How to reformat this table's information so that a column of data becomes the header's for matching data?
3 Ansichten (letzte 30 Tage)
Ältere Kommentare anzeigen
Austin B
am 6 Jun. 2022
Beantwortet: Peter Perkins
am 13 Jun. 2022
I apologize for the unclear title, not sure how to properly describe my request.
I have a table of data that is formatted as such when read into a matlab table:
Items Family
_______ __________
"Item1" "Family 1"
"Item2" "Family 2"
"Item3" "Family 4"
"Item4" "Family 2"
"Item5" "Family 3"
"Item6" "Family 1"
"Item7" "Family 1"
"Item8" "Family 3"
I want to convert this into a differently formatted table that has the items separated by families, like this:
Family 1 Family 2 Family 3 Family 4
________ ________ _________ _________
"Item1" "Item2" "Item5" "Item3"
"Item6" "Item4" "Item8" <missing>
"Item7" <missing> <missing> <missing>
I am struggling to come up with a solution for this. How can this be achieved, and is a table the best way to format this data?
0 Kommentare
Akzeptierte Antwort
Voss
am 6 Jun. 2022
Here's one way to do it:
% construct the initial table:
Items = ["Item1";"Item2";"Item3";"Item4";"Item5";"Item6";"Item7";"Item8"];
Family = ["Family 1";"Family 2";"Family 4";"Family 2";"Family 3";"Family 1";"Family 1";"Family 3"];
t = table(Items,Family)
% group t.Family:
[idx,group_ids] = findgroups(t.Family);
n_groups = numel(group_ids); % number of Family groups
% split t.Items into groups according to Family:
grouped_items = splitapply(@(x){x},t.Items,idx);
n_items = cellfun(@numel,grouped_items); % number of Items in each Family group
% construct 2d string array of Item* Strings, where each column
% is the Items in a Family group, filled out with <missing> String(s):
vals = repmat(string(missing()),max(n_items),n_groups);
for ii = 1:n_groups
vals(1:n_items(ii),ii) = grouped_items{ii};
end
% convert 2d String array into a cell array,
% to be used as an argument list in table():
vals = num2cell(vals,1);
% create the new table:
new_t = table(vals{:},'VariableNames',group_ids)
2 Kommentare
Weitere Antworten (1)
Peter Perkins
am 13 Jun. 2022
I guess this is a useful format? Voss shows one way, here's another:
>> Items = "Item" + (1:8)';
>> Family = "Family" + [1;2;4;2;3;1;1;3];
>> t = table(Items,Family);
>> t2 = rowfun(@(x)[x' repmat(missing,1,3-length(x))],t,"GroupingVariable","Family","OutputVariableName","Items")
t2 =
4×3 table
Family GroupCount Items
_________ __________ _________________________________
"Family1" 3 "Item1" "Item6" "Item7"
"Family2" 2 "Item2" "Item4" <missing>
"Family3" 2 "Item5" "Item8" <missing>
"Family4" 1 "Item3" <missing> <missing>
>> t3 = array2table(t2.Items',"VariableNames",t2.Family)
t3 =
3×4 table
Family1 Family2 Family3 Family4
_______ _________ _________ _________
"Item1" "Item2" "Item5" "Item3"
"Item6" "Item4" "Item8" <missing>
"Item7" <missing> <missing> <missing>
But I might argue that there are better ways to store these "ragged" data, for example as t2.
0 Kommentare
Siehe auch
Kategorien
Mehr zu Matrix Indexing finden Sie in Help Center und File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!