Propose New Type Interval

[latot]

Hi, well thinking again, actually when we add functions to sets we are doing bigger and bigger the files while the type of data are completely differents, so i propose make a new type, actually we have @symfun, @sym and some @logical and @double functions, so do a @interval to call the functions.

I think with this will be easier to work.

This start when i was checking the changes of this pr #328 and your comment:

Lots of questions now that we have sets! Here is special treatment for isempty. What about
numel, and size and length and... and ...

Edited: set its already a function, so now is interval.

Thx. Cya.

[latot]

A little question D:
i have problems with classes, like my firsts steps in this, i make a new folder, @interval then i put the interval file there, with this little changes:

function I = interval(varargin)
  if (nargin < 2 || nargin > 4)
    print_usage();
  end
  varargin = sym(varargin);
  I = python_cmd ('return Interval(*_ins),', varargin{:});
  I = class(I, 'interval');
end

and now trying to execute this:

>> c=interval(1, 9)
c = <class interval>
>> c=interval(1, sym(9))
c = (sym) [1, 9]

This is weird!
The interval function use sym(varargin) so shouldn't be a difference but it only works when it have at least a sym (interval of @sym its generated for octave if i'm right, so both are equals).
What do you think?

Thx.

[cbm755]

Your 2nd example calls @sym/interval, that's why it's different.

[latot]

yes, but @sym/interval its the same as @interval/interval, so i know tis call the other but i cat'n understand why the result is different.

Thx.

[latot]

Okis, checking again, the files are differents, that why with one works and with the other not.

Thx the help!

[latot]

what doc do you can recommend me to learn to write types/classes?
i search about classes but i it feel i have the half of info to start writing it, i read symfun.m and sym.m, and still are things i don't understand.
D:

This is my newbie file: https://github.com/latot/octsympy/blob/interval/inst/%40interval/interval.m

still something fails, ..., basically this:
this relatively works

>> interval(1, 2)
ans = (interval) [1, 2]  (1×4 matrix)

this don't:

>> interval(1, 2, sym(3))
error: Python exception: NotImplementedError: left_open and right_open can have only true/false values, got 3 and False
    occurred at line 1 of the Python code block
error: called from
    python_cmd at line 174 column 5
    interval at line 62 column 5

i use logical so the sym(3) should be converted to True but nothing, and the last:

>> interval(1, sym(2))
ans = (sym) [1, 2]

The class changes, as a note, only exist the file @interval/interval.m, the file @sym/interval.m was removed.

Thx. Cya.

[cbm755]

I used the Octave docs to learn classes, maybe some Matlab docs are useful too.

I don't really understand the superiorto bit (maybe I did but I forgot ;-)

I'm not at all sure that sym(3) will get converted to a python bool...

I'm not sure what we get by making an interval class.

In the future, maybe we want all sym objects to actually be pyobject (see Pytave) and then class(x) would return their Sympy python class.

a = interval(1, 2)
isa(a, 'sym')   % true
isa(a, 'pyobject')   % true
class(a)     % calls @pyobject/class, answers with `py.sympy.Interval` or similar.

[latot]

that looks great, the reason of the issue, is (thinking only in sym and intervals), we have basically there two types of data, and if we split it to @sym and @interval we will don't need start checking every type in the code.
Maybe are other way to do this.

[cbm755]

Perhaps we could try overloading @sym/class right now? No need to wait for full PyTave support.

[cbm755]

we will don't need start checking every type in the code.

Do we do a lot of that right now for intervals?

My goal is to make Symbolic "thinner" around SymPy. I feel adding more classes for particular types of SymPy objects does the opposite. So there will have to be some big benefit; and I don't see what that is.

[latot]

sorry but i have little problems interpreting the overloading "line", what do you mean? (only as a note, i know what is overload a function/operator) (maybe your intention was say start implementing the class def).

execute one action depending if is set or matrix: https://github.com/cbm755/octsympy/blob/master/inst/%40sym/intersect.m
don't is very expensive but i think if its possible avoid that, try split it.

only as idea, its possible make a classdef of classdef?, like a master classdef to handle values types and similars and the other to set the opertaions. (or something similar).

[cbm755]

sorry but i have little problems interpreting the

Let's take that discussion elsewhere: #549.

like a master classdef to handle values types and similars and the other to set the opertaions

Yes it is possible to subclass.

[cbm755]

execute one action depending if is set or matrix:

So the natural place to split is for SymPy class Set (not Interval). Now I understand your comment earlier "set its already a function, so now is interval."

This thread is really just PR #328. I'll close this: it'll be cross-referenced from #328.

[latot]

okidokis