The first one I got has 946201 rows and 16 columns. The second one I got has 946199 rows and 16 columns. I did use your code, and it showed 17 obs. I cannot figure out, because it is only 2 obs difference, but why this code generated 17 obs.
Appreciated if anyone can help!

proc compare就是用起来有点复杂

你说的差了2个obs是数量上的，完全有可能是第一个数据集没有完全包含第二个数据集，虽然在数量上比第二个数据集只多两个观测，但是其中有17个观测是第二个数据集里没有的。

我抽样查了这17个 好像在这两个数据集中都有

用这code 作了另一个比较产生error message

WARNING: A table has been extended with null columns to perform the EXCEPT set operation.
ERROR: Column 5 from the first contributor of EXCEPT is not the same type as its counterpart from
      the second.
ERROR: Column 6 from the first contributor of EXCEPT is not the same type as its counterpart from
      the second.
ERROR: Column 9 from the first contributor of EXCEPT is not the same type as its counterpart from
      the second.
ERROR: Column 10 from the first contributor of EXCEPT is not the same type as its counterpart
      from the second.
ERROR: Column 12 from the first contributor of EXCEPT is not the same type as its counterpart
      from the second.
ERROR: Column 13 from the first contributor of EXCEPT is not the same type as its counterpart
      from the second.
1830  quit;
NOTE: The SAS System stopped processing this step because of errors.

应该是两个数据集不同源，变量属性不同，试试compare吧
proc compare base=old comp=new;listall;run;
hope it help!

Thanks. List all, or listall, seem to be wrong. Delete this, I can still see it works. Any better suggestions?