_pass;
run;
That would then give you what you need (the input and the output).