Thread Subject: Extracting variables from a structure?

Hi, I'm new to structures but was forces to use one as I wanted to use a string as a variable name. What I am left with is a structure called "SETS" like so..

SETS =

    [1x1 struct] [1x1 struct] [1x1 struct]

>> SETS{1}

ans =

    nset_PickedSet2: [1 8 1]

>> SETS{2}

ans =

    nset_PickedSet5: [7 8]

>> SETS{3}

ans =

    nset_PickedSet6: [3 4]


What I want to have is variables...

nset_PickedSet2 = [1 8 1]
nset_PickedSet5 = [7 8]
nset_PickedSet6 = [3 4]

rather than having them in the structure.

Is this possible?

Any help is greatly appreciated.

Thanks.

Mark.

"Mark Hard" <GerryTheLeper@hotmail.com> wrote in message <hkrcn1$v0$1@fred.mathworks.com>...
> Hi, I'm new to structures but was forces to use one as I wanted to use a string as a variable name. What I am left with is a structure called "SETS" like so..
> SETS =
> [1x1 struct] [1x1 struct] [1x1 struct]
> >> SETS{1}
> ans =
> nset_PickedSet2: [1 8 1]
> >> SETS{2}
> ans =
> nset_PickedSet5: [7 8]
> >> SETS{3}
> ans =
> nset_PickedSet6: [3 4]
> What I want to have is variables...
> nset_PickedSet2 = [1 8 1]
> nset_PickedSet5 = [7 8]
> nset_PickedSet6 = [3 4]
> rather than having them in the structure.
> Is this possible?
> Any help is greatly appreciated.
> Thanks.
> Mark.

clearly NOT advisable...

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

one of the solutions

% the data
     clear s; % <- save old stuff(!)...
     s{1}.a=1:3;
     s{2}.bb=1:4;
     s{3}.ccc=1:5;
% the engine
% - convert into CELL array of DOUBLEs
     c=cellfun(@(x) struct2array(x),s,'uni',false).'
%{
% c =
     [1x3 double]
     [1x4 double]
     [1x5 double]
%}
% now, access data by indexing, eg,
     c{1}
% ans = 1 2 3

us

"us " <us@neurol.unizh.ch> wrote in message <hkrgob$ej9$1@fred.mathworks.com>...
> "Mark Hard" <GerryTheLeper@hotmail.com> wrote in message <hkrcn1$v0$1@fred.mathworks.com>...
> > Hi, I'm new to structures but was forces to use one as I wanted to use a string as a variable name. What I am left with is a structure called "SETS" like so..
> > SETS =
> > [1x1 struct] [1x1 struct] [1x1 struct]
> > >> SETS{1}
> > ans =
> > nset_PickedSet2: [1 8 1]
> > >> SETS{2}
> > ans =
> > nset_PickedSet5: [7 8]
> > >> SETS{3}
> > ans =
> > nset_PickedSet6: [3 4]
> > What I want to have is variables...
> > nset_PickedSet2 = [1 8 1]
> > nset_PickedSet5 = [7 8]
> > nset_PickedSet6 = [3 4]
> > rather than having them in the structure.
> > Is this possible?
> > Any help is greatly appreciated.
> > Thanks.
> > Mark.
>
> clearly NOT advisable...
>
> http://matlabwiki.mathworks.com/MATLAB_FAQ#How_can_I_create_variables_A1.2C_A2.2C....2CA10_in_a_loop.3F
>
> one of the solutions
>
> % the data
> clear s; % <- save old stuff(!)...
> s{1}.a=1:3;
> s{2}.bb=1:4;
> s{3}.ccc=1:5;
> % the engine
> % - convert into CELL array of DOUBLEs
> c=cellfun(@(x) struct2array(x),s,'uni',false).'
> %{
> % c =
> [1x3 double]
> [1x4 double]
> [1x5 double]
> %}
> % now, access data by indexing, eg,
> c{1}
> % ans = 1 2 3
>
> us

Thanks for your reply. Unfortunately I have to have the variable names as they are. Te variables aren't being made by a loop they are being extracted from a text file so is it still not advisable to extract them from the structure as there will be at most 5 or 6 of them?

Thanks again.

Mark.

"Mark Hard"
> Thanks for your reply. Unfortunately I have to have the variable names as they are. Te variables aren't being made by a loop they are being extracted from a text file so is it still not advisable to extract them from the structure as there will be at most 5 or 6 of them?
>
> Thanks again.
>
> Mark.

one of the solutions
- if you insist on the mnemonics...

