Thread Subject: Wishlist for R2007b

This thread will be used to indicate the community's requests for
features/bug-fixes for the upcoming Matlab release (R2007b, Sep 1
2007). This is in lieu of an official votable wishlist (see separate
thread on this: Yair Altman, "Proposed Matlab votable wishlist" #, 16 May 2007 11:13 am </WebX?14@@.ef5718f>
# ).

Here are some items I would like to see:
- fix the "clear java" bug (also clears globals as side-effect)
- try-catch should also catch Java exceptions, not just Matlab ones
- add an Excel-like iff function to the basic ML language (equivalent
to C's a?b:c construct) - highly important in anonymous functions
- enable multi-thread use of the Matlab computation engine from
external C++/C#/Java threads
- finally (after who knows how many years) fix the stupid zoom and
similar problems with plotyy

More to follow later - I think this is enough for now to start a
healthy discussion. Please pitch in with your wishlist items.

Yair Altman

In article <ef57195.-1@webcrossing.raydaftYaTP>, Yair Altman wrote:
> - finally (after who knows how many years) fix the stupid zoom and
> similar problems with plotyy

That's got my vote!

On Wed, 16 May 2007 11:24:33 -0400, Yair Altman wrote:

> This thread will be used to indicate the community's requests for
> features/bug-fixes for the upcoming Matlab release (R2007b, Sep 1 2007).
> This is in lieu of an official votable wishlist (see separate thread on
> this: Yair Altman, "Proposed Matlab votable wishlist" #, 16 May 2007
> 11:13 am </WebX?14@@.ef5718f> # ).
>
> Here are some items I would like to see: - fix the "clear java" bug
> (also clears globals as side-effect) - try-catch should also catch Java
> exceptions, not just Matlab ones - add an Excel-like iff function to the
> basic ML language (equivalent to C's a?b:c construct) - highly important
> in anonymous functions - enable multi-thread use of the Matlab
> computation engine from external C++/C#/Java threads
> - finally (after who knows how many years) fix the stupid zoom and
> similar problems with plotyy

1. Allow toggle to turn off file overwrite message with uiputfile
2. Officially support uitable.
3. Support handle objects again.
4. Make boolean functions such as intersect, union, etc. builtin so they
operate faster.
5. Fix listbox highlighted items turning gray when focus shifts to a
different uicontrol.

Dan

Kelvin Hales wrote:
>
>
> In article <ef57195.-1@webcrossing.raydaftYaTP>, Yair Altman
wrote:
>> - finally (after who knows how many years) fix the stupid zoom
> and
>> similar problems with plotyy
>
> That's got my vote!
>
>
>

Agree with the above and please include the "Serial port fprintf
sporatic time out" that has been around for the last 6 releases.

Reed

In article <ef57195.-1@webcrossing.raydaftYaTP>,
altmanyDEL@gmailDEL.comDEL says...
> This thread will be used to indicate the community's requests for
> features/bug-fixes for the upcoming Matlab release (R2007b, Sep 1
> 2007). This is in lieu of an official votable wishlist (see separate
> thread on this: Yair Altman, "Proposed Matlab votable wishlist" #, 16 May 2007 11:13 am </WebX?14@@.ef5718f>
> # ).
>
> Here are some items I would like to see:
> - fix the "clear java" bug (also clears globals as side-effect)
> - try-catch should also catch Java exceptions, not just Matlab ones
> - add an Excel-like iff function to the basic ML language (equivalent
> to C's a?b:c construct) - highly important in anonymous functions
> - enable multi-thread use of the Matlab computation engine from
> external C++/C#/Java threads
> - finally (after who knows how many years) fix the stupid zoom and
> similar problems with plotyy
>
> More to follow later - I think this is enough for now to start a
> healthy discussion. Please pitch in with your wishlist items.
>
> Yair Altman
>

I hate to hassle all of you who are enthusiastic and have great
suggestions. But it would be far more helpful if you would each put in
bug/enhancement requests here, one per bug/feature:

http://www.mathworks.com/support/service_request/login_servicerequests.h
tml?targetURL=/support/service_requests/contact_support.do

We do keep track of the frequency of reported issues and that can figure
into decisions on what to work on.

-- Loren
http://blogs.mathworks.com/loren/

> I hate to hassle all of you who are enthusiastic and have great
> suggestions. But it would be far more helpful if you would each put in
> bug/enhancement requests here, one per bug/feature:

I've been requesting at least one of mine since 1999, and it still hasn't
been addressed.

And I have requested the other ones each release since about Matlab 7.0.
All I get back from tech support is "that is not supported".

Dan

Dan Hensley wrote:
>
>
>> I hate to hassle all of you who are enthusiastic and have great
>> suggestions. But it would be far more helpful if you would
each
> put in
>> bug/enhancement requests here, one per bug/feature:
>
> I've been requesting at least one of mine since 1999, and it still
> hasn't
> been addressed.
>
> And I have requested the other ones each release since about Matlab
> 7.0.
> All I get back from tech support is "that is not supported".
>
> Dan
>
  

The one I refered to "Serial port fprintf sporadically times out";
report ID 250986 has been shown as open since release 7.0.1 Should I
report it again?
Reed

Just out of curiosity, who voted for the "nag me about my code"
colored tags that started to appear in the editor? I would happily
vote that out again ;)

That effort could have been put into bug fixing instead, e.g. sort out
Copy Figure, update all the legacy 'double only' functions and fully
optimize the math operations: mixed real-complex multiplication,
sparse complex. It amazes me The Mathworks needs us to vote to work on
basic stuff like that. Bah humbug!

In article <1179378005.215820.316490@u30g2000hsc.googlegroups.com>,
spasmous@gmail.com says...
> Just out of curiosity, who voted for the "nag me about my code"
> colored tags that started to appear in the editor? I would happily
> vote that out again ;)
>
> That effort could have been put into bug fixing instead, e.g. sort out
> Copy Figure, update all the legacy 'double only' functions and fully
> optimize the math operations: mixed real-complex multiplication,
> sparse complex. It amazes me The Mathworks needs us to vote to work on
> basic stuff like that. Bah humbug!
>
>

several things come to mind, most/all of which you probably know
already:

1) you can turn off mlint in the editor by unchecking a box for the
mlint preferences
2) developers are not entirely interchangeable. those that work on the
math typically have different skills and knowledge base than those who
work on graphics, and they again tend to differ from those who work
directly on the user interface

just some perspective. I believe we need the full breadth of knowledge
and expertise of all these development groups and MATLAB would be the
worse for wear if we concentrated in only one domain.

-- Loren
http://blogs.mathworks.com/loren/

spasmous wrote:
...
> update all the legacy 'double only' functions and fully
> optimize the math operations: mixed real-complex multiplication,
> sparse complex...

MATLAB relies on the BLAS for its performance for dense matrix
operations (and it uses the BLAS for the x=A\b in the sparse case as
well). The BLAS are far faster than what mere mortals writing in
C/C++ can get, so using them is an imperative (i.e, the Intel MKL on
the Intel cpu's, the AMD has their own etc). Each BLAS is optimized
for that particular architecture.

However, the BLAS doesn't support mixed real/complex. It's either
all real or all complex. To do C=A*B when A is real and B is
complex, my guess is that MATLAB does (and this is purely a guess):

C = A*real(B) + 1i*A*imag(B)

since MATLAB stores its matrices with the real and imaginary parts in
different arrays (thus the "+" above doesn't really occur, neither
does the "1i*"). Then this would become two calls to DGEMM.

If The MathWorks does that, my guess is that that's as best it will
get. If they instead do C=A*B by coercing A into complex (appending
a complex part) then it would be slower than the above, I would
predict.

The BLAS stores complex matrices with real/imaginary values
interleaved. So in memory, the BLAS sees real(A(1,1)), imag(A(1,1)),
real(A(2,1)), imag(A(2,1)), etc. But MATLAB stores the real part of
A and the imaginary part of A in 2 different real arrays. So there
will often be a copy made when passing complex matrices to the BLAS,
and I doubt that will ever change.

I could say lots about how sparse complex should be done, but that's
another story and quite a long one at that ...

And, if I were The MathWorks (and I'm not so I know nothing), it's
unlikely I'd be adding new features to a release just a few months
away. So this thread should be entitled "wish list for R2008x"
instead.

On May 17, 10:35 am, "Tim Davis" <d...@cise.ufl.edu> wrote:
> spasmouswrote:
>
> ...
>
> > update all the legacy 'double only' functions and fully
> > optimize the math operations: mixed real-complex multiplication,
> > sparse complex...
>
> MATLAB relies on the BLAS for its performance for dense matrix
> operations (and it uses the BLAS for the x=A\b in the sparse case as
> well). The BLAS are far faster than what mere mortals writing in
> C/C++ can get, so using them is an imperative (i.e, the Intel MKL on
> the Intel cpu's, the AMD has their own etc). Each BLAS is optimized
> for that particular architecture.
>
> However, the BLAS doesn't support mixed real/complex. It's either
> all real or all complex. To do C=A*B when A is real and B is
> complex, my guess is that MATLAB does (and this is purely a guess):
>
> C = A*real(B) + 1i*A*imag(B)
>
> since MATLAB stores its matrices with the real and imaginary parts in
> different arrays (thus the "+" above doesn't really occur, neither
> does the "1i*"). Then this would become two calls to DGEMM.
>
> If The MathWorks does that, my guess is that that's as best it will
> get. If they instead do C=A*B by coercing A into complex (appending
> a complex part) then it would be slower than the above, I would
> predict.
>
> The BLAS stores complex matrices with real/imaginary values
> interleaved. So in memory, the BLAS sees real(A(1,1)), imag(A(1,1)),
> real(A(2,1)), imag(A(2,1)), etc. But MATLAB stores the real part of
> A and the imaginary part of A in 2 different real arrays. So there
> will often be a copy made when passing complex matrices to the BLAS,
> and I doubt that will ever change.
>
> I could say lots about how sparse complex should be done, but that's
> another story and quite a long one at that ...
>
> And, if I were The MathWorks (and I'm not so I know nothing), it's
> unlikely I'd be adding new features to a release just a few months
> away. So this thread should be entitled "wish list for R2008x"
> instead.

Hi Tim, thank you for your input. Just for the record I'm basing my
own comments on comparisons of A * B and A .* B (different
operations). Whereas A * B is BLAS and The Mathworks is probably wise
to stay away, A .* B is definitely within their scope. When one is
complex and the other is real, the operation is more expensive than
pure complex-complex. I'm not really complaining as much as
explaining ;) Bug reports were submitted a year or two ago...

Well, since Loren has already said our pleas are falling on deaf
ears:
a girlfriend. It would be a bonus if she knew MatLab but I'll take
whatever you send.

