Joined: 09 Mar 2011 Posts: 7309 Location: Inside the Matrix
Where does AFP come into this?
You do a "sideways" match.
Your "driver" is the first output.
Read driver until end of file.
With each record from driver:
Read record on subsidiary (the second output).
If records are equal, all is well, continue from Read driver...
If unequal, write to output file.
Continue from Read record on subsidiary...
When end of file on driver, write any remaining records on subsidiary to the output file.
At the end of this you should have your 10 extra records. You could include "record numbers" on the output if you wanted to know where they came from :-)
Note: this is a description of the process, not an indication as to how to structure your program :-)
The attempt to match with SuperCE might(!) imply that the files do not match very well.
You really need to tell us what you are trying to match. We can assume "documents" as you mentioned AFP. Should they be the same? Identical, or logical, for instance if there is a "time" anywhere in the document pages? Does AFP put in control information of some sort, specific to the job?
I'm suspecting you're going to have to rethink it. However, try to answer everything and we'll get a clearer picture of what you have.
If they are "documents" there seems little point in sorting the records.
If not documents you might test sorting seven times on chunks of 4000, with EQUALS. However, you seem to have variable-length records which would throw additional spanners.
Do you have the same, exactly, number of records on each? If not....
If so, you could extract the first 4000 bytes, including the RDW, sort on the whole thing and compare that using JOINKEYS.
I think you'll end up rethinking, no matter what the requirement says...
Sorry for all the confusion . Yes. The two files are documents in AFP formats. Actually this requirement came into picture when
A defective program accepts an afp file as input and inserts some page(a number of records) records in production . As I say its a defective one , we have rectified the program and now the program prouces an output that may more or less be the same as the output that was produced by the defective one but with the correct number of pages( a number of records ) inserted.
Now I would want to compare both these files and send a compare report to my senior officials . The output file has more than 10 million records ( i said as 1000 for an example) . So storing the records in an array may not be feasible. I know I have confused you people a lot . Sorry again
Joined: 09 Mar 2011 Posts: 7309 Location: Inside the Matrix
OK, if the output is supposed to be otherwise identical other than the "extras" then you can proceed as I suggested.
However, things like dates/times of production can mess you up. AFP control information...
Code it up. Test it with your test data (into bad program and good program) and see if you have any problems with information outside of your control varying between the runs.
If you have a problem with that, establish whether it can reasonably be "masked" in its native format.
If it can't be masked natively, how about taking the "spool" files and masking?
