Velocity vector plot with empty cell array

With the following code I made a velocity vector plot over time.

%% Velocity vector plot
close all

x_cent = mean(X(1,1:2));
z_cent = mean(Z(1:2,1));
Nx2 = 49;
Nz2 = 49;
bins_x2 = linspace(x_cent,-x_cent,Nx2);
% Bins defined over the height of the silo
bins_z2 = linspace(0.6-z_cent,z_cent,Nz2);

[X2,Z2] = meshgrid(bins_x2,bins_z2);

for i = 1:N_run
    for t = 1:length(col)
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));
        vz_avr{t,i} = cellfun(@(x)mean(x(:,2)),iwant2(:,:,t,i));
    end
end
for i = 1:N_run
    for t = 1:5
        vx_rot{t,i} = rot90(vx_avr{t,i},2);
        vz_rot{t,i} = rot90(vz_avr{t,i},2);
    end
end

for i = 1:N_run
    figure(i)
    hold on
    for t = 1:length(col)
        quiver(X2,Z2,vx_rot{t,i},vz_rot{t,i})
    end
end

Below you see an example of a velocity vector plot at t = 4 (which represents 2 seconds)

However, when t = 6 (representing 3 seconds) there are no velocity vectors left resulting in an empty iwant2. When I use the function cellfun to take the mean I get of course an error, because it's empty. How can I make sure that matlab ignores the empty iwant2 and just plot an empty quiver instead?

Answer 1

[Tessa Kol]

Solved it myself with an if else statement, see code below.

for i = 1:N_run
    for t = 1:length(col)
        if cellfun(@isempty,iwant2(:,:,t,i))
            vx_avr{t,i} = zeros(Nz2,Nx2);
            vz_avr{t,i} = zeros(Nz2,Nx2);
        else
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));
        vz_avr{t,i} = cellfun(@(x)mean(x(:,2)),iwant2(:,:,t,i));
        end
    end
end

Thank you Adem Danz for your suggestions, it really helped me in finding the final solution to this problem.

Answer 2

[Adam Danz]

We can't run your code due to missing variables but here are two general suggestions.

Before before we get to that,

• always preallocate loop variables as shown below
• use numel() instead of length()

1. If iwant2 is empty, just skip to the next loop and check for empties on all other loops, too. See the 3 conditionals that check for empty arrays.

% Pre allocate loop vars!
vx_avr = cell(numel(col), N_run); 
vz_avr = vx_avr; 
vx_rot = vx_avr; 
vz_rot = vx_avr; 

for i = 1:N_run
    for t = 1:numel(col)
        %if isempty(iwant2(:,:,t,i)); update: this won't work
        %    continue
        %end
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));
        vz_avr{t,i} = cellfun(@(x)mean(x(:,2)),iwant2(:,:,t,i));
    end
end
for i = 1:N_run
    for t = 1:5
        if isempty(vx_avr{t,i})
            continue
        end
        vx_rot{t,i} = rot90(vx_avr{t,i},2);
        vz_rot{t,i} = rot90(vz_avr{t,i},2);
    end
end

for i = 1:N_run
    figure(i)
    hold on
    for t = 1:numel(col)
        if isempty(vz_rot{t,i})
            continue
        end
        quiver(X2,Z2,vx_rot{t,i},vz_rot{t,i})
    end
end

2. If you want to plot "empty" quiver objects, use the same example above but change the last condition to

for i = 1:N_run
    figure(i)
    hold on
    for t = 1:numel(col)
        if isempty(vz_rot{t,i})
            quiver(nan,nan)
        else
            quiver(X2,Z2,vx_rot{t,i},vz_rot{t,i})
        end
    end
end

[Tessa Kol]

Thank you for helping me and the usefull suggestions of preallocate and numel. However, I did not use

numel(col)

because it counts all the elements of the 7x2 array. Thus, this will give me a value of 14 instead of the value 7 I acquired with using

length(col)

But I understand that numel works faster than length.

I did got an error when implementing your piece of code as follow:

vx_avr = cell(length(col), N_run);
vz_avr = vx_avr;

for i = 1:N_run
    for t = 1:length(col)
        if isempty(iwant2(:,:,t,i))
            continue
        end
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));
        vz_avr{t,i} = cellfun(@(x)mean(x(:,2)),iwant2(:,:,t,i));
    end
end

Index in position 2 exceeds array bounds.