 Tim Davis wrote:
>
>
> spasmous wrote:
> ...
>> update all the legacy 'double only' functions and fully
>> optimize the math operations: mixed real-complex
multiplication,
>> sparse complex...
>
> MATLAB relies on the BLAS for its performance for dense matrix
> operations (and it uses the BLAS for the x=A\b in the sparse case
> as
> well). The BLAS are far faster than what mere mortals writing in
> C/C++ can get, so using them is an imperative (i.e, the Intel MKL
> on
> the Intel cpu's, the AMD has their own etc). Each BLAS is
> optimized
> for that particular architecture.
>
> However, the BLAS doesn't support mixed real/complex. It's either
> all real or all complex. To do C=A*B when A is real and B is
> complex, my guess is that MATLAB does (and this is purely a guess):
>
> C = A*real(B) + 1i*A*imag(B)
>
> since MATLAB stores its matrices with the real and imaginary parts
> in
> different arrays (thus the "+" above doesn't really occur, neither
> does the "1i*"). Then this would become two calls to DGEMM.
>
> If The MathWorks does that, my guess is that that's as best it will
> get. If they instead do C=A*B by coercing A into complex
> (appending
> a complex part) then it would be slower than the above, I would
> predict.
>
> The BLAS stores complex matrices with real/imaginary values
> interleaved. So in memory, the BLAS sees real(A(1,1)),
> imag(A(1,1)),
> real(A(2,1)), imag(A(2,1)), etc. But MATLAB stores the real part
> of
> A and the imaginary part of A in 2 different real arrays. So there
> will often be a copy made when passing complex matrices to the
> BLAS,
> and I doubt that will ever change.
>
> I could say lots about how sparse complex should be done, but
> that's
> another story and quite a long one at that ...
>
> And, if I were The MathWorks (and I'm not so I know nothing), it's
> unlikely I'd be adding new features to a release just a few months
> away. So this thread should be entitled "wish list for R2008x"
> instead.

1. Sparse matrices of dimension greater than 2.

2. Support for running Matlab in emacs under Windows. This is
possible using workarounds, but after much searching I have not found
a clean implementation.

spasmous wrote:
>...
> Hi Tim, thank you for your input. Just for the record I'm basing my
> own comments on comparisons of A * B and A .* B (different
> operations). Whereas A * B is BLAS and The Mathworks is probably
> wise
> to stay away, A .* B is definitely within their scope. When one is
> complex and the other is real, the operation is more expensive than
> pure complex-complex. I'm not really complaining as much as
> explaining ;) Bug reports were submitted a year or two ago...

Oh, I see. A dot-times B (A.*B) doesn't exist since it's not linear
algebra ;-) so I know nothing about how it should work. You're
right, A.*B is BLAS-less.

In MATLAB 7.4, A.*B when A and B are both real take 0.4 seconds on
my laptop. When A is real and B is complex, the time goes up to 0.7
seconds. B.*A takes the same time (both A and B are 3000-by-3000).
The real-complex case runs at 25 MFlops, which is slow but then the
operation is memory-bound. When both are complex, the time is 0.8
seconds. These times sound reasonable to me. Am I wrong?

I have a 2Ghz Pentium 4 Mobile, with 1GB of memory.

Trevor wrote:
>
>
> 1. Sparse matrices of dimension greater than 2.

What do you want to do with them?

The sparse data structure for the 2D case is optimized for linear
algebra, not as a "data bucket". Thus, A(i,j) = whatever and
whatever=A(i,j) are not terribly fast ... Although they could be
better they will never be truly speedy.

ND matrices come up in tensor calculations, I think.

What the ND sparse case should look like depends strongly on what you
want to do with the beasts. If you want just sparse sub referencing
and sub assignment, that's one thing. If you want tensors, that's
another. (I'm not a tensor specialist, except in the morning when
I've had one two many tenser products via my morning coffee... ;-)

I want a scale arrow feature for use with the quiver function (ie, a
little arrow on the side of the plot that indicates the scale of the
arrow vectors). This is an absolute necessity when generating figures
for publication or presentations. Right now, I have a hacked version
of quiver that can be messily and painfully used to implement a scale
arrow (by pulling the scaling factor from quiver, overlaying a
separate axis on top of the first quiver plot, drawing a single arrow
to a set scale, and labeling it with text), but it leaves much to be
desired.

To have a switch on sparse matrices such that when an element is set
to zero, the inernal storage is not reset. Quite often I zero a row
in preparation for recomputing the non-zero elements, and the
internal resetting of the storage is needlessly time-consuming.

Richard Hindmarsh (remove '1's in e-mail)

This is, I think, impossible to achieve in R2007b but I hope it'll
come in R2008:

TMW, please, make Simulink for Mac look like the Windows version!

The Mac (unix) version looks ugly (choppy icons and fonts), has no
toolbar and no status bar (you can't tell if the simulation is
running!), and the load/save masks are terrible and not even coherent
with the Matlab ones. I'm sure there is a reason for this... but is
it so difficult to use graphics the Simulink interface?

In article <ef57195.10@webcrossing.raydaftYaTP>,
aggieinmissouri_DUDE@hotmail.com says...
> Well, since Loren has already said our pleas are falling on deaf
> ears:
> a girlfriend. It would be a bonus if she knew MatLab but I'll take
> whatever you send.
>

I deinitely did NOT say that. I just said we have a system for tracking
customer input and the way to use it best is to submit items via the web
form. As for a girlfriend, I am afraid that request will not go into
our queue :-)

-- Loren
http://blogs.mathworks.com/loren/

Richard Hindmarsh wrote:
>
>
> To have a switch on sparse matrices such that when an element is
> set
> to zero, the inernal storage is not reset. Quite often I zero a row
> in preparation for recomputing the non-zero elements, and the
> internal resetting of the storage is needlessly time-consuming.
>
> Richard Hindmarsh (remove '1's in e-mail)

That would be tricky to have a switch. Better one way or the other
with no switch. There are too many repercusions of this change to
make it easy to have a user-settable switch.

It would be better if MATLAB never dropped zeros due cancellation, in
my sparse opinion. Then if you really want them to go away, just do
A=sparse(A), which should be the only function that drops zeros (so
that the result A is the same regardless of whether or not the input
A is full or sparse.

S = alpha*A where alpha is a scalar should "never" drop zeros. But
then you would have odd cases like S = 0*A, which in my mind should
return spones(S) == spones(A), but then ... maybe that's a little
bizarre. I like it, though. It's a nice way of creating numerically
zero entry placeholders (if it existed). (OK, maybe my brain is
bizarre).

Regarding zeroing out a row, though. That's really slow. Try to
operate on columns as much as possible, since the matrices are stored
by column, not by row. Try timing x=A(i,:) and x=A(:,i) for a large
sparse A. The latter is much faster.

In article <ef57195.-1@webcrossing.raydaftYaTP>,
altmanyDEL@gmailDEL.comDEL says...
> This thread will be used to indicate the community's requests for
> features/bug-fixes for the upcoming Matlab release (R2007b, Sep 1
> 2007). This is in lieu of an official votable wishlist (see separate
> thread on this: Yair Altman, "Proposed Matlab votable wishlist" #, 16 May 2007 11:13 am </WebX?14@@.ef5718f>
> # ).
>
> Here are some items I would like to see:
> - fix the "clear java" bug (also clears globals as side-effect)
> - try-catch should also catch Java exceptions, not just Matlab ones
> - add an Excel-like iff function to the basic ML language (equivalent
> to C's a?b:c construct) - highly important in anonymous functions
> - enable multi-thread use of the Matlab computation engine from
> external C++/C#/Java threads
> - finally (after who knows how many years) fix the stupid zoom and
> similar problems with plotyy
>
> More to follow later - I think this is enough for now to start a
> healthy discussion. Please pitch in with your wishlist items.
>
> Yair Altman
>

I know some on this thread have mentioned that they requested a bug fix
and haven't seen it yet. Are you aware that you can now track bugs of
interest to you from the support web site?

http://www.mathworks.com/accesslogin/watchList.do

-- Loren
http://blogs.mathworks.com/loren/

On Fri, 18 May 2007 10:52:33 -0400, Loren Shure wrote:

> In article <ef57195.10@webcrossing.raydaftYaTP>,
> aggieinmissouri_DUDE@hotmail.com says...
>> Well, since Loren has already said our pleas are falling on deaf ears:
>> a girlfriend. It would be a bonus if she knew MatLab but I'll take
>> whatever you send.
>>
>>
> I deinitely did NOT say that. I just said we have a system for tracking
> customer input and the way to use it best is to submit items via the web
> form. As for a girlfriend, I am afraid that request will not go into
> our queue :-)

I think that many people have the impression that even submitting
requests to the system will be a waste of time because Mathworks won't
respond to them (which gives the impression of deaf ears). So many
people are voicing this ("I've been requesting this or reporting this for
the last 6 releases...").

I know that's my impression. I continue to request the same features
each release, but it's disappointing to continually get "that's not
supported and we have no plans for future support, or we won't tell you
if we do".

I even had a recent case where I could easily produce a Matlab crash with
about 3 lines of code. I was told that the functionality I was using
(figure handle) was unsupported. And not only that, because it was
unsupported, Mathworks would NOT even fix the crash. That's pathetic IMO.

Dan

> I know that's my impression. I continue to request the same
> features
> each release, but it's disappointing to continually get "that's not
>
> supported and we have no plans for future support, or we won't tell
> you
> if we do".

<rant>
That's certainly my impression too. I submitted a request for the
ability to rotate image axes with out the image disappearing. The
workaround of using a surf with Z ranging from 0 to 0 was not
practical because it is too slow for large images. I did receive an
email in response, basically saying "Oh, that's not supported". I
sort of got the impression that the person responding to me didn't
fully understand that the reason for requesting new features is that
the feature would be new, so just reminding us that it isn't
supported isn't much of a response.
</rant>

That said, I'd also like to see a couple of new uicontrols. A table,
similar to that used in the UI Property Inspector window, would be
nice. I'd also like to see real UI tabs.

On Fri, 18 May 2007 12:18:33 -0400, Gary wrote:

> That said, I'd also like to see a couple of new uicontrols. A table,
> similar to that used in the UI Property Inspector window, would be nice.
> I'd also like to see real UI tabs.

There's uitable which is extremely functional, but difficult to use
because "it's unsupported". We ended up writing a lot of code to make it
usable and act somewhat like the other uicontrols. And in the process we
ran into a number of bugs (some as simple as typos in the attributes),
but every time we sent in a bug report all we got was "it's unsupported,
so too bad and don't bother us". I've thought about putting it on Matlab
Central but haven't had any free time to select a license, etc. and get
it through company approval.