% the data
     clear s; % <- save old stuff(!)...
     s{1}.a=1:3;
     s{2}.bb1=1:4;
     s{2}.bb2=1:10;
     s{3}.ccc=1:5;
% the engine
     c=cellfun(@(x) [fieldnames(x),struct2cell(x)],s,'uni',false);
     c=cat(1,c{:}).';
     c=struct(c{:});
% the result
     disp(c);
%{
       a: [1 2 3]
     bb1: [1 2 3 4]
     bb2: [1 2 3 4 5 6 7 8 9 10]
     ccc: [1 2 3 4 5]
%}

us

"us " <us@neurol.unizh.ch> wrote in message <hkrk53$hi7$1@fred.mathworks.com>...
> "Mark Hard"
> > Thanks for your reply. Unfortunately I have to have the variable names as they are. Te variables aren't being made by a loop they are being extracted from a text file so is it still not advisable to extract them from the structure as there will be at most 5 or 6 of them?
> >
> > Thanks again.
> >
> > Mark.
>
> one of the solutions
> - if you insist on the mnemonics...
>
> % the data
> clear s; % <- save old stuff(!)...
> s{1}.a=1:3;
> s{2}.bb1=1:4;
> s{2}.bb2=1:10;
> s{3}.ccc=1:5;
> % the engine
> c=cellfun(@(x) [fieldnames(x),struct2cell(x)],s,'uni',false);
> c=cat(1,c{:}).';
> c=struct(c{:});
> % the result
> disp(c);
> %{
> a: [1 2 3]
> bb1: [1 2 3 4]
> bb2: [1 2 3 4 5 6 7 8 9 10]
> ccc: [1 2 3 4 5]
> %}
>
> us

Thanks again for your reply. In your example above what I want is to have a, bb1, bb2 and cc to show up a variables in the workspace where,

a = [1 2 3]
bb1 = [1 2 3 4]
bb2 = [1 2 3 4 5 6 7 8 9 10]
ccc: = [1 2 3 4 5]

rather than have them as part of a structure 'c' so that if I type say bb1 into the command window I'll get back..

bb1 =

     1 2 3 4

I hope I'm making sense.

Thanks again

Mark.

"Mark Hard" <GerryTheLeper@hotmail.com> wrote in message <hkrhtr$rn4$1@fred.mathworks.com>...

> Thanks for your reply. Unfortunately I have to have the variable names as they are. Te variables aren't being made by a loop they are being extracted from a text file so is it still not advisable to extract them from the structure as there will be at most 5 or 6 of them?
=======================

If there will only be 5 or 6 of them, why not just extract them manually

nset_PickedSet2=SET{1}.nset_PickedSet2;
nset_PickedSet5=SET{2}.nset_PickedSet5;
etc...

"Matt J " <mattjacREMOVE@THISieee.spam> wrote in message <hkrr64$e1r$1@fred.mathworks.com>...
> "Mark Hard" <GerryTheLeper@hotmail.com> wrote in message <hkrhtr$rn4$1@fred.mathworks.com>...
>
> > Thanks for your reply. Unfortunately I have to have the variable names as they are. Te variables aren't being made by a loop they are being extracted from a text file so is it still not advisable to extract them from the structure as there will be at most 5 or 6 of them?
> =======================
>
> If there will only be 5 or 6 of them, why not just extract them manually
>
> nset_PickedSet2=SET{1}.nset_PickedSet2;
> nset_PickedSet5=SET{2}.nset_PickedSet5;
> etc...

Well the number of the set can change everytime depending on the automatically generated text file. so I won't know the names I need to variables to be to input it into the code.

"Mark Hard" <GerryTheLeper@hotmail.com> wrote in message <hkrnhq$neh$1@fred.mathworks.com>...

> > % the engine
> > c=cellfun(@(x) [fieldnames(x),struct2cell(x)],s,'uni',false);
> > c=cat(1,c{:}).';
> > c=struct(c{:});
> > % the result
> > disp(c);
> > %{
> > a: [1 2 3]
> > bb1: [1 2 3 4]
> > bb2: [1 2 3 4 5 6 7 8 9 10]
> > ccc: [1 2 3 4 5]
> > %}
> >
> > us
>
> Thanks again for your reply. In your example above what I want is to have a, bb1, bb2 and cc to show up a variables in the workspace where,
>
> a = [1 2 3]
> bb1 = [1 2 3 4]
> bb2 = [1 2 3 4 5 6 7 8 9 10]
> ccc: = [1 2 3 4 5]
>
> rather than have them as part of a structure 'c' so that if I type say bb1 into the command window I'll get back..
>
> bb1 =
>
> 1 2 3 4
>
========================

Once you have the structure c generated by us' engine, you can use my structvars()
tool on the FEX,

http://www.mathworks.com/matlabcentral/fileexchange/26216-structure-fields-to-variables

in conjunction with eval() to transfer the fields of c to separate variables:

>> c

c =

      a: [1 2 3]
    bb1: [1 2 3 4]
    bb2: [1 2 3 4 5 6 7 8 9 10]
    ccc: [1 2 3 4 5]

>> eval(structvars(c).')
>> a,bb1,bb2,ccc

a =

     1 2 3


bb1 =

     1 2 3 4


bb2 =

     1 2 3 4 5 6 7 8 9 10


ccc =

     1 2 3 4 5

I cannot emphasize enough, however, what a bad practice this is. It is a very bad idea to introduce variables into a MATLAB workspace without an explicit assignment statement. You should extricate yourself from the need for this at the earliest opportunity.

