unapplied patches

[Jaroslav Hajek]

hello,

following David's suggestion, here is a list of patches I've sent that
are still unapplied (an perhaps need to be reworked). They're all
extensions, so nothing critical, but still, I'd like to see some of
them to get into Octave a lot.

==============================
save -struct extension

https://www.cae.wisc.edu/pipermail/octave-maintainers/2008-February/006127.html

I'd really like this to be in Octave, since I intend to use it a lot
:) Is there something wrong about this patch?
Perhaps I can remake it against recent sources, since it's quite old.


==============================
binary lookup:

https://www.cae.wisc.edu/pipermail/octave-maintainers/2008-March/006495.html

John didn't like the original code driven by gotos - admittedly, it
was not quite clear. This version is rewritten using STL's <algorithm>
an <functional> template libraries - it had become almost
self-documenting thus (at least if you're familiar with STL and its
metaprogramming) I'm not sure about Octave's attitude to STL though
...

I'd really like to see a true binary search in Octave (i.e. one with
logarithmic complexity per search, rather than linear-at-best as with
the current version), in this way or another. This version (based on
binary search, but optimized for downsampling) will outperform the
current version in most (if not all) cases, so I think it is a viable
approach.

The above message also points out a corner case when using lookup with
the expression `lookup(x(2:end-1))+1', that makes most of
interpolation functions buggy. This really needs to be fixed.

Since the `lookup(x(2:end-1))+1` usage is the most common, I was
thinking about making this kind of usage an option to lookup, rather
than using different expression, something like
lookup (table, y, 'lr'), where options mean: 'l' - treat the leftmost
value as negative infinity (i.e. always return at least 1)
treat the rightmost value as positive infinity (i.e. always return at
most length(table)-1).

If John agrees, I'll resubmit my lookup with these options supported
and then submit a fix for interp1, interp2 etc. to use this new style.


==============================
expm1 & log1p functions

https://www.cae.wisc.edu/pipermail/octave-maintainers/2008-March/006522.html


==============================
pchip interpolation method for interp2

https://www.cae.wisc.edu/pipermail/octave-maintainers/2008-March/006471.html

this was documented as missing in interp2. Similar to bicubic, but
uses SLATEC's dpchim to preserve monotonicity.
Also cleans up the code in interp2 and adds more demos.



==============================
dmult extension
https://www.cae.wisc.edu/pipermail/octave-maintainers/2008-March/006421.html


==============================
banker's rounding (roundb)
https://www.cae.wisc.edu/pipermail/octave-maintainers/2008-March/006420.html



-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz