Is there a reason why {'charVector'} is not considered a text scalar for argument validation purposes?

It seems like a software design inconsistency to me that this passes validation,

mustBeText({'charVector'})
mustBeScalarOrEmpty({'charVector'})

but this does not,

mustBeTextScalar({'charVector'})

Value must be a character vector or string scalar.

Is there a logic to this?

[Steven Lord]

It probably could be clearer, but I suspect one problem that the developers were trying to avoid: is C a text scalar?

S = ['ab';'cd'];
C = {S}

C = 1x1 cell array
    {2x2 char}

isscalar(C)

ans = logical
   1

iscellstr(C)

ans = logical
   1

C is a scalar and it is a cellstr, and so you could argue it is a text scalar. But the text it contains isn't really one piece of text. To me that "feels like" two pieces of text data (stored as a char matrix, not a char vector) that happens to be stored in one cell of a cell array. We could make the description of what mustBeTextScalar does more complicated (it has to be a scalar string or a row char vector or a scalar cell whose one cell contains a scalar string or a row char vector) but that's starting to get harder to parse (for a human, not for MATLAB!) And then if we define "a scalar cell whose one cell contains a scalar string or a row char vector" as a text scalar, does a cell array containing that count as a text scalar?

C2 = {{'hello'}} % text scalar or no?

C2 = 1x1 cell array
    {1x1 cell}

Lest you think that an artificial example, well it is since I wrote it up myself. But if you have space padded data (so that two pieces of text can be concatenated together) it looks a little more realistic:

C3 = {['apple '; 'banana']}

C3 = 1x1 cell array
    {2x6 char}

celldisp(C3)

 
C3{1} =
 
apple 
banana
 

iscellstr(C3)

ans = logical
   1

isscalar(C3)

ans = logical
   1

[Matt J]

C is a scalar and it is a cellstr, and so you could argue it is a text scalar.

That is food for thought, Steve, but to my mind cell arrays of char vectors exclusively are meant to be interpretted as "text", and not more general variables satisfying iscellstr(). Since mustBeText() doesn't even recognize your example as text, that seems to be where the developers were going:

mustBeText( {['ab';'cd']} )

Value must be a character vector, string array, or cell array of character vectors.

Answer 1

[Taylor]

Apologies, I should have said " A cell array is not necessarily interpretted as "text" ". I think I have a more complete understanding of the situation now. mustBeTextScalar should operate like a logical evaluation of "text" AND "scalar" given that mustBeText and mustBeTextScalar return "true" for a cell containing a character array. I have reported this inconsistency to the development team.

[Matt J]

Great. Thanks a lot!

Answer 2

[Taylor]

From the mustBeText documention "mustBeText(value) throws an error if value is not a string array, character vector, or cell array of character vectors". mustBeScalarOrEmpty returns nothing because by enclsoing the character array in curly brackets, it becomes a 1x1 cell. From the mustBeTextScalar documentation "mustBeTextScalar(value) throws an error if value is not a text scalar". The 1x1 cell is scalar but not text hence the error.

[Matt J]

The 1x1 cell is scalar but not text.

@Taylor. Thanks for your reply, but I'm having trouble with your explanation. You seem to say the issue is in whether the variable value={'charVector'} is to be considered text. Shouldn't the suggestively named mustBeText() be the highest and most reliable authority on whether value is to be considered text? If so, then why does mustBeTextScalar appear to rebel against its authority?

[Matt J]

Perhaps you are saying it is an issue of the function naming conventions? Is there meant to be a difference between a "text scalar" and "scalar text"?

[Taylor]

I think it will help to see some of this evaluated in code. I believe the root casue of your confusion is the use of curly brackets around your character array that make it a cell array. The cell array acts as a container around the character array so it has a size of 1x1 and is thus interpretted as a scalar. A cell array is not interpretted as "text" becuase it can contain any type of data.

% Using a character vector within a cell array
class({'charVector'})

ans = 'cell'

size({'charVector'})

ans = 1x2

1 1

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

% A cell containing a standard double array
class({[1 2 3]})

ans = 'cell'

size({[1 2 3]})

ans = 1x2

1 1

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

If you don't use the curly brackets and use single quotes around text, you just have a character array. This will have a size of 1xn where n is equal to the number of characters of text. A character array will be interpretted as scalar as long as the first dimension is one (a single row of text).

% Using a character array
class('charVector')

ans = 'char'

size('charVector')

ans = 1x2

1 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

Now if you used double quotes, you would create a string array. Like a character array, a string array will be interpretted as scalar as long as the first dimenion is one. However, unlike character arrays, each row of a string array does not need to be the same number of characters.

% Using a string array
class("charVector")

ans = 'string'

size("charVector")

ans = 1x2

1 1

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Matt J]

A cell array is not interpretted as "text" becuase it can contain any type of data.

But the special case of a cell array whose cells contain char vectors clearly is interpreted as text (and for good reason, it seems to me). This is seen not only in the mustBeText documentation, but also from the examples below. All of these examples throw no errors, affirming that the input cell arrays are considered text.

mustBeText({'a'})            %throws no error
mustBeText({'a','bb'})       %throws no error
mustBeText({'a','bb','ccc'}) %throws no error

It seems very natural to me that cell arrays containing character vectors should always be considered text, because they are accepted by all text manipulation functions (replace(), contains(),startsWith(), etc...). The only question, to my mind, is, why does mustBeTextScalar deviate from this?

[Stephen23]

"A cell array is not interpretted as "text" becuase it can contain any type of data."

Yet your statement directly contradicts what TMW defines to be "text" here:

https://www.mathworks.com/help/matlab/ref/mustbetext.html

which states "Validate that value is string array, character vector, or cell array of character vectors" (bold added). I can assure you that Matt J is well aware of what a cell array is and what a character vector is and what a container array is and what a string array is. Thus your explanation (gently explaining data types for beginners) completely misses the point.

"Like a character array, a string array will be interpretted as scalar as long as the first dimenion is one."

Lets test this right now. Here is a string array whose first dimension is one, exactly as you wrote:

str = ["hello","world"]

str = 1x2 string array
    "hello"    "world"

size(str,1) % fits your definition of "scalar"...

ans = 1

isscalar(str) % but it is not scalar.

ans = logical
   0

mustBeTextScalar(str) % yep, it fails here too.

Value must be a character vector or string scalar.

[Matt J]

Perhaps another way to ask it is this. Consider the following variables.

T1={'a','bb','ccc'};
T2=string(T1);

I cannot think of any text-manipulating Matlab function (replace(), contains(),startsWith(), etc...) that does not accept either T1 or T2 as input interchangeably. So why wouldn't both T1 and T2 be considered "text"?

[Stephen23]

Another another another way to ask this:

C = {'a','bb','ccc'}; % if this is
mustBeText(C) % text
X = C(1); % then why is this not
mustBeTextScalar(X) % scalar text ?

Value must be a character vector or string scalar.