passing multiple input variables when using pyrunfile

I have a python script named "clean_csv.py" that removes selected rows from a CSV and then writes a new CSV:

import pandas as pd

def clean_csv(input_path, output_path, rows_to_remove):

#Load a CSV file, remove specified rows, and save the new CSV

#Parameters:

# input path (str): full path to the input CSV file

# output_path (str): full path to the output CSV file

# rows_to_remove (list of int): Row indices (0-based) to remove

# Load the CSV into a Dataframe

df=pd.read_csv(input_path)

# drop the specified rows

df_cleaned=df.drop(index=rows_to_remove)

# write the cleaned Dataframe to a new CSV

df_cleaned.to_csv(output_path, index=False)

print(f"Cleaned CSV saved to: {output_path}")

I want to call this from MATLAB, while supplying the input_path, output_path and rows_to_remove, e.g.:

input_path = 'D:\datasets\XYZ\XYZ\XYZ.csv';

output_path = 'D:\datasets\ABC\ABC\ABC.csv';

range = [3:5468, 8859:10876];

I am working in the directory where the Py code resides. I am running 2025b with py version 3.11. I have tried some of MATLAB's posted examples and they work.

I have tried a lot of possibilities for calling clean_csv (they have all failed with various or no listed errors), but here is one of the latest attempts:

pyrunfile("clean_csv.py '" + input_path + " " + output_path + "' range")

This one gives no error, but doesn't run

thanks!

Answer 1

[Taylor]

Have you tried

pyrunfile("clean_csv.py input_path output_path range")

assuming input_path, output_path, and range are defined in the MATLAB workspace?

Also this is pretty straightforward in MATLAB if you don't want to call Python

function clean_csv(input_path, output_path, rows_to_remove)
    % Load a CSV file, remove specified rows, and save the new CSV
    % Parameters:
    % input_path (string): full path to the input CSV file
    % output_path (string): full path to the output CSV file
    % rows_to_remove (vector): Row indices (1-based) to remove
    arguments
        input_path (1,1) {mustBeFile}
        output_path (1,1) {mustBeFile}
        rows_to_remove (1,:) {mustBeNumeric, mustBePositive, mustBeReal, mustBeVector}
    end

    % Load the CSV into a table
    data = readtable(input_path);

    % Remove the specified rows
    data_cleaned = data;
    data_cleaned(rows_to_remove, :) = [];

    % Write the cleaned table to a new CSV
    writetable(data_cleaned, output_path);

    disp("Cleaned CSV saved to: " + output_path);
end

[Kip]

Taylor: I tried this - no success, no error message.

MATLAB cannot work with CSVs that have individual columns with strings and doubles. I also tried MATLAB by breaking the CSV into two parts - one with only doubles the other with strings and then saving one and appending the other to it. This looked to work (to my visual inspection of the merged CSV), but there was still something wrong with the merged file because the python program I then ran on the merged CSV failed with these. Python, on the other hand, has no problem with mixed columns.

Thanks.

[Stephen23]

"MATLAB cannot work with CSVs that have individual columns with strings and doubles"

I doubt the truth of that statement. For a start this would most likely work:

https://www.mathworks.com/help/matlab/ref/readcell.html

And READTABLE may also work, with the correct input options: https://xyproblem.info/

[Taylor]

Change your Python code to

import pandas as pd
import sys
import ast

def clean_csv(input_path, output_path, rows_to_remove):
    # Your existing function code here
    try:
        df = pd.read_csv(input_path)
        df_cleaned = df.drop(index=rows_to_remove)
        df_cleaned.to_csv(output_path, index=False)
        print(f"Cleaned CSV saved to: {output_path}")
    except Exception as e:
        print(f"Error: {e}")

if __name__ == "__main__":
    # Check if correct number of arguments provided
    if len(sys.argv) != 4:
        print("Usage: python clean_csv.py <input_file> <output_file> <rows_to_remove>")
        print("Example: python clean_csv.py input.csv output.csv '[0,2,5]'")
        sys.exit(1)
    
    input_file = sys.argv[1]
    output_file = sys.argv[2]
    
    # Convert string representation of list to actual list
    try:
        rows_list = ast.literal_eval(sys.argv[3])
    except:
        print("Error: rows_to_remove must be a valid list format like '[0,1,2]'")
        sys.exit(1)
    
    clean_csv(input_file, output_file, rows_list)    

and then run this MATLAB code

input_file = "testOriginal.csv";
output_file = "testNew.csv";
rows_to_remove = "[2,3]";

pycode = "clean_csv.py " + input_file + " " + output_file + " " + rows_to_remove;

pyrunfile(pycode)

Also, Stephen is correct, readcell is the correct function to read a text file with hetrogenous data types.

[Kip]

Well, thank you. This led me further and I now think it's working. Since this is running in a loop and the filenames and rows-to_remove are assigned new in each loop iteration, I found:

the filenames were coming in as chars, so I had to change them to a 1x1 using string(filename)

and I had to make the rows_to_remove a 1x1 also with no whitespace, so I did this:

rows_to_remove = [Startinds Stopinds]; % e.g. Startinds=3:500 and Stopinds = 2500:6000

stringParts = arrayfun(@(x) sprintf('%d,', x), rows_to_remove, 'UniformOutput', false);

resultString = [stringParts{:}]; % concatenate

resultString = resultString(1:end-1); % remove the trailing comma

resultString = strcat('[',resultString,']'); % stick the brackets on the ends

rows_to_remove = string(resultString); % string it

Now it just ran through truncating 963 files in about two minutes :)

Thanks for your help, and Stephen's too.