extracting rows from large table

I have a large table (23000000x11) and want to extract rows from the table based on values in the 5th column.

for example the data looks like:

66172 8:17 1 2 1 1 1 1 1 6 8

66842 7:17 1 3 0 1 1 1 1 1 7

.

.

.

and I want to create a smaller table with all the rows whose value of the fifth column are 0. This is of course easy to do with for loops and if statements, but takes enormous amounts of time to run. Is there a way using maybe ismember, intersect, or some other means to carry this out. I'm sure there must be. Any suggestions are appreciated.

Answer 1

[Walter Roberson]

mask = YourTable{:, 5} == 0;
subset0 = YourTable(mask, :);

[Mark Bodner]

Thanks Walter. Amazing answers as always!

[Siddharth Bhutiya]

For the first line of code dot would be much faster than brace.

mask = YourTable.varName == 0; % if you know the name of your fifth variable
mask = YourTable.(5) == 0; % if you dont know the name.

[Walter Roberson]

I was curious about the timing.

In my tests, the first 5 or so {:,5} accesses were reliably much slower than the others, so I skip those in the plot.

I must admit that I do not understand why {:,5} would be 5 or 6 times slower than .(5)

format long g
YourTable = array2table(randi(9, 5000, 7), 'VariableNames', {'a', 'b', 'c', 'd', 'e', 'f', 'g'});

N = 50;
skip = 7;
time_brace = zeros(N,1);
time_named = zeros(N,1);
time_dotnum = zeros(N,1);

for K = 1 : N
    tic; YourTable.(5); t=toc(); time_dotnum(K) = t;
    tic; YourTable{:,5}; t=toc(); time_brace(K) = t;
    tic; YourTable.e; t=toc(); time_named(K) = t;

end

subset = [time_brace(skip+1:end), time_named(skip+1:end), time_dotnum(skip+1:end)];
plot(subset);
legend({'\{:,5\}', '.e', '.(5)'})

mean(subset)

ans = 1×3

1.0e+00 * 7.27209302325581e-05 1.4046511627907e-05 1.01627906976744e-05