I think there is or at least used to be a tab capability too that's been
available for a long time, but "it's unsupported".

Dan

Dan Hensley wrote:
>
>
> On Fri, 18 May 2007 10:52:33 -0400, Loren Shure wrote:
>
>> In article <ef57195.10@webcrossing.raydaftYaTP>,
>> aggieinmissouri_DUDE@hotmail.com says...
>>> Well, since Loren has already said our pleas are falling on
deaf
> ears:
>>> a girlfriend. It would be a bonus if she knew MatLab but
I'll
> take
>>> whatever you send.
>>>
>>>
>> I deinitely did NOT say that. I just said we have a system for
> tracking
>> customer input and the way to use it best is to submit items
via
> the web
>> form. As for a girlfriend, I am afraid that request will not
go
> into
>> our queue :-)
>
> I think that many people have the impression that even submitting
> requests to the system will be a waste of time because Mathworks
> won't
> respond to them (which gives the impression of deaf ears). So many
>
> people are voicing this ("I've been requesting this or reporting
> this for
> the last 6 releases...").
>
> I know that's my impression. I continue to request the same
> features
> each release, but it's disappointing to continually get "that's not
>
> supported and we have no plans for future support, or we won't tell
> you
> if we do".
>
> I even had a recent case where I could easily produce a Matlab
> crash with
> about 3 lines of code. I was told that the functionality I was
> using
> (figure handle) was unsupported. And not only that, because it was
>
> unsupported, Mathworks would NOT even fix the crash. That's
> pathetic IMO.
>
> Dan
>

Dan, that sounds odd.

Can you post the lines of code?

Regards,
Eric Sampson
The MathWorks, Inc.

> ...I think there is or at least used to be a tab capability too
that's
> been
> available for a long time, but "it's unsupported".
>
> Dan
>

How do you locate/find out about these "unsupported" features?

Yair Altman wrote:
> finally (after who knows how many years) fix the
> stupid zoom and similar problems with plotyy

> That's got my vote!
> Kelvin Hales <khales@khace.com>

Yair and Kelvin - I'm curious to know which "stupid" plotyy problems
are annoying you. I'd also be curious what you think of my "plt"
alternative to plotyy and whether it avoids these stupid problems.

You can find plt here:
www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=4936
&objectType=file

Or if you want to check out the documentation first go here:
www.mennen.org/plt/plthome.htm

Although plt represents a modest shift in graphical and coding style
I'd like to think of plt as a viable alternative to plotyy for most
applications.

Perhaps the most significant advantage of plt may be its
documentation. From the many newsgroup questions about plotyy, it
seems few people can figure out how to get plotyy to do what they
want. I get practically no similar questions about plt. Maybe that is
because of the clear and extensive help file and breadth of the
example programs, or maybe it is because nobody is using it. (plt was
once Doug's "pick of the week" and has been downloaded over 2600
times, so I hope it's the former.)

~Paul
paul@m_e_n_n_e_n.org
(remove underscores from my email address)

Just to throw my 2 cents in:

Printing in deployed applications has been an enormous pain in the
a** since R14. The deployprint 'solution' has been a complete kludge
since then . It's painfully slowly improving but I'd like to see a
complete overhaul of this. And yes, I have submitted numerous tech
support/ enhancement requests for this over the years and had to get
a little ugly a couple of times.

Also why can't we get more documentation on the Java underpinnings of
Matlab. There's huge functionality waiting to be tapped there. I know
it's not for newbies but heck I'd even consider paying for the
documentation. We shouldn't have to rely on intrepid explorers like
Yair Altman (the 'Magellan' of Matlab?) to reveal these untapped
riches.

And please can we finally get a usable OOP system released. We see
hints of it using classdef the last couple of releases. See
memmapfile and a few other functions.

I would dearly like to see 64 bit support on Mac OS X. Can someone
from MathWorks comment about whether we will see this in 2007b or not?

Loren Shure wrote:

> I know some on this thread have mentioned that they requested a bug
> fix
> and haven't seen it yet. Are you aware that you can now track bugs
> of
> interest to you from the support web site?
>
> <http://www.mathworks.com/accesslogin/watchList.do>

You can track some bugs of interest, but not all. Of the six bug
reports I've submitted (all of which were confirmed as bugs in my
correspondence with the Mathworks following each report), only three
are listed on the site. Of those three, two were fixed in R2007a
(although I still take issue with the behavior of bufferm; I suppose
it now falls into the category of enhancement rather than bug though)
but one remains open. The three unlisted bugs (in ltlon2val,
setpostn, and polyxpoly) have remained untouched through the last few
releases.

Out of curiosity, how does the Mathworks decide which bugs to list on
the site? And which ones to fix? I do understand that the bugs I've
reported are part of the Mapping Toolbox, which is not one of your
most popular toolboxes, but shouldn't bug fixes be a priority even
before enhancements?

-Kelly

On Fri, 18 May 2007 12:43:38 -0400, Eric Sampson wrote:

>> I even had a recent case where I could easily produce a Matlab crash
>> with
>> about 3 lines of code. I was told that the functionality I was using
>> (figure handle) was unsupported. And not only that, because it was
>>
>> unsupported, Mathworks would NOT even fix the crash. That's pathetic
>> IMO.
>>
>> Dan
>>
>>
> Dan, that sounds odd.
>
> Can you post the lines of code?

Here is the code (I compressed my original example into one line):

uicontrol('parent',handle(uibuttongroup(figure)));

It crashes on both Win32 and Linux64.

On a related note: Please, please make handle objects supported!! They
are so useful! I use them all the time even though they aren't
officially supported. Consider this:

hb=handle(uicontrol(...));
hb.Position(3)=200;

Compared to the official way, which is bloated and harder to read:

hb=uicontrol(...);
pos=get(hb,'Position');
pos(3)=200;
set(hb,'Position',pos);

Dan

On Fri, 18 May 2007 12:50:05 -0400, Gary wrote:

>> ...I think there is or at least used to be a tab capability too
> that's
>> been
>> available for a long time, but "it's unsupported".
>>
>> Dan
>>
>>
> How do you locate/find out about these "unsupported" features?

I don't remember how I found them. It's probably from various sources.
You can learn about a lot of them here (there are many posts about how to
use uitable, which is a pretty popular unsupported feature).

You can also stumble on them using 'lookfor', for example 'lookfor tab'.
You run across one function called 'tabdlg' that looks rather interesting.

And of course if you peruse the toolbox folders in your installation, you
will also run across files that look kind of interesting.

Dan

Dan Hensley <somewhere@somewhere.withnospam> wrote in
news:pan.2007.05.18.18.09.30@somewhere.withnospam:

> uicontrol('parent',handle(uibuttongroup(figure)));

>> uicontrol('parent',handle(uibuttongroup(figure)));
??? Error using ==> uitools.uibuttongroup.uibuttongroup
Invalid input arguments

Error in ==> uibuttongroup at 22
h = uitools.uibuttongroup(varargin{:});

--
Scott
Reverse name to reply

On Fri, 18 May 2007 18:49:54 +0000, Scott Seidman wrote:

> uicontrol('parent',handle(uibuttongroup(figure)));

Oops, I missed a part. Try this:

uicontrol('parent',handle(uibuttongroup('parent',figure)));

It crashes on R2006a, R2006b, and R2007a. Sometimes I get the segfault
display in the Matlab window, sometimes Matlab just disappears. Actually
Matlab always just disappeared until just now when I tried it, and now
sometimes it manages to stay up.

Dan

In article <ef57195.21@webcrossing.raydaftYaTP>,
Gary <grubin698@gmail.com> wrote:
>That said, I'd also like to see a couple of new uicontrols. A table,
>similar to that used in the UI Property Inspector window, would be
>nice. I'd also like to see real UI tabs.

I would really like to be able to write

vc_mainplot.YLim = ylims*0.5;
ylims = vc_mainplot.YLim;

instead of

set(handles.vc_mainplot,'YLim',ylims*0.5);
ylims=get(handles.vc_mainplot,'YLim');

like I can in Delphi.

Francis

Samar Khatiwala wrote:
>
>
> I would dearly like to see 64 bit support on Mac OS X. Can someone
> from MathWorks comment about whether we will see this in 2007b or
> not?

I've been begging for this for quite a while, most of the answers
I've received point to the lack of 64-bit java, and other libraries
within Mac OS X. I'm hoping that this will be solved with the release
of Leopard - which is supposed to have 64-bit everything.

On Fri, 18 May 2007 14:12:19 -0500, Dan Hensley wrote:

> On Fri, 18 May 2007 18:49:54 +0000, Scott Seidman wrote:
>
>> uicontrol('parent',handle(uibuttongroup(figure)));
>
> Oops, I missed a part. Try this:
>
> uicontrol('parent',handle(uibuttongroup('parent',figure)));

Note that if you split the original line I posted into separate lines,
you don't need the 'parent' qualifier for uibuttongroup. Apparently
Matlab parses this differently. Try

hf=figure;
uicontrol('parent',handle(uibuttongroup(hf)));

Dan

On Fri, 18 May 2007 19:27:04 +0000, Francis Burton wrote:

> I would really like to be able to write
>
> vc_mainplot.YLim = ylims*0.5;
> ylims = vc_mainplot.YLim;


You can, if you turn your handle into a handle object.

vc_mainplot=handle(vc_mainplot);

Of course "it's unsupported".

Dan

I'll have you guys know that I was at a meeting this week at
Mathworks for a couple of days. Believe it or not Cleve Moller
(Matlab inventor/chairman/chief scientist) still responds to some bug
reports (he actually was looking at one during our meeting).

Also know that quality control is very important to Mathworks.
However, you must remember that Mathworks like anyone else only has
so many resources and their priorities may not be the same as yours.

For example, if you're interested in parallel computing, Mathworks is
making great strides at developing and improving the DCT. This is
one of their priorities.

Though, I do feel that bug fixes should be a higher priority than
developing new product.

Loren - I must echo Kelly and others' frustration. You're referring
us to the online buglist, but this list is incomplete at best. Just
in the past year I submitted over a dozen confirmed bugs, none of
which (except one) appear in the list (I just rechecked). I can send
you the internal issue IDs if you want. And these are all in fully
documented territory - I don't normally bother submitting issues in
undocumented areas. I'm actually very pleased with the promptness and
professionalism of your tech support staff, but if bugs are not
published, and if bugs are not fixed over a very long time (e.g., the
plotyy zoom bug, which is still unlisted BTW), then this effort is
only half-done and submitters like me get dishearted and frustrated.

Regarding the requested new features, these can't be seen anywhere.

As you can see from this highly active thread, there's an enormous
willingness by ML users to help TMW improve ML, and a frustration
that this willingness appears unanswered by TMW. Perhaps my original
suggestion of a votable wishlist is not the best mechanism. But
surely something can be found to improve the interaction with the ML
community and increase our participation in the dev-roadmap process.
We're all on the same side after all.

