[patch #6650] one pass covariance algorithm

[John Darrington]

Follow-up Comment #20, patch #6650 (project pspp):

> Dropping that code would eliminate a lot of tests. That is OK with me, but
it means that whoever uses covariance-matrix.c would have to remember to use
the right filter. If that's OK with everyone, I'll drop this code out of
covariance-matrix.c and add a comment saying that if the caller wants listwise
deletion of missing values, they'll have to use the right casefilter.

I think that's the right approach.  Even if listwise deletion is handled by
the covariance matrix implementation, then almost all procedures will still
need that filter for things like summary statistics.

> match_nodes still returns an int since it's a hash comparison function. It
used to return -1, 0 or 1, but now it just returns 0 or 1 because I'm not sure
how to make the sign of the return value after the change I made to the hash
table. The code in hash.c will work without the sign, but I thought I would
leave its return type as an int in case I thought of a good signed return
value later.

OK.  In that case, match_nodes could use a comment explaining this.


    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/patch/?6650>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/