robert bristow-johnson <rbj@audioimagination.com> wrote:
> On 4/8/15 8:30 AM, Randy Yates wrote:

(snip)
>> MATLAB = MATRIX Laboratory. So the same reason mathematicians
>> number their elements a_11, a_12, ..., a_21, a_22, etc. Blame
>> the mathematician who started THAT convention!
 
> yeah, but they're not always matrices.  

It took Fortran 20 years before it allowed one to specify the origin.

I haven't used Matlab so much lately, but doesn't the zero subscript
have another meaning? Or negative subscripts? I might be thinking of
R, which I have used more recently.

> in fact, the largest-selling 
> MATLAB toolbox is the Signal Processing Toolbox.  i might venture a 
> guess that the largest portion of MATLAB users are EE's doing signal 
> processing.  for EE's doing signal processing, i'm sure the vast 
> majority of the time, they are "mere" *arrays* not matrices where we are 
> doing matrix arithmetic.  probably most other users of MATLAB use it for 
> general math use and matrix math is used, maybe, 30% of the time.

(snip)

-- glen

On Thu, 09 Apr 2015 00:41:47 -0400, Max <Max@sorrynope.com> wrote:

>"By popular demand, Matlab array indexes will now start from negative
>infinity.  A[1] will refer to the element at A[-infinity + 1],  etc."
>
>That will solve all those problems with Fourier transforms, as long as
>your machine has a lot of memory.

Finally I can implement those ideal lowpass filters.

On 4/9/15 12:41 AM, Max wrote:
> On Wed, 08 Apr 2015 09:52:02 -0400, robert bristow-johnson
> <rbj@audioimagination.com>  wrote:
>
>> AND, with that, we might have origins that are negative so that we can
>> cleanly represent negative frequency in the Fourier transform and
>> negative time for non-causal impulse responses.
>>
>> it's without excuse.
>
> There's my headline for Apr 01, 2016:
> "By popular demand, Matlab array indexes will now start from negative
> infinity.  A[1] will refer to the element at A[-infinity + 1],  etc."
>
> That will solve all those problems with Fourier transforms, as long as
> your machine has a lot of memory.
>
> (Actually, your idea about a selectable starting index is great, rbj)

it was the original idea from 2 decades ago.

 
https://groups.google.com/forum/#!original/comp.soft-sys.matlab/2J15oSvPTGQ/JpPiJVUgItwJ 


the thread and context is

 
https://groups.google.com/forum/#!msg/comp.soft-sys.matlab/2J15oSvPTGQ/JpPiJVUgItwJ

at that time i was able to get Cleve to notice and respond.  not so anymore.

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

On Wed, 08 Apr 2015 09:52:02 -0400, robert bristow-johnson
<rbj@audioimagination.com> wrote:

>AND, with that, we might have origins that are negative so that we can 
>cleanly represent negative frequency in the Fourier transform and 
>negative time for non-causal impulse responses.
>
>it's without excuse.

There's my headline for Apr 01, 2016:
"By popular demand, Matlab array indexes will now start from negative
infinity.  A[1] will refer to the element at A[-infinity + 1],  etc."

That will solve all those problems with Fourier transforms, as long as
your machine has a lot of memory.

(Actually, your idea about a selectable starting index is great, rbj)

On 4/8/15 8:30 AM, Randy Yates wrote:
> Max<Max@sorrynope.com>  writes:
>
>> On Tue, 07 Apr 2015 05:35:50 -0700, Rick Lyons
>> <R.Lyons@_BOGUS_ieee.org>  wrote:
>>
>>>>
>>>> You might want to go back and look at the date of the original post.
>>>>
>>>> ;)
>>>
>>> Hi Eric,
>>>
>>>    To quote Homer Simpson, "DOH."
>>
>> Oops, sorry. I should have made the date of that post more obvious.
>> I really do wish that they hadn't started with 1 though. Gotta wonder
>> why they thought that was a good idea.
>
> MATLAB = MATRIX Laboratory. So the same reason mathematicians
> number their elements a_11, a_12, ..., a_21, a_22, etc. Blame
> the mathematician who started THAT convention!

yeah, but they're not always matrices.  in fact, the largest-selling 
MATLAB toolbox is the Signal Processing Toolbox.  i might venture a 
guess that the largest portion of MATLAB users are EE's doing signal 
processing.  for EE's doing signal processing, i'm sure the vast 
majority of the time, they are "mere" *arrays* not matrices where we are 
doing matrix arithmetic.  probably most other users of MATLAB use it for 
general math use and matrix math is used, maybe, 30% of the time.

and those matrix elements *could* be numbered a_00, a_01, ..., a_10, 
a_11, etc.  wouldn't hurt anything.

AND, (this is the reason the refusal to fix is inexcusable), what MATLAB 
needs to be is have the array origin for each dimension USER SPECIFIED 
with the default value being 1.  there would need to be two new 
functions, comparable to size() and reshape() that i would call 
"origin()" and "reorigin()" (or maybe "base()" and "rebase()")  that 
makes it perfectly backward compatible and no one's existing code breaks.

