Re: Matching Records - End Product

Still not safe. The first when-clause needs to be conditioned by 'not %eof(file1) and not %eof(file2)'. If either file hits eof, the fields still contain the value of the last record read.

And then there's the sequence of the records. The code assumes the records are read in ascending order by key. If that is not the case .... The cycle checks it (runtime error: 'Match field out of sequence').

But the point is made: writing your own matching records logic is trivial. Sort of :-)

Joep Beckeringh


Murphy, Mark schreef:

These are both very nice, but they won't work unless the last record in both files is a matching
record. So you have to add some complexity to prevent an infinite loop in all but one circumstance.
/free
read file1;
read file2;
dow (not %eof(file1) or not %eof(file2));
select;
when old.key = new.key;
update file1; //matched records logic...
read file1;
read file2;
when (old.key < new.key and not %eof(file1))
or %eof(file2);
delete file1; // old file record no longer needed.
read file1;
other; // (old.key > new.key and not %eof(file2))
// or %eof(file1) = New record!
write file1; // using file2 fields
read file2;
endsl;
enddo;
*inlr = *on;
/end-free

You don't have to deal with that for MR logic. Now try making it work for three files. The
complexity increases exponentially. But not with MR logic. With MR logic you have a predefined
behavior that works the same whether you are matching records from two files or three or more. If you
are looking for a different behavior, then you will have to code it yourself. Do you really want to
code it if you don't have to?

BTW, this really won't give you the same progression of records as MR logic. If there are record
formats involved that do not include match fields this won't work, but even if all record formats have
match fields you could get something different. MR logic will select one record at a time for
processing, and the match fields do not have to be primary key fields. This means that each match can
be found on multiple records of both files. Reading both files each time there is a match can result
in a much different record progression than what you get with MR logic.