We're offering our help and advise willingly and freely - please use
it.

Yair Altman

Yair Altman:
<SNIP wish-list evergreen...

> This thread will be used to indicate the community's requests for
features/bug-fixes for the upcoming Matlab release...

i've been somewhat active in CSSM since its birth in 1993 (under
various names/aliases)...
very well-meant requests (similar to this OP) have come up
periodically - they NEVER EVER yielded anything substantial...

...NOR does TMW's official, mind-soothing placebo bug-site, which was
mentioned earlier in this thread...

unfortunately, the very founders of ML (little/moler) have come of
age (like all of us) and - by law of nature - lost touch with the
daily operations of their own domain... (i do remember the days when
both of them vividly and visibly participated in this NG)

anyhow ,this thread is a good vessel to vent your hopes for a better
(ML) world, frustration, or even anger...

us

Dear <us> wrote:

> Yair Altman:
> <SNIP wish-list evergreen...
> ...
> very well-meant requests (similar to this OP) have come up
> periodically - they NEVER EVER yielded anything substantial...

Well, I more than rarely have the same feeling. But, at a second
look, over time many good suggestions have been included in new
versions, but not necessarily in the order and pace that I/we/user(s)
would have wished (which typically is pronto if not yesterday :-).
Having said that, there are some rather long-standing requests that
tend to come up again and again here at CSSM.

One example, that currently causes much problem for me is the
unfortunate mixture of graph rendering and display engine that forces
me to do any serious plotting in interactive mode (suggestions in a
threadlast summer did not work out). Searching CSSM ("batch plotting"
etc) we find that this is a decadal-long story:
<342B416D.D2DBAEB6@mirinz.org.nz> is the earliest I found in a
quick check.

I aired my "batch plotting" problems last summer here at CSSM
<ef3c3ec.-1@webcrossing.raydaftYaTP> and was interacting with
TMW Support (and through them, I take it, with the developers). The
submitted bug/RFE is not even visible in
> ... TMW's official, mind-soothing placebo bug-site,

On the other hand, I understand that there is a completely new
graphical engine on its way (question is whether this is soon enough)
...

> anyhow ,this thread is a good vessel to vent your hopes for a
> better
> (ML) world, frustration, or even anger...

And for 2007b, I wish for a
TMW official, mind-blowing, pleasantly accurate bug-site,
that I will not need to use anyway...

Best,
Lars

Dan Hensley wrote:
> And of course if you peruse the toolbox folders in your
> installation, you
> will also run across files that look kind of interesting.
>
> Dan
>
  
Dan,
Thanks for pointing these out. It is nice to see that the uitab and
uitable do exist (sort of). I would still like to see uitab as a
GUIDE object.

In article <ef57195.25@webcrossing.raydaftYaTP>, Paul Mennen wrote:
> Yair Altman wrote:
> > finally (after who knows how many years) fix the
> > stupid zoom and similar problems with plotyy
>
> > That's got my vote!
> > Kelvin Hales <khales@khace.com>
>
> Yair and Kelvin - I'm curious to know which "stupid" plotyy problems
> are annoying you. I'd also be curious what you think of my "plt"
> alternative to plotyy and whether it avoids these stupid problems.

The fact that zoom does not work simultaneously on both axes (zooms only
on the currently selected axis).

I've not tried your "plt" alternative. The alternative 'zoom' function
<http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectI
d=1137> solved the problem; but has not been updated for a long time
now.

Kelvin Hales wrote:
>
>
> In article <ef57195.25@webcrossing.raydaftYaTP>, Paul Mennen
wrote:
>> Yair Altman wrote:
>> > finally (after who knows how many years) fix the
>> > stupid zoom and similar problems with plotyy
>>
>> > That's got my vote!
>> > Kelvin Hales <khales@khace.com>
>>
>> Yair and Kelvin - I'm curious to know which "stupid" plotyy
> problems
>> are annoying you. I'd also be curious what you think of my
"plt"
>> alternative to plotyy and whether it avoids these stupid
> problems.
>
> The fact that zoom does not work simultaneously on both axes (zooms
> only
> on the currently selected axis).
>
> I've not tried your "plt" alternative. The alternative 'zoom'
> function
> <http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectI
> d=1137> solved the problem; but has not been updated for a long
> time
> now.

For whomever is interested, the simple fix to the plotyy-zoom problem
is to call linkaxes for the two axes handles returned by the plotyy
function (see below). So it works for me, but this should be placed
within Matlab's zoom.m so that *everyone* enjoys the fix. For old
Matlab versions that don't have linkaxes, there are other solutions -
one of which Kelvin pointed out above. I just used this bug to point
out that MathWork's bug-list is, well, buggy...

[BTW, MathWorks finally added a comment line about linkaxes to the
bottom of plotyy's help comment, but never bothered to fix the code
itself...]

linkaxes(axout,'xy');
set(ax,'YTickMode','auto');

Yair Altman

Yair Altman wrote:
> [BTW, MathWorks finally added a comment line about linkaxes to the
> bottom of plotyy's help comment, but never bothered to fix the code
> itself...]

scrap this side-note: I mixed up two unrelated issues. I stand behind
the rest of what I posted.

Yair Altman

I suppose it depends on how you want to define threads, but I would
like to make a couple of comments. A Mathworks person might tell you
that Matlab is already multithreaded. Of course what you're really
interested in is multiple computational threads.

Of course you could just start multiple Matlab sessions. I don't
like this method either.

Another alternative is the distributed processing toolbox. With
2007a, you now have the option of having 4 workers on your local
machine without having to buy the computing engine.

 ryan kelly wrote:
>
>
> I want THREADS!!!!!!!!!!!
>
> Yair Altman wrote:
>>
>>
>> Kelvin Hales wrote:
>>>
>>>
>>> In article <ef57195.25@webcrossing.raydaftYaTP>, Paul
>> Mennen
>> wrote:
>>>> Yair Altman wrote:
>>>> > finally (after who knows how many years) fix the
>>>> > stupid zoom and similar problems with plotyy
>>>>
>>>> > That's got my vote!
>>>> > Kelvin Hales <khales@khace.com>
>>>>
>>>> Yair and Kelvin - I'm curious to know which "stupid"
> plotyy
>>> problems
>>>> are annoying you. I'd also be curious what you think of
> my
>> "plt"
>>>> alternative to plotyy and whether it avoids these
stupid
>>> problems.
>>>
>>> The fact that zoom does not work simultaneously on both
axes
>> (zooms
>>> only
>>> on the currently selected axis).
>>>
>>> I've not tried your "plt" alternative. The alternative
'zoom'
>>> function
>>> <http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectI
>>> d=1137> solved the problem; but has not been updated for
a
>> long
>>> time
>>> now.
>>
>> For whomever is interested, the simple fix to the plotyy-zoom
>> problem
>> is to call linkaxes for the two axes handles returned by the
> plotyy
>> function (see below). So it works for me, but this should be
> placed
>> within Matlab's zoom.m so that *everyone* enjoys the fix. For
old
>> Matlab versions that don't have linkaxes, there are other
> solutions
>> -
>> one of which Kelvin pointed out above. I just used this bug to
>> point
>> out that MathWork's bug-list is, well, buggy...
>>
>> [BTW, MathWorks finally added a comment line about linkaxes to
> the
>> bottom of plotyy's help comment, but never bothered to fix the
> code
>> itself...]
>>
>> linkaxes(axout,'xy');
>> set(ax,'YTickMode','auto');
>>
>> Yair Altman

Gary wrote:
> Thanks for pointing these out. It is nice to see that the uitab
and
> uitable do exist (sort of). I would still like to see uitab as a
> GUIDE object.

Just FYI, these are already available in GUIDE, but only internally
to TMW developers. TMW does not wish to promote Java-related
components (like the ui* family). The semi-official reason I heard is
that it's very easy for non-experienced programmers to hang Matlab
and to cause errors, due to conflicts between the Matlab
computational thread (which is still single-threaded for historical
reasons, although this is slowly being changed) and the Java event
thread. Personally I believe we're grown ups and Matlab should let us
decide whether or not we wish to take the risk. But apparently TMW
thinks otherwise.

Yair Altman

This is a very interesting discussion that provided a lot of
information that is usually not publicly known. I have the impression
that exactly this makes the difference: The existing buglist is not
completely visible to everyone and seems to be treated like a
business secret by TMW. Result: There is no fruitful interaction
between people (users) who are actually interested in getting the
bugs fixed. Also, there is no exchange of workarounds, which might
depend on application. Having not only one, but perhaps a set of
workarounds, discussing their pros and cons, etc, can change the
amount of trouble that users have with the related bugs completely.

It appears to me that as much open communication as possible is a key
to maximizing the usefulness of complex software tools. You can study
this in the open source world: There are quite a lot of free/open
source software products now that challenge commercial products with
a big headcount of developers. It is safe to assume that much more
workforce is available on the commercial side, so how is this
possible? I think it comes from all those wikis/forums/mailing lists
that enable developers and users to get in touch directly. It is
especially useful if scientists/engineers are on both sides of the
counter.

Conclusion: If open communication is a possible reason for open
source software possibly being developed efficiently, anyone who
wants to sell software should study how to benefit from it.

CSSM is a good example. Traditionally, TMW support is expected to be
the prime source of information if you have a Matlab-related problem.
But, at least for me, it is CSSM. Consequently, and very good, some
TMW employees are very active here. But why not take the next step
and provide more information, organized in a modern way? So here is
my suggestion to TMW:

1. create a (second?), publicly available website with databases for
bugs, workarounds, vote for new features etc.
2. everything on this site is visible to everyone, even without a
login, and with fast and easy browsing
3. host the site, but give control over it to a committee of
long-term, active, respected users that you can trust
4. it should be possible to find members for this commitee here, as
there are regulars who are obviously willing to invest time in Matlab
5. instead of letting everybody publish their latest problem there,
let the commitee decide:
5.1 everybody can input reports,
5.2 but these become visible to the public only if approved by a
commitee member
5.3 provide software to make it fast and easy for the commitee
members to do you this favor in a reasonably amount of time, i.e.:
user input is presented in a form together with options how to deal
with it in a mouseclick, like
5.3.1 reject without reply (SPAM, pure rant, ...)
5.3.2 reject with comment, link to CSSM, etc.
5.3.3 insert to database as new item
5.3.4 add to existing item
6. understand that such a database can be useful, even if it can put
you under pressure sometimes

Another approach would be to modify the concept of a resident
engineer that is common to some industries, and let this person run
the described site. Or a combination of both. The basic idea is to
separate the wishlist-assembly from other interests, such that an
input for the wishlist still might not get implemented, but not
because it does not get known or discussed.

