While PROC COMPARE technically supports ODS OUTPUT, the actual data output is basically just lines of text, which would need to be parsed in order to extract values from the text. Suggest to have PROC COMPARE produce useful ODS OUTPUT datasets with values stored in meaningful variables.
This code:
data class ;
set sashelp.class(drop=weight) ;
if name='Mary' then height=33 ;
run ;
ods trace on ;
ods output CompareDatasets=CompareDatasets CompareSummary=CompareSummary CompareDifferences=CompareDifferences ;
proc compare base=sashelp.class compare=class ;
run ;
ods output close ;
ods trace off ;
produces three output datasets, all of them just free text:
CompareDatasets
Obs type batch
1 h The COMPARE Procedure
2 h Comparison of SASHELP.CLASS with WORK.CLASS
3 h (Method=EXACT)
4 h
5 h Data Set Summary
6 h
7 h Dataset Created Modified NVar NObs Label
8 d
9 d SASHELP.CLASS 05AUG20:21:16:10 05AUG20:21:16:10 5 19 Student Data
10 d WORK.CLASS 05NOV22:18:58:12 05NOV22:18:58:12 4 19
11 d
12 d
13 h Variables Summary
14 h
15 d Number of Variables in Common: 4.
16 d Number of Variables in SASHELP.CLASS but not in WORK.CLASS: 1.
CompareSummary
Obs type batch
1 d
2 d
3 h Observation Summary
4 h
5 h Observation Base Compare
6 d
7 d First Obs 1 1
8 d First Unequal 14 14
9 d Last Unequal 14 14
10 d Last Obs 19 19
11 d
12 d Number of Observations in Common: 19.
13 d Total Number of Observations Read from SASHELP.CLASS: 19.
14 d Total Number of Observations Read from WORK.CLASS: 19.
15 d
16 d Number of Observations with Some Compared Variables Unequal: 1.
17 d Number of Observations with All Compared Variables Equal: 18.
18 d
19 d
20 h Values Comparison Summary
21 h
22 d Number of Variables Compared with All Observations Equal: 3.
23 d Number of Variables Compared with Some Observations Unequal: 1.
24 d Total Number of Values which Compare Unequal: 1.
25 d Maximum Difference: 33.5.
26 d
27 h The COMPARE Procedure
28 h Comparison of SASHELP.CLASS with WORK.CLASS
29 h (Method=EXACT)
30 h
31 h Variables with Unequal Values
32 h
33 h Variable Type Len Ndif MaxDif
34 d
35 d Height NUM 8 1 33.500
36 d
CompareDifferences
Obs type batch
1 d
2 d
3 h Value Comparison Results for Variables
4 h
5 d __________________________________________________________
6 d || Base Compare
7 h Obs || Height Height Diff. % Diff
8 d ________ || _________ _________ _________ _________
9 d ||
10 d 14 || 66.5000 33.0000 -33.5000 -50.3759
11 d __________________________________________________________
Consider the CompareDifferences table. This would be much for useful it had variables Obs, Variable, BaseValue, CompareValue, Difference.
The current output tables are basically as cumbersome as the pre-ODS approach to reading values from output, where you had to use PROC PRINTTO to send the results to a .lst file, and then input that .lst file as data and parse out the values you want.
PROC COMPARE is one of my favorite PROCS, but the lack of useful output datasets is a big weakness.
(Yes, I realize PROC COMPARE has an out= option, but I don't love that format either : )
