how do extract data from a multi dimensional structure array

I am working with Australian weather data from the web. I read my data into structure array S using JSON formatted data:

S = webread("http://www.bom.gov.au/fwo/IDQ60901/IDQ60901.95551.json" );

I can display the data in my command window by typing the desired field:

>> S.observations.data.rel_hum

ans =

58

ans =

63

ans =

69

ans =

69

etc...

But if I try to enter this data into a simple array for analysis I just get the first value:

>> RH = S.observations.data.rel_hum

RH =

58

>>

How can I get the data into simple variables for analysis and plotting?

Answer 1

[Walter Roberson]

put [] around the whole S.whatever expression

[WIlliam Main]

It works, I understand now

RH = [S.observations.data.rel_hum];

Thank you so much!

[Sven Baumgaartner]

Hello,

I have a similar Problem: How do you fetch the data from the second or third hour ?

I tried thinks like the following:

RH = [S.observations.data(1).rel_hum]

RH = [S.observations.data{1}.rel_hum]

But this dosent work...

Greetings from Germany

[Walter Roberson]

@Sven Baumgaartner

If you look at S.observations.data(K).local_date_time for several different K, then you can see that at least for that particular page, the observations are most recent first, and there are two observations for each hour. To locate the information for a particular day and hour, you could try something like

dt = datetime({S.observations.data.local_date_time_full}, 'InputFormat', 'yyyyMMddHHmmss');

after which you could use isbetween to determine which entries are a particular year / month / day / hour . Or you could use ymd and hms on the objects to determine the day number and the hour number (that might be an easier interface.)

Once you have a logical mask indicating which entries to select,

selected_relhum = [S.observations.data(LOGICALMASK).rel_hum];

Answer 2

[madhan ravi]

you may also look into arrayfun() and structfun() to plot struct array

RH = S.observations.data.rel_hum(:)

[WIlliam Main]

This does not appear to work, what am I doing wrong ?

>> RH = S.observations.data.rel_hum(:)

Expected one output from a curly brace or dot indexing expression, but there were 201 results.

[madhan ravi]

can you upload the data as a .mat file?

Answer 3

[AnaelG]

More generally, I've realized that s_struct(:).data is trying to output all the contents of a cell array, like calling c_cell{:}. So if you assign it to a variable it only gives you the first element.

The trick is to assign it to each cell of a pre-declared cell array of the same size:

c_array= cell(size(s_struct));
[c_array{:}]= s_struct.data

So in your case:

RH= cell(size(S.observations.data));

[RH{:}]= S.observations.data.rel_hum

But of course if it's scalar data and you already know the type you're dealing with, you can just create the array on the right hand side:

RH= [S.observations.data.rel_hum] %if doubles

or

RH= {S.observations.data.rel_hum} %if cells

etc.

[Stephen23]

"More generally, I've realized that s_struct(:).data is trying to output all the contents of a cell array"

More generally it actually returns a comma-separated list:

https://www.mathworks.com/matlabcentral/answers/1656435-tutorial-comma-separated-lists-and-how-to-use-them

There are absolutely no cell arrays involved in s_struct(:).data, just a non-scalar structure:

s_struct(1).data = 1;
s_struct(2).data = 2;
iscell(s_struct)

ans = logical
   0

iscell(s_struct(1).data)

ans = logical
   0

iscell(s_struct(2).data)

ans = logical
   0

Nope, there are absolutely no cell arrays involved anywhere in that code.

"The trick is to assign it to each cell of a pre-declared cell array of the same size"

You could do that, but by no means is that the only thing that is possible to do with comma-separated lists.

[AnaelG]

Stephen this is great info! I never knew there was such a thing as comma-separated lists in matlab. I just see it as an annoyance that it outputs the data as separate entries when I slice a cell array or struct or N-dim matrix...

Assigning to new variables in a for loop or as [c1,c2,...,cN] is not particularly practical or elegant so i'm glad I found the right combination of operators with [C{:}] ! It seems to work relatively universally with different data types.

Thanks for linking a tutorial, I'll save that reference for future use.