How to show a whole table in Matlab variable window

Hi,

When I make a large table and try to view it in the variable view window in the Matlab UI, instead of seeing all the values and variable names I just see a summary of the size of the table. Is there any way to view the whole table in some other way, or to extend the number of columns shown?

Thanks for any help,

R.

% Here you can see the whole table:
x = rand(10,10);
y = array2table(x);

% Here you will get a summary:
x = rand(10,10000);
y = array2table(x);

[A. Sawas]

Variables editor:

"The contents of a table are only visible and modifiable when the number of variables is fewer than 5000. When the number of variables equals or exceeds 5000, you can only view the table properties."

It seems this is a rigid limitation in the Variables Editor in Matlab. Why not to save the table as csv format and view it using Excel ?

[Neuropragmatist]

Hmmm, that's interesting, I couldn't find the documentation on this so thanks for that.

The odd thing is that in the table I want to view I don't have more than 5000 variables. However, some of the variables are cell arrays containing 100x100 (RxC) matrices. So maybe Matlab is counting them as 100 variables wide or something?

This is also the reason I don't want to save as a csv file, because I haven't found a way to write a table containing cell arrays to csv.

Answer 1

[Neuropragmatist]

I found why this was happening in my case - if you assign an empty value to a table column the variable editor switches from showing the contents of the table to showing the properties only.

Normally Matlab will not let you assign an empty [] value to a table but if you assign it to all columns of a table row (i.e. table1.variable1(1,:) ) it works but removes the table editor view.

I'm not sure if this is a bug or a feature?

The effect can be demonstrated here:

% this table can be viewed in the variable editor perfectly fine and the contents of v1 and v2 can be viewed and edited
t = table;
t.v1(1) = {rand(10,10)};
t.v2(1,:) = {rand(10,10)};

% this table cannot be viewed, only the properties can be viewed, despite the table actually being smaller
t = table;
t.v1(1) = {rand(10,10)};
e_array = [];
t.v2(1,:) = e_array;

[Peter Perkins]

Those two tables are quite different. The first table contains two "array of arrays" cell variables, the second contains a cell variable and an Nx0 double variable.

>> t = table;
>> t.v1(1) = {rand(10,10)};
>> t.v2(1,:) = {rand(10,10)}
t =
  1×2 table
          v1                v2      
    ______________    ______________
    {10×10 double}    {10×10 double}
>> 
>> t2 = table;
>> t2.v1(1) = {rand(10,10)};
>> e_array = [];
>> t2.v2(1,:) = e_array
t2 =
  1×2 table
          v1               v2     
    ______________    ____________
    {10×10 double}    [1×0 double]

I think maybe you want something more like this, but it's hard to tell.

>> t3 = table;
>> t3.v1(1) = {rand(10,10)};
>> t3.v2(1) = {e_array}
t3 =
  1×2 table
          v1               v2     
    ______________    ____________
    {10×10 double}    {0×0 double}

I'm guessing the variable editor is unhappy about t2 because of the variable v2 that has no columns, but I will make a note to have this looked into. Thanks for pointing it out.

[Neuropragmatist]

I agree that they are different, but there isn't really a reason why the second table can't be shown in the variable editor with an empty column for v2.

I can do as you suggest, but then I will lose some of the functionality of the table (ease of indexing etc) so I would rather stick with numerical arrays if possible (I know I can just cell2mat(table.column(:)) but that's a second step which seems like it should be unecessary).

For now I have just removed the possibility that my variables will be empty or I have used your solution. It's not a big deal, just a little strange.

I could also use structures instead (they can have empty fields) but I am currently converting my code from using a structure to using a table, so that would defeat the whole purpose. I really like Matlab tables, especially the ease with which you can see the table contents in the variable editor and index into columns.

I have submitted this already as a possible bug (Support Case #03565710) and the respose from TMW, in case anyone is interested, was:

Technically speaking, I believe that this behavior can be considered a bug, so I have reported it to our Development Team.

However, depending on what your end goal is, I expect that one of the following approaches could be more suitable for your work:

1) Use structures instead of tables:

>> s.v1 = rand(10); >> s.v2 = [];

With this approach there is no need to declare "s" beforehand.

2) Enclose the empty array in a cell array before assigning it to the table:

>> t = table; >> t.v1 = { rand(10) }; >> t.v2 = { [] };

[Peter Perkins]

Metioche, I'm not sure I've made myself clear.

In my previous post, take a close look at t2 and t. t2 is not an "edge case" of t, it is a completely different kind of table. I understood your question to mean, "I have a cell array in my table that contains other arrays. Why does the Variable Editor behave differently for this edge case?"

But it's not an edge case. But perhaps I have misunderstood your question.

[Neuropragmatist]

I'm not sure I understand what you mean by an 'edge case'.

I was merely pointing out that you can't view a table in the variable editor if one of the variables has been set to empty in the way shown above.

The fact that matlab will show the table when the column contains an empty cell array instead suggests to me that the former case may be a bug, but I'm not certain that's actually the case.

It could be a bug that matlab lets you do this at all:

t2.v2(1,:) = e_array

It only works in this case because the table t2 has a single row. As soon as the v2 column has contents on other rows matlab will not let you assign an empty value to that column because the sizes don't match.

So maybe this 'bug' is actually irrelevant, because how often would someone be working with a single row table, accidentally assign an empty value to a column and then want to view it in the variable editor?

As you pointed out, using a cell array {e_array} solves the issue in all cases.

Hope this helps,

R.