Re: [boost] [time_series] Review

From: Steven Watanabe (steven_at_[hidden])
Date: 2007-08-08 19:50:50

AMDG

Steven Watanabe <watanabesj <at> gmail.com> writes:

> I still need to think a bit more before I vote

Ok.  I vote to reject.

ordered_inserter is wrong.
Floating point offsets need to be fixed.

If these are fixed or if I can be convinced
that they are not problems then I would change my
vote to a yes. I have made suggestions for both of
these, but I am not sure that they work as I haven't
tried them yet.

A little more about floating point offsets:
I do not think that floating point offsets can
just be eliminated.  The piecewise constant forms
behave much like integral offsets.  As I mentioned
in another post sparse series is needed if the piecewise
constant series are available.  Also, I gave a whole
list of algorithms that don't work for sparse floating
point sparse series but on the other side of the coin partial_sum
makes sense for them but not for piecewise constant series.
If point series cannot be made to fit properly within the concepts
then the concepts need to be adjusted.

Finally, I am not knowledgeable at all about the problem domain.
Therefore, I am not qualified to express an opinion about
the usefulness of the library.

In Christ,
Steven Watanabe