How can I get a class property to be overridden and handled as a variable?

7 次查看（过去 30 天）
Hakon Haugnes
Hakon Haugnes 2019-6-4
评论： Hakon Haugnes 2019-6-4
Some of the Matlab classes have a great features, e.g. TABLE, by letting the class be called with a property that is really a variable inside the class.
For example:
somedata = {'a','b','c'}
T = table (somedata)
And you can then access the row of "a" by calling
T.a
and you can even call it with "modifiers", e.g.
T.a(end-1:end)
to get the last two rows
How can this be replicated in my own class definition / how does a class "capture" that this is not a real method/property, but to be treated as a variable? And how is the "end-1:end" command captured in this case so it can be used in the class?

请先登录，再回答此问题。

采纳的回答

Guillaume
Guillaume 2019-6-4
There are various ways to customise properties such as dynamic properties.
However, what you're looking at here with table has nothing to do with customised or overriden properties. The T.xxx T{xxx, yyy} and T(xxx, yyy) syntax override the default indexing (subsref) and default assignment (subsasgn) methods. As for end, it is the end method that is overriden.
For example the line
T{end-1:end, 'a'}
is actually
  • two calls to end(T, 1)
  • a call to numArgumentsFromSubscript
  • finally a call to subsref(t, struct('type', '{}','subs', {[calculatedindex calculatedindex], 'a'})
The line
T.a(end-1:end)
is actually a call to
subsref(t, struct('type', '.', 'subs', {'a'});
followed by a plain matrix indexing.
You can do the same for your own class. It's all explained under Obejct Indexing. However, be aware that it's a lot of work to get right. You can actually see how it's implemented for tables. It's all in m files. The simples way is to use the debugger to step through a line that index a table. Or you can find the implementation of the overriding methods in
fullfile(matlabroot, 'toolbox\matlab\datatypes\@tabular')
  3 个评论
显示 1更早的评论
Guillaume
Guillaume 2019-6-4
Personally, I don't bother. As you can see it's a complex machinery and I dislike the fact that even if you just want to override one type of indexing (e.g. {}) you lose the built-in indexing for the other two so have to reimplement () and . indexing yourself.
So I just implement my own indexing method,e g. myclass.at(index). Less headache.
Hakon Haugnes
Hakon Haugnes 2019-6-4
Excellent advice and I agree with this approach until absolutely necessary to bite the bullet.
Maybe I can ask the following which then bugs me:
classdef class1 < handle
properties
data = [] %does not really matter
end
methods
function self = class1() %constructor
a=123; %does not matter
end
% some method that returns an array that we may want to index into as a chain
function v=mydata(arg1)
switch arg1
case 'a'
v = [1 2 3];
end
end
%some indexing method
function v=at(arg1)
fprintf('At');
disp(arg1)
end
end
end
I wonder if it is possible to chain class calls somehow like this?
c=class1;
c.v % Return the whole v array. Trivial. Works.
% Now can we return just part of v somehow in a single chain
% without invoking arguments on the call on v itself ?
% the problem is that v is just an array, not an instance of the class so doesn't have methods!
c.v(end) % this would work but I'd like to keep the argument for other things
c.v.at(end) % Separate index method, but this chaining does not work

请先登录，再进行评论。

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Loops and Conditional Statements 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by