How to Read and plot CSV file and delete infinity values from CSV file

Hi

I want to read csv file and later plot it by using some formula.

I used csvread, but it doesn’t work as the file contains text (column title) and infinity values which MATLAB read as ∞.

I tried with readtable it can read (c=readtable('sensor.csv');. But again I can’t delete infinity values from the file. When I try with c(isnan(c))=0; or c(~any(isinf(c))) = 0; it always give error

“Undefined function 'isinf' or ‘isnan’ for input arguments of type 'table'”

Can anyone help how to read large csv file (xlsread doesn’t work, file size>1gb) and delete the infinity values which MATLAB read ∞.

Small section of large file is attached sensor.csv

Using Matlab R2017b

[Walter Roberson]

Which MATLAB version are you using?

[Khalil]

Matlab R2017b

Answer 1

[per isakson]

Matlab provide many ways to read your csv-files. Here are two variants based on textscan, Read formatted data from text file or string .

sensor.csv contains "???", which I assume is your "infinity value". AFAIK, there is no way to make textscan() or any other reading function directly convert "???" to the numerical value, Inf.

In the script, R1, textscan() converts "???" to NaN and in a second step replaces NaN by Inf. That's fine if the file doesn't have missing values, since they also would be converted to Inf.

In the script, R2, the entire file is read to a character array, chr. Next all '???' are replaced by 'Inf' and finally textscan() parses chr. R2 requires more memory to store chr.

%% R1 
fid = fopen( 'sensor.csv', 'r' );
cac = textscan( fid, '%f%f%f%f%f'                       ...
            ,   'HeaderLines',3, 'CollectOutput',true   ...
            ,   'Delimiter',',', 'TreatAsEmpty','???'   );
[~] = fclose( fid );  
num = cac{1};
% num(isnan(num)) = inf;    % why not just keep the NaNs
num(isnan(num)) = 0;        % or replace NaNs by zero
display( num )

%% R2
chr = fileread( 'sensor.csv' );
% chr = strrep( chr, '???', 'inf' );
chr = strrep( chr, '???', '0' );    % or replace '???' by '0'
cac = textscan( chr, '%f%f%f%f%f'                       ...
            ,   'HeaderLines',3, 'CollectOutput',true   ...
            ,   'Delimiter',','                         );
num = cac{1};
display( num )

Both outputs

num =
            0          Inf       1.8947      0.19107    -0.015871
        4e-06       1.9911       1.9105      0.19107            0
        8e-06       1.9911       1.8947      0.19107            0
      1.2e-05       1.9751       1.9105      0.19107            0
      1.6e-05       1.9911       1.9421      0.17514            0
        2e-05       1.9751       1.9262      0.17514            0
      2.4e-05       1.9911       1.8947      0.17514            0
      2.8e-05       1.9911       1.8947      0.17514            0
      3.2e-05          Inf       1.9105      0.19107     0.015871
      3.6e-05       1.9911       1.9262      0.19107            0
        4e-05       1.9911       1.9105      0.19107            0
      4.4e-05       1.9911       1.9262      0.19107            0
      4.8e-05       1.9911       1.8947      0.19107            0
>> 

That's before I realised that you don't want Inf in your matrix.

Answer 2

[Bhaskar R]

opt = detectImportOptions('sensor.csv', 'MissingRule', 'fill', 'NumHeaderLines', 2);
T = readtable('sensor.csv', op);
header = {'Time',	'Sensor_A',	'Sensor_B',	'sensor_C',	'Sensor_D'};
T.Properties.VariableNames = header;

[Khalil]

first line giving me error message

"Error using detectImportOptions

'MissingRule' is not a recognized parameter. For a list of valid name-value pair arguments, see the documentation for

detectImportOptions."

Answer 3

[the cyclist]

I find that sometimes with these finicky imports it can be helpful to use the Import Data Tool.

I used that as a basis to make the following import script:

% If dataLines is not specified, define defaults
if nargin < 2
dataLines = [4, Inf];
end

% Set up the Import Options and import the data
opts = delimitedTextImportOptions("NumVariables", 5);

% Specify range and delimiter
opts.DataLines = dataLines;
opts.Delimiter = ",";

% Specify column names and types
opts.PreserveVariableNames = true;
opts.VariableNames = ["Time", "Sensor A", "Sensor B", "sensor C", "Sensor D"];
opts.VariableTypes = ["double", "double", "double", "double", "double"];

% Specify file level properties
opts.ExtraColumnsRule = "ignore";
opts.EmptyLineRule = "read";

% Import the data
sensorDataTable = readtable(filename, opts);

% Convert to numeric
sensorDataArray = table2array(sensorDataTable);

[the cyclist]

That script will put NaN where the input file has ????.

You could then do

sensorData(isnan(sensorData)) = Inf;

to convert to Infinity.

[Walter Roberson]

Putting an explicit function header on it:

function sensorDataArray = ReadSensorTable(filename, dataLines)
    % If dataLines is not specified, define defaults
    if nargin < 2
    dataLines = [4, Inf];
    end
    
    % Set up the Import Options and import the data
    opts = delimitedTextImportOptions("NumVariables", 5);
    
    % Specify range and delimiter
    opts.DataLines = dataLines;
    opts.Delimiter = ",";
    
    % Specify column names and types
    opts.PreserveVariableNames = true;
    opts.VariableNames = ["Time", "Sensor A", "Sensor B", "sensor C", "Sensor D"];
    opts.VariableTypes = ["double", "double", "double", "double", "double"];
    
    % Specify file level properties
    opts.ExtraColumnsRule = "ignore";
    opts.EmptyLineRule = "read";
    
    % Import the data
    sensorDataTable = readtable(filename, opts);
    
    % Convert to numeric
    sensorDataArray = table2array(sensorDataTable);
end

Answer 4

[Walter Roberson]

https://www.mathworks.com/help/matlab/ref/rmmissing.html can be used since R2016b. That is, you use readtable() and let the ??? be replaced by nan; after that rmmissing will remove the rows that have any nan, which seems to be what you are asking for.

There is also fillmissing() which tries to deduce what reasonable values might be in place of each nan.

[Khalil]

During readtable ??? is not replaced by nan, it shows ‘???’ as it. And then rmmissing is not removing those rows.

When I change table to array, the ??? are read as '∞'. And can’t be removed by rmmissing.

Fillmissing is not replacing those ???, even if the row before ‘???’ is empty and I wrote t1=fillmissing(t,’previous’), it fills that empty row with ???

[Walter Roberson]

The techniques with import options showed by The Cyclist should translate the ??? to nan. You would rmmissing after that.

[Khalil]

it gives this error "You can only call nargin/nargout from within a MATLAB function. "

[Walter Roberson]

I posted a version of his code, above, made into a function.