Can I assign a value from textscan to anything other than a cell array?

I want to assign the value from a textscan directly to an simple variable, can I do this? Example:

str = '223 87 87';
[Nx, Ny, Nz] = textscan(str, '%f %f %f');

I know this doesn't work, because textscan returns a cell array, rather than a list of numbers. So I could do something like

Narray = textscan(str, '%f %f %f');
Nx = Narray{1}; Ny = Narray{2}; Nz = Narray{3}; clear Narray;

But then it takes two lines and an extra cell array that I never wanted to exist in the first place...it seems inefficient and less readable. Is there a way that I can have the output from textscan assigned directly to the variables I want, or is an intermediate cell array always necessary? Thanks!

[per isakson]

Short answer: NO!

textread did that

[A,B,C, ...] = textread('FILENAME','FORMAT')

However, "Note: textread is not recommended. Use textscan instead."

Answer 1

[dpb]

The insistence of textscan to return only cell array is annoying often, indeed. Only way around what you're doing I'm aware of is to not generate the three separate variables but use an array instead--then you can at least eliminate the explicit temporary

>> Nxyz=cell2mat(textscan(str,'%f'))
  Nxy =
     223
      87
      87
>>

If you're not reading a string but a file, I often still revert to textread for the same reason; in that case you can have the multiple outputs as well.

ADDENDUM

It seems one should be able to deal the outputs, but I couldn't manage to force textscan to generate a comma-separated list without the explicit temporary so ended up with the array propagated over the three variables instead of the three elements distributed. But, it might be some better than the explicit evaluation...

>> C=textscan(str,'%f%f%f')
C = 
  [223]    [87]    [87]
>> [Nx Ny Nz]=deal(C{1,:})
Nx =
 223
Ny =
  87
Nz =
  87
>>

The difficulty here is that one must use a counted format string to force the creation of the cell array of N elements instead of the single cell of an N-sized array. And, since there's no syntax for array or cell array access after a function, the explicit temporary must be available for the dereferencing expression since

>> [Nx Ny Nz]=deal(textscan(str,'%f%f%f'))
Nx = 
  [223]    [87]    [87]
Ny = 
  [223]    [87]    [87]
Nz = 
  [223]    [87]    [87]
>>

as previously noted which is same result as is

>> [Nx Ny Nz]=deal(C)

The latter seems an incorrect choice for the implementation of deal to me but is as documented.

ADDENDUM

"The difficulty here is that one must use a counted format string ..."

Actually, it just dawned on me in thinking over Walter's most-excellent addition of dealcell that the facility of textscan to automagically parse the input record and return the proper number of elements if the formatSpec field is the empty string comes into play here...I think the best end solution is then

[Nx Ny Nz]=dealcell(textscan(str,''));

[Emily]

Thank you!

[Walter Roberson]

You need the temporary to be able to do cell expansion.

[dpb]

@Walter -- I had thought I'd read/heard that Octave had addressing facilities after a function reference but a look-see at the doc's didn't reveal it if so...I'm sure it'd be a nightmare to implement but have often wished could make such references to save the temporaries hanging around...

Answer 2

[Image Analyst]

I agree it's annoying that textscan() is not very flexible. You can use cell2mat() to put it all into one array:

N = cell2mat(textscan(str, '%f %f %f'))

You can then split N up into 3 separate individual variables if you want, or use it conveniently as an array.

[Emily]

Hmmm...it would seem like I could use this to do

 [Nx, Ny, Nz] = cell2mat(textscan(str, '%f %f %f'))

but it seems it doesn't work that way, as matlab says cell2mat has too many output arguments. I wonder why I can't make that assignment. Thanks for your answer.

[dpb]

Because cell2mat only converts to a single array and you've asked for three outputs. deal is the closest but it expands the RHS to all the LHS arguments for anything excepting a comma-separated list which you can only generate starting from textscan as the explicit dereferencing expression of the cell array illustrated above.

It seems to me the use for propagating a full array to multiple copies as happens is likely far less useful than the more-true-to-the-name expansion of scattering the RHS members to the LHS members on a one-to-one association but that isn't what TMW chose to do.

[Walter Roberson]

dealcell = @(C) deal(C{:});

[Nx, Ny, Nz] = dealcell(textscan(str, '%f %f %f'));

[dpb]

That's a good solution, Walter..."have your cake and eat it, too!" :)

Actually, I overlooked that this is another case where the empty-string for the formatSpec argument is a major benefit in that you then don't have to count the input record to build the format string whereas the single specific format causes the array duplication even with anonymous function wrapper.

[Emily]

Thanks for your help Walter and dpb. Would you be willing to explain this answer to me a bit? I don't understand what the line

dealcell = @(C) deal(C{:});

does. And I assume it is not the same C as in the addendum, but C is somehow a new variable, that doesn't exist before this command?

Finally, if this is a better answer, is it possible for me to unaccept the first answer so this can be recorded as the correct answer?

[dpb]

The line is defining an "anonymous function" assigned to the "function handle" dealcell. C is the dummy argument to the function and is simply a placeholder for the actual argument that will be used when the function is invoked.

Read up on anonymous functions and function handles to get the full scoop but essentially it creates a named function programmatically inline that can be used just as if there were an m-file of the same name containing the same code. Exception being that it only exists for the scope and duration of the program unit in which it is defined (or to which the handle is passed as an argument but that's more than need to get into here).

I already edit'ed the ACCEPT'ed Answer in the ADDENDUM to incorporate what I thought the best solution would be using the anonymous function so it's already been accepted. :) Although I didn't replicate the anonymous function definition therein; I simply referred to it as Walter's baby... :) But, you need both it and the subsequent line that uses it to get the end result; the function definition by itself does nothing useful (well, it creates the function handle, but nothing happens until it is subsequently invoked).