Re: ROC is ready for review

[John Darrington]

Thanks for the review.  I addressed most of the issues you raised
and pushed a futher set of changes.

On Sat, Jul 18, 2009 at 09:39:38PM -0700, Ben Pfaff wrote:
     
     I don't really understand the algorithm used by
     casereader_create_distinct.  Why are there two layers
     (casereader_create_filter_func and casereader_create_translator)?

Ease of implementation I think.

     Why is it necessary to use casereader_peek (which forces all of
     the input data to stay around) instead of reading cases one at a
     time (and possibly being able to throw away data that has been
     consumed)?

I couldn't find any other solution which actually worked.
     
     process_group does a *lot* of manipulation of cases and
     caseprotos and readers and writers.  I didn't check it in detail
     because it wasn't obvious to me what the overall goal was.  It
     could really use a high-level comment.

I added a comment which explains that it actually does 3 things. In
order to keep down the number of data passes, it's necessary to have this
multi-purpose function.
     
J'

-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.

signature.asc
Description: Digital signature