AND, with that, we might have origins that are negative so that we can 
cleanly represent negative frequency in the Fourier transform and 
negative time for non-causal impulse responses.

it's without excuse.

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

Max <Max@sorrynope.com> writes:

> On Tue, 07 Apr 2015 05:35:50 -0700, Rick Lyons
> <R.Lyons@_BOGUS_ieee.org> wrote:
>
>>>
>>>You might want to go back and look at the date of the original post.
>>>
>>>;)
>>
>>Hi Eric,
>>
>>   To quote Homer Simpson, "DOH."
>
> Oops, sorry. I should have made the date of that post more obvious. 
> I really do wish that they hadn't started with 1 though. Gotta wonder
> why they thought that was a good idea.

MATLAB = MATRIX Laboratory. So the same reason mathematicians 
number their elements a_11, a_12, ..., a_21, a_22, etc. Blame
the mathematician who started THAT convention!
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

On Tue, 07 Apr 2015 05:35:50 -0700, Rick Lyons
<R.Lyons@_BOGUS_ieee.org> wrote:

>>
>>You might want to go back and look at the date of the original post.
>>
>>;)
>
>Hi Eric,
>
>   To quote Homer Simpson, "DOH."

Oops, sorry. I should have made the date of that post more obvious. 
I really do wish that they hadn't started with 1 though. Gotta wonder
why they thought that was a good idea.

On Tue, 07 Apr 2015 00:38:27 GMT, eric.jacobsen@ieee.org (Eric
Jacobsen) wrote:

>On Mon, 06 Apr 2015 17:18:16 -0700, Rick Lyons
><R.Lyons@_BOGUS_ieee.org> wrote:
>
>>On Wed, 01 Apr 2015 06:24:47 -0400, Max <Max@sorrynope.com> wrote:
>>
>>>I was glad to hear that Mathworks has finally corrected their array
>>>indexing scheme to start with 0 instead of 1, so they're now
>>>compatible with all the other languages out in the real world.
>>>
>>>Apparently they knew all along that array indexes actually start at
>>>zero, but a bug in the initial release of the code caused the 'off by
>>>one' problem.  An admission of such a glaring math error would have 
>>>been disastrous, so they claimed that they meant to do it. 
>>>
>>>Mark this day on your calendar as the day when corresponding source
>>>code needs to be corrected.  And perhaps the day that I get hunted
>>>down by some humorless Mathworks salesman :-)
>>
>>Hello Max,
>>   Thanks for the good news.
>>
>>I wonder, how does this affect users of older 
>>versions of Matlab?  That is, is there any 
>>way older versions of Matlab can take advantage 
>>of the updated zero-indexing capabilities? 
>>
>>Thanks,
>>[-Rick-]
>
>You might want to go back and look at the date of the original post.
>
>;)

Hi Eric,

   To quote Homer Simpson, "DOH."

[-Rick-]

>> On Wed, 01 Apr 2015 06:24:47 -0400, Max<Max@sorrynope.com>  wrote:
>
>>> I was glad to hear that Mathworks has finally corrected their array
>>> indexing scheme to start with 0 instead of 1, so they're now
>>> compatible with all the other languages out in the real world.
>
> (snip)
>
>>> Mark this day on your calendar as the day when corresponding source
>>> code needs to be corrected.  And perhaps the day that I get hunted
>>> down by some humorless Mathworks salesman :-)
>

Rick Lyons<R.Lyons@_bogus_ieee.org>  wrote:
>> I wonder, how does this affect users of older
>> versions of Matlab?  That is, is there any
>> way older versions of Matlab can take advantage
>> of the updated zero-indexing capabilities?

On 4/6/15 11:40 PM, glen herrmannsfeldt wrote:
>
> Only on April 0th.

On 4/6/15 8:38 PM, Eric Jacobsen wrote:
>
> You might want to go back and look at the date of the original post.

but still, oh how i wish Cleve Moler would notice.

even 15 or 20 years later, it isn't too late to correct a mistake.


-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

Rick Lyons <R.Lyons@_bogus_ieee.org> wrote:
> On Wed, 01 Apr 2015 06:24:47 -0400, Max <Max@sorrynope.com> wrote:
 
>>I was glad to hear that Mathworks has finally corrected their array
>>indexing scheme to start with 0 instead of 1, so they're now
>>compatible with all the other languages out in the real world.

(snip)

>>Mark this day on your calendar as the day when corresponding source
>>code needs to be corrected.  And perhaps the day that I get hunted
>>down by some humorless Mathworks salesman :-)

(snip)
> I wonder, how does this affect users of older 
> versions of Matlab?  That is, is there any 
> way older versions of Matlab can take advantage 
> of the updated zero-indexing capabilities? 

Only on April 0th.

-- glen