Heinrich

"Heinrich Acker" <Firstname.Lastname@web.de> wrote in news:ef57195.47
@webcrossing.raydaftYaTP:

> , such that an
> input for the wishlist still might not get implemented, but not
> because it does not get known or discussed.
>

If nothing else, doesn't this thread show that anybody who wants to post a
"wish" can get it discussed? Wouldn't keeping an easy-to-find trace of it
involve no more than making sure "wish" appears in the subject?

Even without the subject line standard, a google groups search for "wish"
in *.matlab yielded mkore than 5,000 hits.


--
Scott
Reverse name to reply

Scott Seidman wrote:
>
>
> Even without the subject line standard, a google groups search for
> "wish"
> in *.matlab yielded mkore than 5,000 hits.

Why don't you just google for 'wish matlab'? This ensures that you
get even more hits from the all-knowing, all-seeing trash heap.

Seriously, what I had in mind is an organized process that solves a
problem that some posters described: That their reasonable input got
only 'not supported' as reply. Which is not exactly a discussion. As
you know, everybody can publish anything these days. Getting someone
who counts to listen is something different. You can't expect a good
result from distilling your 5,000 hits to a wish list.

Heinrich

"Heinrich Acker" <Firstname.Lastname@web.de> wrote in news:ef57195.49
@webcrossing.raydaftYaTP:

> That their reasonable input got
> only 'not supported' as reply. Which is not exactly a discussion. As
> you know, everybody can publish anything these days. Getting someone
> who counts to listen is something different.


And the wishlist organization will guarantee that the powers that be listen
by ....?


--
Scott
Reverse name to reply

I guess I have lower expectations. :) I have been very happy with
the improvements in MATLAB over the last couple of years:

1. bsxfun
2. statistical arrays in the stats toolbox
3. publishing
4. 64bit matlab
5. inputParser
6. .net compiler
7. java compiler
8. optimization tool graphical interface
9. mac / intel
to name a few

Recently I switched careers to a software developer for a
professional software company, I can appreciate what it takes to get
a release out.

Stephen

I think my absolute number one wish right now would be a function to
programmatically install a database as an ODBC data source. The ODBC
Administrator always seems easy until I have to show somebody how to
use software I've written, and the digging around in the Windows
control panel applications becomes absolutely embarassing.

Darik Gamble wrote:
>
>
> I think my absolute number one wish right now would be a function
> to
> programmatically install a database as an ODBC data source.
[snip]

It may take a few days but somewhere in my archives I have code to do
that. I forget if it was a vb com dll I wrote w/ source and callable
from matlab, or a registry entry I did with matlab.

I'll post it here when I find it unless someone comes up with a
better answer.

Stephen

Scott Seidman wrote:
>
>
> "Heinrich Acker" <Firstname.Lastname@web.de> wrote in
> news:ef57195.49
> @webcrossing.raydaftYaTP:
>
>> That their reasonable input got
>> only 'not supported' as reply. Which is not exactly a
discussion.
> As
>> you know, everybody can publish anything these days. Getting
> someone
>> who counts to listen is something different.
>
>
> And the wishlist organization will guarantee that the powers that
> be listen
> by ....?
>
>
> --
> Scott
> Reverse name to reply
>

Scott, I don't understand why you are so negative about this, in this
thread and the other wish-list related. Do you think that it is
better not to try things that do not guarantee success?

Heinrich

"Heinrich Acker" <Firstname.Lastname@web.de> wrote in news:ef57195.54
@webcrossing.raydaftYaTP:

> Scott Seidman wrote:
>>
>>
>> "Heinrich Acker" <Firstname.Lastname@web.de> wrote in
>> news:ef57195.49
>> @webcrossing.raydaftYaTP:
>>
>>> That their reasonable input got
>>> only 'not supported' as reply. Which is not exactly a
> discussion.
>> As
>>> you know, everybody can publish anything these days. Getting
>> someone
>>> who counts to listen is something different.
>>
>>
>> And the wishlist organization will guarantee that the powers that
>> be listen
>> by ....?
>>
>>
>> --
>> Scott
>> Reverse name to reply
>>
>
> Scott, I don't understand why you are so negative about this, in this
> thread and the other wish-list related. Do you think that it is
> better not to try things that do not guarantee success?
>
> Heinrich
>


I'm not negative on it, I just don't see the point of pursuing it. Your
latest description of a committee, or a coordinating engineer, is not
viable, as it is hard to implement and maintain, and frankly, I simply do
not understand the function you're proposing that couldn't be served
right now and on this group. I don't understand what a vote gets us.

The Mathworks, right now, is in the unenviable position of trying to get
people to pay a fairly high price for updates in a subscription model of
an EXTREMELY mature product. They can either sit back on their heels and
let hardware and OS compatibility issues drive updates, hope that the
userbase continuously expands and they sell more copies through growth,
or make their product more exciting so people will want to upgrade--even
if people aren't forced to because of OS issues. I'm confident that they
have people closely following what upgrades, improvements, and new
features people want, and then they make a business-savvy decision that
includes such matters as value added vs. resources available, not to
mention the capabilities of competing products and other options their
user base must have.

Adding a rogue wishlist committee is a) not likely to have any impact in
a way that matters, b) likely redundant with what's going on internally
at Mathworks, and c) doing Mathworks' work for them, and for free-- if
such a structure would be valuable to Mathworks, they should be footing
the bill.

We all have our wishlists. For example, I've had a large number of
graphical issues over the years, particularly with stuff like hard copy
looking different from what's on the screen. It's better than it was,
but still not great. I've got some wonderful graphical applications that
don't seem to care what renderer I'm using, but they give me pretty good
and realiable wysiwyg. Mathworks obviously knows that there are issues
here that haven't been optimized. They also have a really good estimate,
I'm sure, of what it would cost to fix such problems, how many customers
will seek an alternative product because of such problems, how well their
competitors deal with such problems, whether there are more pressing
problems that trump this one (boy, people drop a product like flies if
the latest version doesn't work with latest OS or processor), and a host
of other decision making inputs. Given all this, it's pretty clear that
these graphics problems I'm referring to don't end up having an immediate
priority when spit out of the decision machine-- and push comes to shove,
Mathworks must be correct, because I haven't opted to vote with my feet
despite such issues.

I guess what I'm saying is we can provide a wishlist through a variety of
mechanisms, but we don't have much insight to any of the other inputs to
this fairly complex multi-input system. I'm also suggesting that the
Mathworks already has a fairly good estimate of what this wishlist would
work out to be, and that if they appear unresponsive, its more likely
because of these inputs we don't know about than because they don't know
how users would prioritize the wishlist. Mathworks is a business, and
they don't have to let their customers and competitors in on the decision
making process.

Given all this, I just don't see the utility of a voting wishlist. I
don't believe it would do any more than would submitting an enhancement
suggestion and posting in parallel to this group.

--
Scott
Reverse name to reply

Scott Seidman wrote:
>
>
> Steve.Amphlett@ricardo.com wrote in news:1179926503.567326.67230
> @p77g2000hsh.googlegroups.com:
>
>> One thing I
>> can add though is that it's very rare for customer wishlists
> (that
>> I've seen in the course of my work) to ever contain strategic
> items
>> that ensure future success of the product. They are generally
> tweaks
>> and extensions that are the pet peeves of a limited number of
>> customers - often "required" to solve problems that have much
> better
>> solutions already, or already in the pipeline.
>>
>
> I actually agree with this pretty closely. Strategic items are
> sort of
> "what the customer REALLY wants in terms of enhanced functionality,
> but
> doesn't even know enough to ask for it"
>
> --
> Scott
> Reverse name to reply
>

I understand your and Steve's explanation and agree with your
description of what probably happens at TMW. What I still don't
understand is this course:

Somebody has a suggestion -> I don't like it -> I think it
won't help -> but it doesn't harm me -> so I write a long
article in order to kill the idea!

What's the benefit?

This thread has shown that some customers are not satisfied with some
things (we have been talking about bugs too, btw). If your
explanation of the business decisions is correct, somebody at the TMW
might see a need to improve a process once this feedback is above a
certain threshold, whatever the metric is. It is less likely if other
customers essentially explain - for free! - that it's no use to
question the business decisions of successful companies. If you too
have unfixed bugs or something else on your wishlist, I doubt that
your last posts were in your own interest (In the sense that either
nothing changes anyway as a result of what we write, or, if changes
are possible, you helped to make them less likely to happen.).

Again, I think that everything you wrote is correct. But the
strategic component (of a customer, not a supplier!) is missing.

For me, it was not just an abstract idea when I wrote about open
source software challenging commercial products. I still use Matlab a
lot, but I'm currently learning Python. The difference I see is that
there are no business decisions between the developers and the users,
only standing in the way from a user's perspective.

TMW's business decisions probably can't be influenced by what we
write here, but I can tell that I'm influenced by them regarding a
possible switch to Python.

Heinrich

On Wed, 23 May 2007 12:24:50 -0400, Heinrich Acker wrote:


> This thread has shown that some customers are not satisfied with some
> things (we have been talking about bugs too, btw). If your explanation
> of the business decisions is correct, somebody at the TMW might see a
> need to improve a process once this feedback is above a certain
> threshold, whatever the metric is. It is less likely if other customers
> essentially explain - for free! - that it's no use to question the
> business decisions of successful companies. If you too have unfixed bugs
> or something else on your wishlist, I doubt that your last posts were in
> your own interest (In the sense that either nothing changes anyway as a
> result of what we write, or, if changes are possible, you helped to make
> them less likely to happen.).


I agree. TMW is behaving like most other large software companies.
While the business model is successful, I think it's time for it to be
challenged.

One problem is that a lot of resource is spent on new flashy gee-whiz
features that make good demos and make it easy for the sales guys to wow
the PHB's and make more sales. But these resources take away development
resource for fixing bugs and making other improvements. For example, a
lot of resource has been spent on nice new interactive plot features, but
I don't even know what's available because I don't use them.

More sales are necessary to provide the resources necessary for
development, but wouldn't it be novel for just one release to be 100%
focused on fixing bugs and responding to longtime customer requests?
Think it will ever happen with commercial software? Probably not a
chance.


> Again, I think that everything you wrote is correct. But the strategic
> component (of a customer, not a supplier!) is missing.
>
> For me, it was not just an abstract idea when I wrote about open source
> software challenging commercial products. I still use Matlab a lot, but
> I'm currently learning Python. The difference I see is that there are no
> business decisions between the developers and the users, only standing
> in the way from a user's perspective.


I am happy that open source is successful, because it does challenge the
status quo. It's a very different way of approaching things, and of
course companies are still trying to figure out viable business models
from this, but many have and are quite successful. Commercial software
companies have a lot to learn from the open source model.