Error in SiloV1_results (line 158)
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));

Error in SiloV1_results (line 158)
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));
        

Instead of using continue I used a while loop:

vx_avr = cell(length(col), N_run);
vz_avr = vx_avr;

for i = 1:N_run
    for t = 1:length(col)
        while isempty(iwant2(:,:,t,i))
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));
        vz_avr{t,i} = cellfun(@(x)mean(x(:,2)),iwant2(:,:,t,i));    
        end
    end
end

But this code returns empty cell array, which are not supposed to be empty

[Adam Danz]

I guessed that col was a vector. If col is a matrix, use szdim = size(A,dim) instead of length().

Also, I commented-out the first conditional in my answer because I realized iwant2 is a cell array and that conditional test will not work with cell arrays.

There are too many undefined variables. I can troubleshoot further if you provide something I can run on my end.

[Tessa Kol]

iwant2 is created with the following code:

% Number of bins over the x-axis
Nx = 75;
% Number of bins over the z-axis
Nz = 67;
% Bins defined over the width of the silo
bins_x = linspace(-0.3077,0.2964,Nx);
% Bins defined over the height of the silo
bins_z = linspace(0.224,0.7628,Nz);

[X,Z] = meshgrid(bins_x,bins_z);

for i = 1:N_run
    for t = 1:length(col)
        % Plotting the figures is optional and only serves as visualisation purpose
%         figure()
        xc{t,i} = expData_struc{1,i}{1,row(t,i)}(:,1);
        zc{t,i} = expData_struc{1,i}{1,row(t,i)}(:,3);
        vx{t,i} = velData_struc{1,i}{1,row(t,i)}(:,1);
        vz{t,i} = velData_struc{1,i}{1,row(t,i)}(:,3);
%         plot(X,Z,'r',X',Z','r')
%         hold on
        for m = 1:Nz-1
            for n = 1:Nx-1
                P = [X(m,n) Z(m,n) ; 
                     X(m,n+1) Z(m,n+1) ; 
                     X(m+1,n+1) Z(m+1,n+1) ;
                     X(m+1,n) Z(m+1,n)] ;

                idx = inpolygon(xc{t,i}(:),zc{t,i}(:),P(:,1),P(:,2)); 
                iwant{m,n,t,i} = [xc{t,i}(idx) zc{t,i}(idx)] ;
                iwant2{m,n,t,i} = [vx{t,i}(idx) vz{t,i}(idx)];
%                 plot(xc{t,i}(idx),zc{t,i}(idx),'.')
%                 drawnow
            end
        end
    end
end

I results in a 4D-cell. When it comes to

iwant2(:,:,7,1)

the cell array is empty (see pictures)

Unfortunatly, I can't provide the whole code nor can I transfer the big data files. I hope this is enough information.

[Adam Danz]

If I could work with a minimal working example that reproduces the problem it would probably take a couple of minutes to see the problem. But without that, I'm required to read lines of code, guess the values of variables and their sizes, and I have to reverse engineer what you're doing which takes much more time that I unfortunately don't have today.

[Tessa Kol]

I realized that you only need the data of iwant2 (attached to this post), which is not a big .mat file. I should have realized it sooner, thus sorry for the inconvenience. Is that enough to serve as a working example?

[Adam Danz]

N_run, col, and possibly other variables are also not defined.

[Tessa Kol]

N_run = 2 and length(col) = 7

[Adam Danz]

There are other missing vars, too. One example is expData_struc.

You could clear your workspace and try to run only the code you provided so you can see which vars are not defined.

[Tessa Kol]

expData_struc is very big (over 1 GB), so that is problematic. I thought that sending the iwant2 in the .mat file iit could reproduce the problem enough. Since I get errors with the following code:

vx_avr = cell(length(col), N_run);
vz_avr = vx_avr;
for i = 1:N_run
    for t = 1:length(col)
        if isempty(iwant2(:,:,t,i))
            continue
        end
        vx_avr{t,i} = cellfun(@(x)mean(x(:,1)),iwant2(:,:,t,i));
        vz_avr{t,i} = cellfun(@(x)mean(x(:,2)),iwant2(:,:,t,i));
    end
end

For that piece of code I thought you only need the vars iwant2, length(col) and N_run.