Thread Subject: deleting 0s from an array

  Is there any easy way to delete all zero values from an
array (i.e., other than going through each element and
rebuilding)?

  This is a 1-D array whose values are used in a
histogram. Sometimes the bins are too wide in value to get
much information, so I wanted to narrow down by looking at
individual bins with more granularity. I can find the
maximum value range for the bins, then the next highest,
etc. and generate histograms.

   Or is there a better idea?

Thanks

James:
<SNIP no-zeros, please, evergreen...

> Is there any easy way to delete all zero values from an
array...

one of the copious solutions

% the data
     d=[0,0,0,1,0,0,0,2,0,0,0,3]
% ...without zeros
     d(d==0)=[]

us

In article <fa08s1$r4b$1@fred.mathworks.com>, us <us@neurol.unizh.ch> wrote:
>James:
><SNIP no-zeros, please, evergreen...

>> Is there any easy way to delete all zero values from an
>array...

>one of the copious solutions

>% the data
> d=[0,0,0,1,0,0,0,2,0,0,0,3]
>% ...without zeros
> d(d==0)=[]

I can "golf" that!

d(~d)=[] %provided d does not contain a nan
--
  If you lie to the compiler, it will get its revenge. -- Henry Spencer

In article <fa3kot$4mp$1@fred.mathworks.com>, us <us@neurol.unizh.ch> wrote:
>Richard Brown:
><SNIP pseudo-golf...

>> > d(~d)=[] %provided d does not contain a nan

>> I can tie:
>> d=d(~~d)

>again, and as <walter roberson> pointed out: as long as <d>
>is finite... now, to test this would (probably) cost you an
>extra few chars...

d(~d)=[] and d=d(~~d) work for inf and -inf but neither works for NaN.

--
   Okay, buzzwords only. Two syllables, tops. -- Laurie Anderson

Walter Roberson:
<SNIP stumbled over <us>'s <finite>...

> d(~d)=[] and d=d(~~d) work for inf and -inf but neither
works for NaN...

correct! hasty use of the word <finite>...

us

roberson@ibd.nrc-cnrc.gc.ca (Walter Roberson) wrote in m
> >> I can tie:
> >> d=d(~~d)

You did better than tie. The above statement is twice as
fast as the other two alternatives. I tried d of size 1e6
with 10% nonzeros.

All of these solutions, however, destroy the location
information of the nonzeros in the array. Just as fast as
d=d(~~d), and cleaner in my mind, is the simple

d=sparse(d)

Then if you want the location information, use find(d). You
can't use find(d) with the other solutions.

Do function calls count as one "character" in MATLAB golf?
If so, I win :-).

Note, however, that if you want to subreference this array,
make sure that d is a row vector. Otherwise d(i) where i is
a scalar, will be slow, if d is a column vector.

"Walter Roberson" <roberson@ibd.nrc-cnrc.gc.ca> wrote in message
news:fa3n8c$1an$1@canopus.cc.umanitoba.ca...
> In article <fa3kot$4mp$1@fred.mathworks.com>, us <us@neurol.unizh.ch>
> wrote:
>>Richard Brown:
>><SNIP pseudo-golf...
>
>>> > d(~d)=[] %provided d does not contain a nan
>
>>> I can tie:
>>> d=d(~~d)
>
>>again, and as <walter roberson> pointed out: as long as <d>
>>is finite... now, to test this would (probably) cost you an
>>extra few chars...


Add one more character and you get a version that works for inf, -inf, NaN,
and finite numbers.

> d(~d)=[] and d=d(~~d) work for inf and -inf but neither works for NaN.
>
> --
> Okay, buzzwords only. Two syllables, tops. -- Laurie Anderson

% This line doesn't count in the score; it generates sample data to test the
solution
d = [0 1 2 -5 0 324.65 Inf 0 -Inf NaN 0];

% Hole in 9
d=d(d~=0)

--
Steve Lord
slord@mathworks.com

"Tim Davis" <davis@cise.ufl.edu> wrote in message
> Then if you want the location information, use find(d). You
> can't use find(d) with the other solutions.
>

Umm. I guess my brain is too sparse. I should have
mentioned that to use find(d) you would do

[i j x] = find(d)

then x is the vector you're looking for, with zeros deleted,
and i is the list of d(i)'s that are nonzero (assuming d is
a column vector).

The above works if d is full or sparse; you don't have to do

d = sparse (d)

first. I think the above is really the simplest way to do
it, personally. But then I'm used to seeing "find"
statements flying around in m-files.