Unclear cause of "Target must be a numeric array of positive integer node indices" error with .csv file

I am attempting to load two large .csv files (~9 million data entries each) into column vectors and then graphing them using g = graph(s,t). I have done significant data scouring to try to find an entry in the data that is not a positive integer, but there are none. Still, I am getting the following errors when I try to graph:

Error using matlab.internal.graph.MLGraph

Target must be a numeric array of positive integer node indices.

Error in matlab.internal.graph.constructFromEdgeList (line 237)

G = matlab.internal.graph.MLGraph(s, t, totalNodes);

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Error in graph (line 329)

matlab.internal.graph.constructFromEdgeList(...

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Error in Project2Debugging (line 4)

g = graph(s,t);

I've tried using readtable, readmatrix, and csvread. readtable takes too long to run, and I get the same error using both csvread and readmatrix to onboard the data.

Here is the code using readmatrix:

s = transpose(readmatrix("s_file.csv"));
t = transpose(readmatrix("t_file.csv"));

g = graph(s,t);

Here is the code using csvread:

s = csvread("s_file.csv");
t = csvread("t_file.csv");

g = graph(s,t);

Here is the code I used to check for improper data entries. I tested it on small sample vectors to ensure it would detect negative numbers, zeros, decimals, and NaN's

for i=1:1:8935316
    
    if s(i,1)<1 | isinf(s(i,1)) | floor(s(i,1)) ~= s(i,1)
        disp("bad s" + i)
    end

    if t(i,1)<1 | isinf(t(i,1)) | floor(t(i,1)) ~= t(i,1)
        disp("bad t" + i)
    end

end

[Torsten]

Where do you check for NaN values in s and t (using isnan) ? What is the maximum node number in both arrays ?

I'd try to convert s and t to int32 before using the graph-function and see if you get results different from 0 using:

si = int32(s)
ti = int32(t)
nnz(si-s~=0)
nnz(ti-t~=0)

[Molly]

Thank you for your response! I tried plugging in the code you provided and got the following error:

Error using -

Integers can only be combined with integers of the same class, or scalar doubles.

Error in Project2Debugging (line 6)

nnz(si-s~=0);

I did use disp(nnz(s)) and disp(nnz(t)) to make sure that the number of elements was correct (which it was).

The maximum node number in both is 1.0026e+21, which is larger than I expected. Pretty sure matlab can handle doubles that size, though.

Not sure if I actually checked for NaN values, now that I think about it. When I test ran the for loop on a smaller data set, it did detect an NaN I threw it, though.

Edit: I used isnan() and nnz() to make sure the number of NaN's was 0, which it was for both matrices.

[Torsten]

The maximum node number in both is 1.0026e+21

Node numbers must be integers:

Target must be a numeric array of positive integer node indices.

That's why I recommended the above check, and it's not the case for your s and t.

But if the number of elements of s and t is 8935316, you should be able to renumber the nodes such that the integer constraint holds.

[Molly]

Renumbering the elements worked! I didn't realize that the automatic scientific notation would cause an issue since there aren't any "decimals" otherwise. Thanks for your help!