Why would command syntax split string when using double quotes?

disp 'str ing' % good
disp "str ing" % error, Too many input arguments

I am on R2020a. It seems that the latter is equivalent to

disp str ing

I thought, at least

disp "str ing"

should be interpreted as

disp('"str ing"')

why would matlab split the string into several parts?

Answer 1

[Walter Roberson]

Not exactly. The latter is equivalent to disp('"str', 'ing"')

N 'str ing'

nargin = 1
 
varargin{1} =
 
str ing
 

N "str ing"

nargin = 2
 
varargin{1} =
 
"str
 
 
varargin{2} =
 
ing"
 

function N(varargin)
   fprintf('nargin = %d\n', nargin);
   celldisp(varargin)
end

why would matlab split the string into several parts?

Because that is what is documented:

https://www.mathworks.com/help/matlab/matlab_prog/command-vs-function-syntax.html

With command syntax, MATLAB passes all inputs as character vectors (that is, as if they were enclosed in single quotation marks) and does not assign outputs to variables. To pass a data type other than a character vector, use the function syntax. To pass a value that contains a space, you have two options. One is to use function syntax. The other is to put single quotes around the value. Otherwise, MATLAB treats the space as splitting your value into multiple inputs.

In order for disp "str ing" to be treated as disp("str ing") then that would directly violate the rule that MATLAB passes all inputs as character vectors.

It does not explain why MATLAB does not convert the string into a character vector, as-if disp('str ing') had been called, but the documentation never implies anything like that would happen. Instead, the documentation talks only about single-quotes to include blanks, not about double-quotes.

It would not be unreasonable for MATLAB to be enhanced to handle double-quoted strings... but it is not a bug that it does not do so at the moment.

[Xingwang Yong]

In function syntax, double quotes and single quotes make no difference, i.e. disp("str ing") == disp('str ing'). I thought they would be the same in command syntax, which is not the case though. Although documented, this is counterintuitive.

[Walter Roberson]

It is not correct that single quotes or double quotes make no difference. Consider

N('str ing')

numel = 7
#1: s
#2: t
#3: r
#4:  
#5: i
#6: n
#7: g

N("str ing")

numel = 1
#1: str ing

function N(S)
   NE = numel(S);
   fprintf('numel = %d\n', NE);
   for K = 1 : NE
      fprintf('#%d: %s\n', K, S(K));
   end
end

Character vectors with single quotes are not the same as string objects with double quotes

[Stephen23]

" In function syntax, double quotes and single quotes make no difference,"

No, strings and character arrays are very different things:

str = "not the same things";
chr = 'not the same things';
numel(str)

ans = 1

numel(chr)

ans = 19

double(str)

ans = NaN

double(chr)

ans = 1×19

110 111 116 32 116 104 101 32 115 97 109 101 32 116 104 105 110 103 115

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

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

"this is counterintuitive."

Not once you appreciate the fact the strings are a container class and have very different properties to character arrays. It seems that most confusion comes from users who attempt to smush them together into one great big mushy "I can use string and character arrays interchangeably everywhere"-class.

[Steven Lord]

We (MathWorks) went to a good bit of effort to make sure that functions that supported char row vectors prior to the introduction of string arrays behaved 'sensibly' (waving my hands a little bit) when passed in string scalars. It is true that there are some functions that behave differently; length, size, numel, and class are notable functions in this category.

f = @(x) cell2table({length(x), size(x), numel(x), class(x)}, ...
    VariableNames = ["length", "size", "numel", "class"]);
charResults = f('MATLAB')

charResults = 1×4 table
    length     size     numel     class  
    ______    ______    _____    ________

      6       1    6      6      {'char'}

stringResults = f("MATLAB")

stringResults = 1×4 table
    length     size     numel      class   
    ______    ______    _____    __________

      1       1    1      1      {'string'}

But if you want to call for example something like strlength that 'knows' what to do with char vectors and strings, it returns a sensible answer on both char row vectors and scalar strings.

[strlength('MATLAB'), strlength("MATLAB")]

ans = 1×2

6 6

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

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

Command form is another place where the behavior does differ. I don't remember off the top of my head when this was discussed (and that discussion may have happened above my pay grade and/or among a different subset of MathWorks development than the design discussions I participate in) but I would be mildly surprised if it hadn't been discussed. Backwards compatiblity (not changing the behavior of commands that already accepted text including double quotes and treated those double quotes as part of the data rather than part of the syntax) might have factored into those discussions and the decision.

[Stephen23]

And then there is the behavior with (overloaded) operators, where they diverge quite significantly.