Dan

Dan Hensley wrote:
>
>
>
> I am happy that open source is successful, because it does
> challenge the
> status quo. It's a very different way of approaching things, and
> of
> course companies are still trying to figure out viable business
> models
> from this, but many have and are quite successful. Commercial
> software
> companies have a lot to learn from the open source model.

I really wish it were that simple. But in our experience, you get
what you pay for. Thousands of man-hours wasted on trying to build
software on ever-moving and ill-defined foundations. The big boys
(e.g. Rational/IBM) will only support commercially supported Linux
distros, which kind of says it all for me (note: at home now, not
company view).

Steve Amphlett wrote:
>
>
> Dan Hensley wrote:
>>
>>
>>
>> I am happy that open source is successful, because it does
>> challenge the
>> status quo. It's a very different way of approaching things,
and
>> of
>> course companies are still trying to figure out viable business
>> models
>> from this, but many have and are quite successful. Commercial
>> software
>> companies have a lot to learn from the open source model.

> The big boys
> (e.g. Rational/IBM) will only support commercially supported Linux
> distros, which kind of says it all for me

Just a tiny little note.
There was not such a thing as a "commercially supported Linux"
untill it reached success

On Wed, 23 May 2007 16:21:20 -0400, Steve Amphlett wrote:

> Dan Hensley wrote:
>>
>>
>>
>> I am happy that open source is successful, because it does challenge
>> the
>> status quo. It's a very different way of approaching things, and of
>> course companies are still trying to figure out viable business models
>> from this, but many have and are quite successful. Commercial software
>> companies have a lot to learn from the open source model.
>
> I really wish it were that simple. But in our experience, you get what
> you pay for. Thousands of man-hours wasted on trying to build software
> on ever-moving and ill-defined foundations. The big boys (e.g.
> Rational/IBM) will only support commercially supported Linux distros,
> which kind of says it all for me (note: at home now, not company view).

I'm talking about stuff like not being opaque about the development
process or future plans, etc. The goal of some projects to release when
the code is done, not when a certain date has been reached, etc. Release
early, release often (which has pros and cons).

For every example (open source, commercial, you name it) there is an
equal and opposite example.

Dan

Matlab should have tasks like in Ada.

StephenLL wrote:
> It may take a few days but somewhere in my archives I have code to
> do
> that. I forget if it was a vb com dll I wrote w/ source and
> callable
> from matlab, or a registry entry I did with matlab.
>
> I'll post it here when I find it unless someone comes up with a
> better answer.
>
> Stephen

I would really, really appreciate that, Stephen.

Joseph wrote:
>
>
> Well, since Loren has already said our pleas are falling on deaf
> ears:
> a girlfriend. It would be a bonus if she knew MatLab but I'll take
> whatever you send.
>

What a silly question; why ask The MathWorks for something that is
present in R2007a already? Just use the MATLAB statement:

find (triu (lu (ver)), 'first') ;

For best results, this should be followed by the unary parenthetical
negation operator, as in:

find (triu (lu (ver)), 'first') ;-)

See? All good things come to them that know MATLAB.

>> find (triu (lu (ver)), 'first') ;
??? Undefined function or method 'lu' for input arguments of type 'struct'.

I guess MATLAB is a believer of free love (DISCLAIMER: in my personal
opinion). ;-)

---Bob.

"Tim Davis" <davis@cise.ufl.edu> wrote in message
news:ef57195.68@webcrossing.raydaftYaTP...
> Joseph wrote:
>>
>>
>> Well, since Loren has already said our pleas are falling on deaf
>> ears:
>> a girlfriend. It would be a bonus if she knew MatLab but I'll take
>> whatever you send.
>>
>
> What a silly question; why ask The MathWorks for something that is
> present in R2007a already? Just use the MATLAB statement:
>
> find (triu (lu (ver)), 'first') ;
>
> For best results, this should be followed by the unary parenthetical
> negation operator, as in:
>
> find (triu (lu (ver)), 'first') ;-)
>
> See? All good things come to them that know MATLAB.

Two things I can think of that I'd love.
1. A ++ operator (and related --, +=, etc...)
2. We have built in fid 1 (stdout) and 2 (stderr). I'd
also like 0, which would mean, write it to "nowhere."

Also, I'll echo the desire for a
( a ? b : c )
or ternary operator.

In article <fa7edm$r14$1@fred.mathworks.com>,
John <johnmarcovici@junkhotmail.com> wrote:
>Two things I can think of that I'd love.
>1. A ++ operator (and related --, +=, etc...)

That would significantly change the semantics of the language.
At present, the only way to modify a value is to have it appear
at the left side of an assignment -- which has implications for
optimizations and JIT. ++ and -- and += modify values in-place,
and -potentially- make the optimization / JIT much harder.
Unless, that is, you restrict the use of the operator to be the
sole assignment on the line, such as

  count++;
  count += 5;

But what should we make of the semantics of

  [X, Y, Z]++;
  [X, Y, Z] += 5;
  [X, Y, Z] += [P,Q,R];
  [X, Y, Z] += P{:};
  [X, Y, Z] += deal([P,Q,R]);

??


>2. We have built in fid 1 (stdout) and 2 (stderr). I'd
>also like 0, which would mean, write it to "nowhere."

0 is stdin, already used

If you want a fd that writes to "nowhere", then

if pc
  nullfile = 'NUL:'
else
  nullfile = '/dev/null'
end
fd = fopen(nullfile, 'w')

This should work on all currently supported platforms (use
'NL:' if you are still using VMS)

>Also, I'll echo the desire for a
>( a ? b : c )
>or ternary operator.

Someone suggested using :: because : is already in use.

I'm not too sure of the meaning of,

 (a ? b{:} :: c(:))

The way {:} is used semantically in Matlab,

 (a ? b{:})

should stand for (a ? b{1} :: b{2}) if b has exactly two elements
and b{:} should be an error in the middle position of the ternary
operator if b does not have exactly 1 or 2 elements.
--
  Prototypes are supertypes of their clones. -- maplesoft

roberson@ibd.nrc-cnrc.gc.ca (Walter Roberson) wrote in
message <fa7glf$ftb$1@canopus.cc.umanitoba.ca>...
> In article <fa7edm$r14$1@fred.mathworks.com>,
> John <johnmarcovici@junkhotmail.com> wrote:
> >Two things I can think of that I'd love.
> >1. A ++ operator (and related --, +=, etc...)
>
> That would significantly change the semantics of the
language.
> At present, the only way to modify a value is to have it
appear
> at the left side of an assignment -- which has
implications for
> optimizations and JIT. ++ and -- and += modify values in-
place,
> and -potentially- make the optimization / JIT much
harder.
> Unless, that is, you restrict the use of the operator to
be the
> sole assignment on the line, such as
>
> count++;
> count += 5;
>
> But what should we make of the semantics of
>
> [X, Y, Z]++;
> [X, Y, Z] += 5;
> [X, Y, Z] += [P,Q,R];
> [X, Y, Z] += P{:};
> [X, Y, Z] += deal([P,Q,R]);
>
> ??
>
>
> >2. We have built in fid 1 (stdout) and 2 (stderr). I'd
> >also like 0, which would mean, write it to "nowhere."
>
> 0 is stdin, already used
>
> If you want a fd that writes to "nowhere", then
>
> if pc
> nullfile = 'NUL:'
> else
> nullfile = '/dev/null'
> end
> fd = fopen(nullfile, 'w')
>
> This should work on all currently supported platforms
(use
> 'NL:' if you are still using VMS)
>
> >Also, I'll echo the desire for a
> >( a ? b : c )
> >or ternary operator.
>
> Someone suggested using :: because : is already in use.
>
> I'm not too sure of the meaning of,
>
> (a ? b{:} :: c(:))
>
> The way {:} is used semantically in Matlab,
>
> (a ? b{:})
>
> should stand for (a ? b{1} :: b{2}) if b has exactly
two elements
> and b{:} should be an error in the middle position of
the ternary
> operator if b does not have exactly 1 or 2 elements.
> --

Thank you for the slick nullfile tip. (Now think how clean
your example code would be with the ternary!) That will be
functionally equivalent to what I want although I'll
continue to wish for a built in "nowhere" fid.

How to interpret all the possibilities of ++, --, etc....?
That's we pay TMW TBB (the big bucks.)

In article <fa7i33$5iu$1@fred.mathworks.com>,
John <johnmarcovici@junkhotmail.com> wrote:

<Snip -- no need to quote everything you are replying to>

>Thank you for the slick nullfile tip. (Now think how clean
>your example code would be with the ternary!)

In many cases, you don't need the full power of ?:

CTern = @(condition,thecell) thecell{condition+1}

fopen( CTern(ispc, {'/dev/null', 'NUL:'}), "w" );

(Note: I used pc before where I should have used ispc)

Similarily,

ATern = @(condition,thearray) thearray(condition+1)
--
  Is there any thing whereof it may be said, See, this is new? It hath
  been already of old time, which was before us. -- Ecclesiastes

For my money, I'd like a section (perhaps also m-lint
guidelines) on optimising my m-files for multithreaded
computing.

I know that the Distributed Computing Toolbox is pretty
comprehensive; but most modern computers have multiple cores
now. Without the DCT, it's still possible to enable
multithreaded computation within Matlab- but I don't think
my m-files make the best use of the capability.

What else? Ummm... Publishing directly to PDF format would
be useful.

Free lollipop in the upgrade box.

An update to the help browser would be fantastic - even
Internet Explorer has tabbed browsing these days!

[0] Optional stack and heap control. If a calculation will
run out of memory, allow the ability to optionally use disk
space as virtual memory **independent** of the OS. For
example, inform Matlab that K:\ (windows) which has a 250 GB
drive can be used for up to 50 GB (say) of virtual memory.
Let matlab open memory-mapped file(s) to use as "last
resort" memory.

[1] Command window unicode: allow fprintf-style outputs
other than English. For example, Asian scripts.

[2] Command window font control: fprintf-style outputs with
colors, font control and so on. Combine with [1]?

[3] Support at least one open source C++ compiler for MEX
generation for win32/64 platforms (even if its limited to
gcc). Limiting use to the Lcc C-compiler is rather annoying.

[4] Fast kill option. Too many times matlab locks in a deep
calculation I'd like to quickly kill (sometimes even
taskmangler can't kill it). Use a separate process to
monitor the 'fast-kill' debug button. If this is not
realistic, allow matlab to be queried by a command window to
determine running processes within it and allow them to be
individually terminated. Without bringing the entire thing down.

