Question

I have a MS Access database. I am required to process and update tables in this database with SAS. The tables can be very large. Some of the processing requires the use of pure SAS functions. With a SAS/ACCESS LIBREF to MS Access, you cannot update or alter tables with PROC SQL. Therefore, I see two approaches:

1. Using LIBREF: Copy the entire table into SAS. Delete the original table in Access. Process the table in SAS. Write the table back to Access.
2. Using PROC SQL passthrough: Read the rows I need into SAS. Process these rows using SAS. Update the rows in Access.

My question: is number 2 even possible? With SQL functions (rather than SAS), the process outlined in number 2 is many orders of magnitude faster than number 1. I understand that this difference comes from the data being processed entirely by the DBMS, but number 2 would still have to be faster, right?

Answer 1

For future readers: I asked this question because SAS does not support a PROC SQL UPDATE statement on a MS Access library. To update an Access table, you must create a new table in the database and use a pass-through statement. My first instinct was to use an UPDATE statement assigning with a WHERE clause to update a subset of records. While I was researching this, I read some bad information that said Jet SQL did not support updating a subset of records. This is incorrect. Using an INNER JOIN, you are able to update a subset. Example code below:

/* Access database is "C:/database.mdb".
   Original table in database is "original". 
   Processed table is "work.updates". */

libname mdbfile "C:/database.mdb";

proc sql;
create table mdbfile.updates as select * from work.updates;
connect to access (path="C:/database.mdb");
execute (
  update original as a inner join updates as b on a.id = b.id
  set a.variable1 = b.variable1
  where condition = 1
) by access;
execute (drop table updates) by access;
disconnect from access;
quit;