Mark Hard wrote:
> "Matt J " <mattjacREMOVE@THISieee.spam> wrote in message

>> If there will only be 5 or 6 of them, why not just extract them manually

>> nset_PickedSet2=SET{1}.nset_PickedSet2;
>> nset_PickedSet5=SET{2}.nset_PickedSet5;
>> etc...

> Well the number of the set can change everytime depending on the
> automatically generated text file. so I won't know the names I need to
> variables to be to input it into the code.

If you do manage to create them as individual variables and you don't
know the names ahead of time, how is your code going to know which
variables to manipulate? Are you going to use a who bunch of
exist('PickedSet2','var') calls, or are you going to use who() and run
through the output? Wouldn't it be easier to just keep the structure and
use fieldnames() to know what got read in?

Walter Roberson <roberson@hushmail.com> wrote in message <hks017$5ua$1@canopus.cc.umanitoba.ca>...
> Mark Hard wrote:
> > "Matt J " <mattjacREMOVE@THISieee.spam> wrote in message
>
> >> If there will only be 5 or 6 of them, why not just extract them manually
>
> >> nset_PickedSet2=SET{1}.nset_PickedSet2;
> >> nset_PickedSet5=SET{2}.nset_PickedSet5;
> >> etc...
>
> > Well the number of the set can change everytime depending on the
> > automatically generated text file. so I won't know the names I need to
> > variables to be to input it into the code.
>
> If you do manage to create them as individual variables and you don't
> know the names ahead of time, how is your code going to know which
> variables to manipulate? Are you going to use a who bunch of
> exist('PickedSet2','var') calls, or are you going to use who() and run
> through the output? Wouldn't it be easier to just keep the structure and
> use fieldnames() to know what got read in?

Thanks for the reply. The same variables names should be found again later on in the text file. When textscan finds that sting containing the variable name it will look at the sting after it in the text file and work on the variable accordingly. That will be my next step presuming I can get this working.

"Mark Hard" <GerryTheLeper@hotmail.com> wrote in message <hks3ck$h7$1@fred.mathworks.com>...
> Walter Roberson <roberson@hushmail.com> wrote in message <hks017$5ua$1@canopus.cc.umanitoba.ca>...
> > Mark Hard wrote:
> > > "Matt J " <mattjacREMOVE@THISieee.spam> wrote in message
> >
> > >> If there will only be 5 or 6 of them, why not just extract them manually
> >
> > >> nset_PickedSet2=SET{1}.nset_PickedSet2;
> > >> nset_PickedSet5=SET{2}.nset_PickedSet5;
> > >> etc...
> >
> > > Well the number of the set can change everytime depending on the
> > > automatically generated text file. so I won't know the names I need to
> > > variables to be to input it into the code.
> >
> > If you do manage to create them as individual variables and you don't
> > know the names ahead of time, how is your code going to know which
> > variables to manipulate? Are you going to use a who bunch of
> > exist('PickedSet2','var') calls, or are you going to use who() and run
> > through the output? Wouldn't it be easier to just keep the structure and
> > use fieldnames() to know what got read in?
>
> Thanks for the reply. The same variables names should be found again later on in the text file. When textscan finds that sting containing the variable name it will look at the sting after it in the text file and work on the variable accordingly. That will be my next step presuming I can get this working.

yet another and very strong reason to use the STRUCT approach(!)...
this will make your next task extremely(!) simple...

us

> > > Well the number of the set can change everytime depending on the
> > > automatically generated text file. so I won't know the names I need to
> > > variables to be to input it into the code.
> >
> > If you do manage to create them as individual variables and you don't
> > know the names ahead of time, how is your code going to know which
> > variables to manipulate? Are you going to use a who bunch of
> > exist('PickedSet2','var') calls, or are you going to use who() and run
> > through the output? Wouldn't it be easier to just keep the structure and
> > use fieldnames() to know what got read in?
>
> Thanks for the reply. The same variables names should be found again later on in the text file. When textscan finds that sting containing the variable name it will look at the sting after it in the text file and work on the variable accordingly. That will be my next step presuming I can get this working.

Solving problems like yours can be very confusing at first, especially when you don't know where to start. Now that you've seen a couple of solutions, I too would like to emphasize that this is a perfect scenario for using structures with dynamic field names. If you open the Matlab help and search for "creating field names dynamically", it will bring you to the page on structures, and hopefully to the part on dynamic field names, which will you give you a slightly better idea of how useful this can be.

When you're just doing some initial exploration of your data, creating simple variables like "Picked2 = [1 8 1]" is intuitive and straightforward, but when you get to the point of coding a solution that you will run on many many files, using a structure with dynamic field names can potentially really simplify your code and make it easier to debug.