[5] Optional figure redraws. With multiple figure windows
open or particularly complicated 3d plots being rotated,
matlab will slow to a crawl trying to redraw them. One at a
time, front to back. Allow user to specify certain figures
as "yes/no automatic refresh"

[6] Help file: A definitive, official matlab-language
specification including operator precedence and so on.

As you can see, my email address is an incinerator, so
please post all replies here.

> I know that the Distributed Computing Toolbox is pretty
> comprehensive; but most modern computers have multiple cores
> now. Without the DCT, it's still possible to enable
> multithreaded computation within Matlab- but I don't think
> my m-files make the best use of the capability.

The latest Matlab releases (starting with R2007a) include
support for multiple cores. However, Matlab will only take
advantage of multiple cores in evaluating certain
computations like an FFT or a FIR filtering operation.
Matlab will never be able to determine if, for example,
consecutive function calls in a for-loop are independent of
each other.

This is a citation of the documentation in the multicore
toolbox on Matlab Central. Check it out here:
http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=13775

I think it much easier to work with this package than with
the Distributed Computing Toolbox, because it is all plain
Matlab. And it is cheaper...

Markus

Walter Roberson wrote:

>
>
>> [2] Command window font control: fprintf-style outputs with
>> colors, font control and so on. Combine with [1]?
>
> If I have a user interaction that needs colours or fonts, then
> I would put it into a GUI window. If I'm outputting something to
> the command window, then my assumption would be that the command
> window has at least a 1 in 3 chance of being run in a non-graphical
> environment, -nodesktop, with the only available user interactions
> those that are representable by character streams that are interpreted
> by the user's terminal or terminal emulator. There is an ANSI X3.64
> standard for a small number of foreground and background colours
> via escape sequences (8 of each); those are about the closest you are
> going to get to a standard way of representing colour in terminal
> emulators. There is no, however, no sequence within X3.64 for
> representing font choice.


If the terminal doesn't support colors, then don't display in color. If
it does, then do. Frequently it does not make sense to put output in a
GUI just to get color. I would really like Matlab to be able to support
  colored, underlined, etc. output.

Or maybe said in a different way: if the user is running the Java
desktop, display with all of the formatting the user specified. If
writing to a text file (i.e. diary, or -nodesktop, etc.) then strip off
all formatting.


>> [4] Fast kill option. Too many times matlab locks in a deep
>> calculation I'd like to quickly kill (sometimes even
>> taskmangler can't kill it). Use a separate process to
>> monitor the 'fast-kill' debug button. If this is not
>> realistic, allow matlab to be queried by a command window to
>> determine running processes within it and allow them to be
>> individually terminated. Without bringing the entire thing down.
>
> I don't understand. Matlab always runs as a single process, unless
> you are using the multiprocesor toolbox (MPI extensions.) Newer
> Matlab support multiple -threads- within the same process; being
> able to examine individual -threads- is not, I think, particularily
> useful. I guess it could happen that one of the threads happened to
> get caught in a tight loop while the other calculation threads waited
> for it to catch up, but has that proven to be a substantial problem?


This is a problem on Windows, since Windows does not support signals.
Thus Matlab has to poll the keyboard for ^C and so on. When it's really
busy in a computation, frequently the keyboard is not very responsive.

Dan

Well, here is my wish - allow sparse(single(m)) or
equivalently single(sparse(m)). This is pretty important I
think. Both sparse and single are useful memory-saving for
large sparse matrices (which come up all the time in
applications). I was surprised to find out that this is not
supported in matlab.

And here's another thing I find really annoying - why can't
you define functions within matlab scripts? Why do they have
to be in separate files? Please allow this in the next version.

Paul

I can't be the only person that finds the formatting syntax
strings for printing to text files with the "fprintf"
command very hard to use?

Matlab does amazing things but some of the features make me
feel like I'm trapped in fortran class in the 80's.

Is a totally incomprehensible string of alphanumerics really
the only possible way to format text output?

In article <fm75bj$n1c$1@fred.mathworks.com>,
Tim <timbrown@nospamhere.nope> wrote:
>I can't be the only person that finds the formatting syntax
>strings for printing to text files with the "fprintf"
>command very hard to use?

>Matlab does amazing things but some of the features make me
>feel like I'm trapped in fortran class in the 80's.

>Is a totally incomprehensible string of alphanumerics really
>the only possible way to format text output?

Of course it isn't the only *possible* way -- but even legendarily
verbose old COBOL uses incomprehensible strings of alphanumerics
to format text output.

Let me turn the question around a bit: what would you -suggest-
as an alternative mechanism to indicate formatting of a floating
point number with a certain number of decimal leading decimal places,
a certain number of trailing decimal places, an option to zero-fill
if the number is smaller than is needed to fill the leading decimal
places, and an option to control left or right justification?
(And this is the simple case, as it doesn't include formatting of
an exponent.)
--
   So you found your solution
   What will be your last contribution?
   -- Supertramp (Fool's Overture)

Good question.

For starters, if "save" just exported the array the way it
was currently formatted that would help. I can set numeric
formatting in matlab, and open an array and look at it and
it looks fine - I go to save it and even small integers are
converted to scientific notation -- this almost seems like a
bug it is so inconvenient. Also, some of the other save
commands take formatting strings, why doesn't the "save"
command do it? Being able to choose between 8 and 16 digit
ASCII is hardly very helpful.

But on to your real question... for the matlab save commands
that do support formatting strings -- well I'm not sure. The
whole thing seems clunky and confusing in its current
implementation. When you go to save a figure, you can set
the output formatting in a menu, which works very well. When
you try to save text to a file you have a huge range of
choices ("save", "frpintf", "sprintf", "dlwrite",
"csvwrite", etc). That all do essentially the same thing but
with different requirements, different options for
formatting and slightly different results. It seems like
everytime I go to save data I have to use some new
collection of save terms that all require a slightly
different usage.

So I agree it is no easy thing to think of a simple way to
handle advanced formatting, but it really seems like the who
means by which data are saved need a real overhaul to
simplify them.

I'm hardly a beginner at matlab, most of the coding I did
for my phd I just finished was in matlab. When it only takes
me a few minutes to write code to load my data, do a bunch
of complex calculations and and output the results, but then
it takes me 3 hours to figure out to just write 3 text
headers and then a 3 column array to a text file, it seems
like this is an area for improvement. (actually I couldn't
even figure it out, I gave up and mailed this forum and
someone explained it to me).

Tim


roberson@ibd.nrc-cnrc.gc.ca (Walter Roberson) wrote in
message <fm82de$792$1@canopus.cc.umanitoba.ca>...

> Of course it isn't the only *possible* way -- but even
legendarily
> verbose old COBOL uses incomprehensible strings of
alphanumerics
> to format text output.
>
> Let me turn the question around a bit: what would you
-suggest-
> as an alternative mechanism to indicate formatting of a
floating
> point number with a certain number of decimal leading
decimal places,
> a certain number of trailing decimal places, an option to
zero-fill
> if the number is smaller than is needed to fill the
leading decimal
> places, and an option to control left or right justification?
> (And this is the simple case, as it doesn't include
formatting of
> an exponent.)
> --
> So you found your solution
> What will be your last contribution?
> -- Supertramp (Fool's Overture)

In article <fmgsj2$9j2$1@fred.mathworks.com>,
Tim <timbrown@nospamhere.nope> wrote:

>For starters, if "save" just exported the array the way it
>was currently formatted that would help. I can set numeric
>formatting in matlab, and open an array and look at it and
>it looks fine - I go to save it and even small integers are
>converted to scientific notation -- this almost seems like a
>bug it is so inconvenient. Also, some of the other save
>commands take formatting strings, why doesn't the "save"
>command do it? Being able to choose between 8 and 16 digit
>ASCII is hardly very helpful.

You should consider the primary purpose of 'save' to be to save
exact copies of the variables as binary. The -ascii option should
be considered to be just thrown on as a convenience in transfering
data to other programs.


>But on to your real question... for the matlab save commands
>that do support formatting strings -- well I'm not sure. The
>whole thing seems clunky and confusing in its current
>implementation. When you go to save a figure, you can set
>the output formatting in a menu, which works very well. When
>you try to save text to a file you have a huge range of
>choices ("save", "frpintf", "sprintf", "dlwrite",
>"csvwrite", etc). That all do essentially the same thing but
>with different requirements, different options for
>formatting and slightly different results. It seems like
>everytime I go to save data I have to use some new
>collection of save terms that all require a slightly
>different usage.

The only two commands that I would consider to be "save commands"
are save() and saveas(), as both of them work with exact binary
representations (of a workspace or of a figure respectively.)

fprintf() is derived from the C programming language and is
quite similar to what is available in C. fprintf() is just one step
above "raw" output (fwrite() outputs raw binary), and has oodles of
options because it is the primary flexible output formatter that
underlies pretty much everything else you list (except save()).

sprintf() creates a string but does not output it to a file, and
accepts the same format specifiers as fprintf() does. By itself it
is not used to save anything to a file, but it is fairly common
to use sprintf() to construct strings in pieces before final output.

dlmwrite() should essentially just be considered a layer over
fprintf(). csvwrite() *is* just a layer over dlmwrite(). Both of
them are just shortcuts for some common formatting tasks. Neither of
them are suitable for anything at all complex -- for example, neither
of them are at all good at writing out strings into spreadsheet cells.
Not even column headers. But they save you from having to know the
details of working with spreadsheet "sheets", which is a win if you
happen to need those (I never have though).

In short, if you need something more specific then the easy outputs
of csvwrite() or dlmwrite() or save -ascii, then you will need to
use fprintf(): it is the only "save command" you need to learn if
you are not doing a matlab binary workspace save().


>So I agree it is no easy thing to think of a simple way to
>handle advanced formatting, but it really seems like the who
>means by which data are saved need a real overhaul to
>simplify them.

Which would you eliminate: the "I don't need to know the details"
simple csvwrite() and dlmwrite(), or the "I need very fine-grained
control" of fprintf() ?


>For starters, if "save" just exported the array the way it
>was currently formatted that would help. I can set numeric
>formatting in matlab, and open an array and look at it and
>it looks fine - I go to save it and even small integers are
>converted to scientific notation -- this almost seems like a
>bug it is so inconvenient.

save -ascii doesn't even put in the variable names -- it isn't
a real "save" at all, just a simple data dump. It can't even
handle a simple structure:

>> baz.a = foo;baz.b = bar;
>> save /tmp/foo.txt -ascii baz
Warning: Attempt to write an unsupported data type to an ASCII file.
Variable 'baz' not written to file.

