Preallocating table makes it slower?

12 views (last 30 days)
I have to add data to a table with a loop. I can estimate how large the table approx will be at the end. But i dont know the exact size. I also dont know the exact size of the data added with every iteration. I notized that the needed time grows nearly exponetialy when using vertcat. So i tried to preallocate the table. It made things worse!
Can some one help me out?
thanks in advance
Christian
Wrote some code to demostrate the problem:
rows = 1e6;
columns = 6;
aproxRows = rows * 1.5;
Table= table();
preallTable = array2table(zeros(aproxRows, columns));
preallTime = [];
preallTime(1) = 0;
Time = [];
Time(1) = 0;
start = 1;
for i = 1:1e3
i
addTable = array2table(rand(1000, columns)); %size of table is unknown in reality for every loop pass
End = start-1 + height(addTable);
preallTic = tic;
preallTable(start:End, :) = addTable;
preallTime(end+1) = preallTime(end) + toc(preallTic);
start = End+1;
Tic = tic;
Table = vertcat(Table,addTable);
Time(end+1) = Time(end) + toc(Tic);
end
%cut off unused space
preallTable(start:end, :) = [];
%plot Time
figure
hold on
plot(preallTime)
plot(Time)
legend({'preallTime', 'Time'})

Accepted Answer

Star Strider
Star Strider on 5 May 2020
I would preallocate the array, then do the array2table call at the end, after the array was created and is stable (nothing more to be added to it).
  2 Comments
Christian Tieber
Christian Tieber on 5 May 2020
The data to add, comes as a table. Not shure if a can change that.
Star Strider
Star Strider on 5 May 2020
Youi can easily change it to an array using the table2array function. If the variable names in the table are important, you can extract them with:
VarNames = T.Properties.VariableNames;
where ‘T’ is the original table name. When you have finished changing the array, you can the re-define the variable names in the new table (created with array2table) using ‘VarNames’ as the cell array value for the 'VariableNames' name-value pair (providing that the number of the variable names has not changed from the original table).

Sign in to comment.

More Answers (0)

Products


Release

R2019b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!