Is there a Faster alternative to containers.Map function ?

Hello everybody,

I have a table with big size of row.

and with this I want to define the place using the right two digits in the first column.

I tried to find this place using the containers.Map function. But I feels it is quite slow...

With the below code, it takes about 44 minutes.

Is there a faster function or a way to replace it?

clc; % Clear the command window.

close all; % Close all figures (except thos1e of imtool.)

clear; % Erase all existing variables. Or clearvars if you want.

load sample.mat

colData.PLACE = extractAfter(colData.REQ_NO,10);

keys = ["AT","AF","ET","EF","CT","CF","JP","HP","CP","DP","MP","IP","TP","KP","KF","KI","RD"];

values = ["ATT","ATT Field","ETT","ETT Field","CTT","CTTField","Gahang","Ganho","Chung","Daelim","Hung","Indu","Tenn","Ksan","Ksan Field","Ksan RD","RD"];

lookup = containers.Map(keys,values,'UniformValues', true); % containers.Map(keySet,valueSet)

num = length(colData.PLACE);

for i = 1:num

if isKey(lookup,colData.PLACE{i}) % isKey(M,keySet) M

colData.PLACE{i} = lookup(colData.PLACE{i});

else

colData.PLACE{i} = 'Not Defined';

end

end

[Walter Roberson]

Improvements are planned, I know.

Answer 1

[Bruno Luong]

I would use array/string array/cellarray because in your case the number of keys is limited and you can enumerate them wih reasonable upper bound

load sample.mat
colData.PLACE = extractAfter(colData.REQ_NO,10);

nalphabet = length('A':'Z');

keys = ["AT","AF","ET","EF","CT","CF","JP","HP","CP","DP","MP","IP","TP","KP","KF","KI","RD"];
values = ["ATT","ATT Field","ETT","ETT Field","CTT","CTTField","Gahang","Ganho","Chung","Daelim","Hung","Indu","Tenn","Ksan","Ksan Field","Ksan RD","RD"];

lookup = initlookup(keys, values);

for i = 1:num
    if isKey(lookup,colData.PLACE{i})
        colData.PLACE{i} = getval(lookup,colData.PLACE{i}); 
    else        
        colData.PLACE{i} = 'Not Defined'; 
    end
end

function lookup = initlookup(keys, values)
lookup = string(missing);
lookup(idxfun(keys)) = values;
end

function [b, val] = iskey(lookup, key)
letter = char(key)-'A'+1; 
b = all(letter > 1 & letter < 26);
if b
    val = lookup(idxfun(key));
    b = ~ismissing(val);
else
    val = string(missing);
end
end

function val = getval(lookup, key)
[~, val] = iskey(lookup, key);
end

function idx = idxfun(keys)
idx = zeros(size(keys));
for k = 1:numel(idx)
    letters = char(keys(k))-'A'+1;
    idx(k) = sub2ind([26 26], letters(1), letters(2));
end
end

I agree that containerMap on the paper using hash should be fast but MATLAB implementation just kills the performance. No wonder not many people use it in practice.

[Smithy]

As your proposal, I changed my table to cell array and run the same code as before. then it is much faster than before. Now it takes 10 seconds. (before it took around 40 minutes).

Answer 2

[Bruno Luong]

In new release R2022b the dictionary is new implementation of hashing search/insertion. It looks great.