If that kind of format is all you want, then,

dlmwrite('/tmp/foo.txt', foo, 'delimiter', '', 'precision', '%9.5g')


>I'm hardly a beginner at matlab, most of the coding I did
>for my phd I just finished was in matlab. When it only takes
>me a few minutes to write code to load my data, do a bunch
>of complex calculations and and output the results, but then
>it takes me 3 hours to figure out to just write 3 text
>headers and then a 3 column array to a text file, it seems
>like this is an area for improvement. (actually I couldn't
>even figure it out, I gave up and mailed this forum and
>someone explained it to me).

The main trick to remember for fprintf is this: it writes a row at a
time, one value from each variable in turn, but extracts values down
columns.

The first value of the first variable will be matched with the first
format descriptor; the first value of the second variable will be
matched to the second format descriptor, and so on until you reach the
end of the list of variables, at which point it will start reusing the
variable list from the beginning, outputting the second value of the
first variable, then the second value from the second variable, and so
on. If at any point it runs out of format string, it will restart from
the beginning of the format string but that does *not* change where it
is in the variable list. The "second value from the first variable" is
thefirstvariable(2), which is indexing going down columns

The exception here is that for a %s or %q format, the entire row
of the corresponding string will be output, not just a single character
from the string.

Thus, the equivilent of the above dlmwrite is,

fprintf(fid, [repmat('%9.5g', 1, size(foo,2)), '\n'], foo .')


>> dlmwrite('/tmp/foo.txt', foo, 'delimiter', '', 'precision', '%9.5g')
>> !more /tmp/foo.txt
  0.81472 0.95751 0.42176 0.67874 0.27692 0.43874 0.70936 0.95974
  0.90579 0.96489 0.91574 0.75774 0.046171 0.38156 0.75469 0.34039
        7 0.15761 0.79221 0.74313 0.097132 0.76552 0.27603 0.58527
  0.91338 0.97059 0.95949 0.39223 0.82346 0.7952 0.6797 0.22381
  0.63236 0.95717 0.65574 0.65548 0.69483 0.18687 0.6551 0.75127
  0.09754 0.48538 0.035712 0.17119 0.3171 0.48976 0.16261 0.2551
   0.2785 0.80028 0.84913 0.70605 0.95022 0.44559 0.119 0.50596
  0.54688 0.14189 0.93399 0.031833 0.034446 0.64631 0.49836 0.69908

>> sprintf([repmat('%9.5g', 1, size(foo,2)), '\n'], foo .')

ans =

  0.81472 0.95751 0.42176 0.67874 0.27692 0.43874 0.70936 0.95974
  0.90579 0.96489 0.91574 0.75774 0.046171 0.38156 0.75469 0.34039
        7 0.15761 0.79221 0.74313 0.097132 0.76552 0.27603 0.58527
  0.91338 0.97059 0.95949 0.39223 0.82346 0.7952 0.6797 0.22381
  0.63236 0.95717 0.65574 0.65548 0.69483 0.18687 0.6551 0.75127
  0.09754 0.48538 0.035712 0.17119 0.3171 0.48976 0.16261 0.2551
   0.2785 0.80028 0.84913 0.70605 0.95022 0.44559 0.119 0.50596
  0.54688 0.14189 0.93399 0.031833 0.034446 0.64631 0.49836 0.69908

The repmat() is there to repeat the '%9.5g' format as many times
as there are columns in the array, so that the newline (\n) can be
placed properly. Note the .' there to flip the array around so
that when printf effectively re-flips it, it ends up coming out
in the order we would normally expect.

So, write one format item per column, and transpose your arrays.

--
  "All is vanity." -- Ecclesiastes

There are lots of interesting questions raised in this
thread, and lots of interesting comments. However, I think
it's time to close this thread and to discontinue any
discussion with it. I suggest starting a new thread, with
one topic per thread. The current thread is far too
unwieldy to read, and not very useful in the newsgroup
archive because the topics are all over the map. In any
case, R2007b has been released for quite some time now.

In the future, I would suggest a seperate thread for each
wishlist item, perhaps something of the form:

Subject: wishlist for R200whatever: (specific topic here)

and replace "specific topic here" with whatever you like,
for all threads with "wishlist ..."

So, please, let this be the last post on this thread.
Start a new one instead.

Hi
I am shan from university of Liverpool.
I have problem in getting logic equation for the numbers
bellow for matlab programing
r1 =1 1 3 5 7 9
r2 =3 1 1 5 7 9
r3 =5 3 1 1 3 5
r4 =7 5 3 1 1 3
r5 =9 7 5 3 1 1.
kindly help me if you have any idea or logic for this
problem.
thank you
with regards
Shanmukha

Hi
I am shan from university of Liverpool.
I have problem in getting logic equation for the numbers
bellow for matlab
r1 =1 1 3 5 7 9
r2 =3 1 1 5 7 9
r3 =5 3 1 1 3 5
r4 =7 5 3 1 1 3
r5 =9 7 5 3 1 1.
kindly help me if you have any idea or logic for this
problem.
thank you
with regards
Shanmukha

Not sure what you mean, but I guess this equation makes sense to use:
abs([9:-2:-9])

Best wishes, Marcus
Tomlab Optimization Inc.
http://tomopt.com/tomlab/
http://tomdyn.com/

"Shanmukha Moturu" <Shanmukharao.Moturu@liv.ac.uk> wrote in message
<g3dru9$odo$1@fred.mathworks.com>...
> Hi
> I am shan from university of Liverpool.
> I have problem in getting logic equation for the numbers
> bellow for matlab
> r1 =1 1 3 5 7 9
> r2 =3 1 1 5 7 9
> r3 =5 3 1 1 3 5
> r4 =7 5 3 1 1 3
> r5 =9 7 5 3 1 1.
> kindly help me if you have any idea or logic for this
> problem.
> thank you
> with regards
> Shanmukha

DON'T post a question like this in an existing
thread!!! Especially not if your question has a
different topic.

DO post it in a new thread, with a proper
subject, that describes your problem!!!

toeplitz(1:2:9,[1,1:2:9])

John

In article <g3dsd9$qk$1@fred.mathworks.com>,
John D'Errico <woodchips@rochester.rr.com> wrote:
>"Shanmukha Moturu" <Shanmukharao.Moturu@liv.ac.uk> wrote in message
><g3dru9$odo$1@fred.mathworks.com>...

>> I have problem in getting logic equation for the numbers
>> bellow for matlab
>> r1 =1 1 3 5 7 9
>> r2 =3 1 1 5 7 9
>> r3 =5 3 1 1 3 5
>> r4 =7 5 3 1 1 3
>> r5 =9 7 5 3 1 1.

>toeplitz(1:2:9,[1,1:2:9])

Looks to me like that's probably what the original poster actually
wanted, but it doesn't produce the same answer as what the original
poster actually asked for.

>> toeplitz(1:2:9,[1,1:2:9])

ans =

     1 1 3 5 7 9
     3 1 1 3 5 7
     5 3 1 1 3 5
     7 5 3 1 1 3
     9 7 5 3 1 1


This differs from the original solution in the 4th, 5th, and 6th
places of the second row.


Unfortunately the matlab documentation for toeplitz is very sparse
as to what it actually -does-, so I'm lost either way ;-(
--
  "Ignorance has been our king... he sits unchallenged on the throne of
  Man. His dynasty is age-old. His right to rule is now considered
  legitimate. Past sages have affirmed it. They did nothing to unseat
  him." -- Walter M Miller, Jr

I am trying to solve problem as explain below
the equation is like this

Vij = (1/ 4επL)* ∑_(m=1)to(m=n){(Q_m )/r}--> (1)
where L is lenth.
r is distance from that point.
where Vij = 1 at P1; 4επL=6.672e-12.
 
-*----@-----*-----@-----*-----@-----*-----@-----*----@-----*

Let "*" be Q1.....Q6 and "@" be P1....p5
the distance between every Q and P is 1/2m let m=6

the above equation (1) can be writen as below


V1 At P1 = (1/ 4επL)[Q1/(1/12) + Q2/(1/12) + Q3/(3/12) + Q4/
(5/12)+ Q5/(7/12)+ Q6/(9/12)
V2 at p2 = (1/ 4επL)[Q1/(3/12) + Q2/(1/12) + Q3/(1/12) + Q4/
(3/12)+ Q5/(5/12)+ Q6/(7/12)
V3 at p3= (1/ 4επL)[Q1/(5/12) + Q2/(3/12) + Q3/(1/12) + Q4/
(1/12)+ Q5/(3/12)+ Q6/(5/12)
V4 at p4 = (1/ 4επL)[Q1/(7/12) + Q2/(5/12) + Q3/(3/12) + Q4/
(1/12)+ Q5/(1/12)+ Q6/(3/12)
V5 at p5= (1/ 4επL)[Q1/(9/12) + Q2/(7/12) + Q3/(5/12) + Q4/
(3/12)+ Q5/(1/12)+ Q6/(1/12)

From the above equations it is clear that
denominator is changing as fellows
1/12, 1/12,3/12,5/12,7/12,9/12 for V1
3/12,1/12,1/12,3/12,5/12,7/12 for V2
5/12,3/12,1/12,1/12,3/12,5/12 for V3
7/12,5/12,3/12,1/12,1/12,3/12 For V4
9/12,7/12,5/12,3/12,1/12,1/12 for V5

i need logic for this values so that i can put for n values
and get similar equation


Kindly help me
Thank you
with regards
Shanmukha

"Shanmukha Moturu" <Shanmukharao.Moturu@liv.ac.uk> wrote in message
<g3e235$668$1@fred.mathworks.com>...
> I am trying to solve problem as explain below

(snip)

> From the above equations it is clear that
> denominator is changing as fellows
> 1/12, 1/12,3/12,5/12,7/12,9/12 for V1
> 3/12,1/12,1/12,3/12,5/12,7/12 for V2
> 5/12,3/12,1/12,1/12,3/12,5/12 for V3
> 7/12,5/12,3/12,1/12,1/12,3/12 For V4
> 9/12,7/12,5/12,3/12,1/12,1/12 for V5
>
> i need logic for this values so that i can put for n values
> and get similar equation

(Sigh.) Did you bother looking at the solution
I posted?

Try it. The extension to larger matrices is trivial.

John

i got it, its working thank you

In matlab can we get out put as
Q1+Q2+Q3+Q4+Q5+...+Qn

if so how can we do it.

In article <g3e4r9$6r2$1@fred.mathworks.com>,
Shanmukha Moturu <Shanmukharao.Moturu@liv.ac.uk> wrote:

>In matlab can we get out put as
>Q1+Q2+Q3+Q4+Q5+...+Qn
>if so how can we do it.

http://matlabwiki.mathworks.com/MATLAB_FAQ#How_can_I_create_variables_A1.2C_A2.2C....2CA10_in_a_loop.3F

Note in particular the first sentance of the answer, as it contains
critical information.

--
  "After all, what problems has intellectualism ever solved?"
                                              -- Robert Gilman