Matlab column extraction sematics

I'm learning Matlab and taking the Signal Processing Onramp. Can someone explain the difference between the two statements below

form1 = quakes(:,"WANC")
form2 = quakes.WANC

form1 returns a timetable with 2 columns, time and the WANC data column.

form 2 returns a column vector with just the WANC data.

quakes is a timetable with several columns, one of which is named WANC.

I should also note that the 2nd statement of these two fail which sort-of makes sense since quakes.WANC doesn't contain time info.

bandpass(quakes(:,"WANC"), [2 10])
bandpass(quakes.WANC, [2 10])

I expected form1 to be the same as form2. Is there some guide to datatypes and operations on them in Matlab that would cover this?

Answer 1

[Stephen23]

" Is there some guide to datatypes and operations on them in Matlab that would cover this?"

Of course: https://www.mathworks.com/help/matlab/matlab_prog/access-data-in-a-table.html

Note that you can easily find all of the documentation on tables: https://www.mathworks.com/help/matlab/tables.html

These documentation pages are also easy to lcoate using your favorite internet search engine.

As the documentation explains, there are actually three types of indexing into tables and timetables. For accessing the content of just one column/variable, the following are equivalent:

form3 = quakes{:,"WANC"} % note curly braces, not parentheses
form2 = quakes.WANC

"I expected form1 to be the same as form2."

In MATLAB the meaning of indexing with parentheses is very very simple: you will always get a sub-array of the same type as the array that you are indexing into. For example, if you index into a cell array using parentheses then you will get another cell array. If you index into a character array with parentheses you will get another character array. If you index into a timetable with parentheses then you will get another timetable (and not its content, which is returned by using curly-braces or dot indexing). So if you know QUAKES is a timetable, then you already know FORM1 is too:

form1 = quakes(:,"WANC") % parentheses returns same type as QUAKES is

Note that timetable FORM1 actually only has one column/variable, because the times are considered to be something like rownames (and the rownames/times are not a column/variable in the table). Confirming this is left as an exercise for the reader.

[Walter Roberson]

Stephen, there is an exception for the extracted values always being the same class when you use () indexing.

https://www.mathworks.com/help/matlab/ref/matlab.mixin.heterogeneous-class.html#mw_68b4b805-5adc-4d46-9f54-fa6d55fc2176

When you extract elements from a mixin heterogeneous array then the subset can be a different class, which is explicitly documented.

[Max Yaffe]

Thank you for your quick response, Stephen23. Now I see I have to understand several different things - Matrix vs Table vs TimeTable, and [] vs () vs {}. I think the only thing I can to to grok this is explore what happens in Matlab by playing with the data and seeing the results. And reading the manual pages you recommend. Sorry - but I'm an old C programmer so I need to know datatypes and operations to really understand the power.

[Stephen23]

"I have to understand several different things - Matrix vs Table vs TimeTable..."

MATLAB does not have a MATRIX class, but it does have TABLE and TIMETABLE classes, so the comparison does not make much sense. A matrix is merely a description of the size of an array, which could be of any class:

https://www.mathworks.com/help/matlab/matlab_prog/fundamental-matlab-classes.html

It is like comparing "dogs vs terriers vs alsatians"

"...and [] vs () vs {}"

Curly braces and parentheses can be used for indexing, but square brackets are not used for indexing. Square brackets are a concatenation operator (not a list operator like in some other languages), for deleting array elements, or for defining multiple function outputs:

https://www.mathworks.com/help/matlab/matlab_prog/matlab-operators-and-special-characters.html#bvg3oy_-5

So square brackets could be used to construct arrays which are used for indexing, but they are not an indexing operator in and of themselves.

[Max Yaffe]

Doesn't have a Matrix class? But references to "Matrix" abound, especially on the fundamental-matlab-classes page (thanks, I think that is actually what I'm looking for.) Maybe my confusion is around the term class which, typical of OO dynamic languages (Smalltalk, Python), everything is an object and every object has a class. I'll keep reading. This has been helpful.

[Walter Roberson]

There are some (many) programming languages in which an item of data type double is considered to be a different class than matrix of double . In some of those languages, any matrix of double is considered to be the same data type as any other matrix of double but in other languages, each different size of matrix is considered to be a different class, or a different class is used when fixing all but one dimension of the matrix (for example 5 by 3 by 9 matrix of double might be a compatible data type with 5 by 3 by 11 matrix of double

But in MATLAB, each object has a size attribute that is different from its class. A 5 by 3 by 9 matrix of double in MATLAB is the same class as an 1 x 99 matrix of double. There is a size() operation that is supported for every object.

There are some classes that deliberately write size restrictions into their semantics. For example the "function handle" class deliberately disallows constructing non-scalar function-handle objects -- you cannot for example do f(2) = @sin . And some classes might interpret size information a bit differently, such as indicating the number of inputs and outputs for a control system. But a tf control system with 1 input and 2 outputs is the same class as a control system with 3 inputs and 1 output.

The class for a numeric array is most often double but other numeric data types such as unsigned 16 bit integer also exist, and are different class than double .

MATLAB does not allow you to directly say that the permitted functions (methods) for [say] 3 x 3 arrays of double is diferent than the permitted functions for [say] 4 x 4 arrays of double. You can define different classes that happen to only support 3 x 3 arrays as properties and define functions for that class, but asking about the class of such an object would return the name of the class, not double

Numeric arrays and character arrays and logical arrays are fundamental data types in which every element of the arrays are of the same data types. There are also composite data types such as struct and cell arrays and table objects, in which different portions of the composite data might be different data types. string arrays require that the contents are all the same data type, but they can also be treated as a composite data type, similar to a cell array of character vectors together with some extra methods.