Multi Struct to Matrix
Afficher commentaires plus anciens
From Structs of Arrays:
s(1).x = [1 2 3; 4 5 6; 7 8 9];
s(2).x = [11 12 13; 14 15 16; 17 18 19];
I would like to get a matrix made up of, say, the second rows only - without using a loop
ans = [4 5 6; 14 15 16]
But typing
s.x(2,:) or [s.x(2,:)] or {s.x(2,:)}
gives the error
"Expected one output from a curly brace or dot indexing expression, but there were 2 results."
Any help appreciated. Jack
2 commentaires
I am not sure that those attempted syntaxes would make much sense anyway...
s.x
s(1).x, s(2).x, ... s(end).x
It makes no sense to apply indexing to the last variable in a comma separated list:
A,B,C,D(:,2)
and then expect that the indexing is applied to all variables within the comma separated list. This would be counter-intuitive. The only logical place would be after the square brackets/curly braces, in which case this becomes a standard application of indexing into a single temporary variable, like this:
[s.x](:,2)
Jack
le 18 Fév 2016
Réponses (2)
Walter Roberson
le 18 Fév 2016
cell2mat( arrayfun(@(c) c.x(2,:), (1:length(s)).', 'Uniform', 0) )
8 commentaires
Jack
le 18 Fév 2016
Delyle Polet
le 6 Août 2016
Modifié(e) : Delyle Polet
le 7 Août 2016
Nice, just missing a reference to s. Should be
cell2mat( arrayfun(@(c) c.x(2,:), s(1:length(s)).', 'Uniform', 0) )
Could also be written more simply as
cell2mat( arrayfun(@(c) c.x(2,:), s.', 'Uniform', 0) )
Walter Roberson
le 6 Août 2016
I wonder why I didn't code
cell2mat( arrayfun(@(c) c.x(2,:), s(:), 'Uniform', 0) )
Ben Oeveren
le 31 Oct 2017
Modifié(e) : Guillaume
le 31 Oct 2017
For everyone still interested. The following seems to do the job. But it is not specifically fast. It works even when you have a struct with multiple field layers. Just insert your desired string.
function m = struct2mat(s,fields)
% Convert a structure to a matrix.
% fields (string) - s(i).my1stfield.my2ndfield.my3thfield
% Ben van Oeveren,
m = [];
for i = 1:length(s)
mnew = eval(['s(i).' fields]);
m = [m;mnew];
end
end
@Ben,
Please format your code properly.
it's not specifically fast and worse it uses eval when it's not needed at all.
m = [];
fieldlist = strsplit(fields, '.');
for i = 1:numel(s)
m = [m; getfield(s(i), fieldlist{:}));
end
It would be better to get the list of field names as a cell array of individual fields rather than as a char array of field names joined by dots as this would avoid the need for the strsplit line.
Note that your code, and Walter's first answer, only work properly with vector structures, due to the problematic use of length.
eval should be avoided for such trivial code:
getfield and setfield work perfectly with multiple field names, see:
Ben Oeveren
le 31 Oct 2017
Thank you good idea. Just tested the eval on timing. The difference in time is not really big.
For 106 fields, in total a double of 10883839x3 Elapsed time is 14.001423 seconds. eval Elapsed time is 13.644880 seconds. normal loop
Nevertheless, getfield and setfield seems like good alternative solution.
Stephen23
le 31 Oct 2017
@Ben Oeveren: consider not just just the simple timing in a loop, but also that JIT does not work, debugging tools do not work, variable highlighting does not work, the security risk, etc.
And thinking that "my code does not need to be secure" is exactly why doing this is so easy:
Walter's suggestion is compact and nice. Internally this contains loops also. So the actual problem I would solve is this:
... without using a loop
Create a loop, care for a proper pre-allocation, export this to a secific M-function if you want to keep you main program clean and lean.
1 commentaire
Jack
le 18 Fév 2016
Catégories
En savoir plus sur Structures dans Centre d'aide et File Exchange
Produits
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!