I have an .csv file made in Excel and that Excel uses ; as delimiter. I don't like that, but I cannot change it.

So I have to accept it. I want to read that .csv file in GNU Octave, but I can't really do it because the float numbers inside the .csv file is separated with , e.g 15,25 and not separated with . e.g 15.25 because in Excel, it assumes that 15.25 is a text string and 15,25 is a number. Yes, I know, it's weird.

If I change all , to . in my .csv file, then I can read the .csv file with dlmread or csvread. But in this case, I don't want to change , to . because that's a standard Excel configuration.

So my question is:

If you have a .csv file were the float numbers are displayed with ,. How can you read that .csv file in GNU Octave then?

CodePudding user response：

Sample file

i1;i2;i3
1,234;34,134;5,987
3,14;6,96;85,05

Script

clearvars
clc

nrows = 2;
ncolumns = 3;

data = textscan (fopen("data.csv", 'r'), "%s", "Delimiter", ";", "HeaderLines", 1);
data = cell2mat(data);
data = str2double(strrep(data,',','.'));
data = reshape(data, ncolumns, nrows); % elements are accessed in column-major order
data = transpose(data)

Logic

1. We read the contents of the file as strings separated by ;.
2. We replace the , present in the strings with ..
3. Convert the strings to doubles.
4. The data is reshaped to a matrix.

References

• Specify decimal separator for .dat file in matlab

CodePudding user response：

For octave specifically, here is a one-liner solution.

str2double( strrep( csv2cell( 'testo.csv', ';' ), ',', '.' ) )

Requires the io pkg to be loaded, for the csv2cell function.