claude-scientific-skills/scientific-skills/matlab/references/python-integration.md

Python Integration Reference

Table of Contents

1. Calling Python from MATLAB
2. Data Type Conversion
3. Working with Python Objects
4. Calling MATLAB from Python
5. Common Workflows

Calling Python from MATLAB

Setup

% Check Python configuration
pyenv

% Set Python version (before calling any Python)
pyenv('Version', '/usr/bin/python3');
pyenv('Version', '3.10');

% Check if Python is available
pe = pyenv;
disp(pe.Version);
disp(pe.Executable);

Basic Python Calls

% Call built-in functions with py. prefix
result = py.len([1, 2, 3, 4]);  % 4
result = py.sum([1, 2, 3, 4]);  % 10
result = py.max([1, 2, 3, 4]);  % 4
result = py.abs(-5);            % 5

% Create Python objects
pyList = py.list({1, 2, 3});
pyDict = py.dict(pyargs('a', 1, 'b', 2));
pySet = py.set({1, 2, 3});
pyTuple = py.tuple({1, 2, 3});

% Call module functions
result = py.math.sqrt(16);
result = py.os.getcwd();
wrapped = py.textwrap.wrap('This is a long string');

Import and Use Modules

% Import module
np = py.importlib.import_module('numpy');
pd = py.importlib.import_module('pandas');

% Use module
arr = np.array({1, 2, 3, 4, 5});
result = np.mean(arr);

% Alternative: direct py. syntax
arr = py.numpy.array({1, 2, 3, 4, 5});
result = py.numpy.mean(arr);

Run Python Code

% Run Python statements
pyrun("x = 5")
pyrun("y = x * 2")
result = pyrun("z = y + 1", "z");

% Run Python file
pyrunfile("script.py");
result = pyrunfile("script.py", "output_variable");

% Run with input variables
x = 10;
result = pyrun("y = x * 2", "y", x=x);

Keyword Arguments

% Use pyargs for keyword arguments
result = py.sorted({3, 1, 4, 1, 5}, pyargs('reverse', true));

% Multiple keyword arguments
df = py.pandas.DataFrame(pyargs( ...
    'data', py.dict(pyargs('A', {1, 2, 3}, 'B', {4, 5, 6})), ...
    'index', {'x', 'y', 'z'}));

Data Type Conversion

MATLAB to Python

MATLAB Type	Python Type
double, single	float
int8, int16, int32, int64	int
uint8, uint16, uint32, uint64	int
logical	bool
char, string	str
cell array	list
struct	dict
numeric array	numpy.ndarray (if numpy available)

% Automatic conversion examples
py.print(3.14);         % float
py.print(int32(42));    % int
py.print(true);         % bool (True)
py.print("hello");      % str
py.print({'a', 'b'});   % list

% Explicit conversion to Python types
pyInt = py.int(42);
pyFloat = py.float(3.14);
pyStr = py.str('hello');
pyList = py.list({1, 2, 3});
pyDict = py.dict(pyargs('key', 'value'));

Python to MATLAB

% Convert Python types to MATLAB
matlabDouble = double(py.float(3.14));
matlabInt = int64(py.int(42));
matlabChar = char(py.str('hello'));
matlabString = string(py.str('hello'));
matlabCell = cell(py.list({1, 2, 3}));

% Convert numpy arrays
pyArr = py.numpy.array({1, 2, 3, 4, 5});
matlabArr = double(pyArr);

% Convert pandas DataFrame to MATLAB table
pyDf = py.pandas.read_csv('data.csv');
matlabTable = table(pyDf);  % Requires pandas2table or similar

% Manual DataFrame conversion
colNames = cell(pyDf.columns.tolist());
data = cell(pyDf.values.tolist());
T = cell2table(data, 'VariableNames', colNames);

Array Conversion

% MATLAB array to numpy
matlabArr = [1 2 3; 4 5 6];
pyArr = py.numpy.array(matlabArr);

% numpy to MATLAB
pyArr = py.numpy.random.rand(int64(3), int64(4));
matlabArr = double(pyArr);

% Note: numpy uses row-major (C) order, MATLAB uses column-major (Fortran)
% Transposition may be needed for correct layout

Working with Python Objects

Object Methods and Properties

% Call methods
pyList = py.list({3, 1, 4, 1, 5});
pyList.append(9);
pyList.sort();

% Access properties/attributes
pyStr = py.str('hello world');
upper = pyStr.upper();
words = pyStr.split();

% Check attributes
methods(pyStr)          % List methods
fieldnames(pyDict)      % List keys

Iterating Python Objects

% Iterate over Python list
pyList = py.list({1, 2, 3, 4, 5});
for item = py.list(pyList)
    disp(item{1});
end

% Convert to cell and iterate
items = cell(pyList);
for i = 1:length(items)
    disp(items{i});
end

% Iterate dict keys
pyDict = py.dict(pyargs('a', 1, 'b', 2, 'c', 3));
keys = cell(pyDict.keys());
for i = 1:length(keys)
    key = keys{i};
    value = pyDict{key};
    fprintf('%s: %d\n', char(key), int64(value));
end

Error Handling

try
    result = py.some_module.function_that_might_fail();
catch ME
    if isa(ME, 'matlab.exception.PyException')
        disp('Python error occurred:');
        disp(ME.message);
    else
        rethrow(ME);
    end
end

Calling MATLAB from Python

Setup MATLAB Engine

# Install MATLAB Engine API for Python
# From MATLAB: cd(fullfile(matlabroot,'extern','engines','python'))
# Then: python setup.py install

import matlab.engine

# Start MATLAB engine
eng = matlab.engine.start_matlab()

# Or connect to shared session (MATLAB: matlab.engine.shareEngine)
eng = matlab.engine.connect_matlab()

# List available sessions
matlab.engine.find_matlab()

Call MATLAB Functions

import matlab.engine

eng = matlab.engine.start_matlab()

# Call built-in functions
result = eng.sqrt(16.0)
result = eng.sin(3.14159 / 2)

# Multiple outputs
mean_val, std_val = eng.std([1, 2, 3, 4, 5], nargout=2)

# Matrix operations
A = matlab.double([[1, 2], [3, 4]])
B = eng.inv(A)
C = eng.mtimes(A, B)  # Matrix multiplication

# Call custom function (must be on MATLAB path)
result = eng.myfunction(arg1, arg2)

# Cleanup
eng.quit()

Data Conversion (Python to MATLAB)

import matlab.engine
import numpy as np

eng = matlab.engine.start_matlab()

# Python to MATLAB types
matlab_double = matlab.double([1.0, 2.0, 3.0])
matlab_int = matlab.int32([1, 2, 3])
matlab_complex = matlab.double([1+2j, 3+4j], is_complex=True)

# 2D array
matlab_matrix = matlab.double([[1, 2, 3], [4, 5, 6]])

# numpy to MATLAB
np_array = np.array([[1, 2], [3, 4]], dtype=np.float64)
matlab_array = matlab.double(np_array.tolist())

# Call MATLAB with numpy data
result = eng.sum(matlab.double(np_array.flatten().tolist()))

Async Calls

import matlab.engine

eng = matlab.engine.start_matlab()

# Asynchronous call
future = eng.sqrt(16.0, background=True)

# Do other work...

# Get result when ready
result = future.result()

# Check if done
if future.done():
    result = future.result()

# Cancel if needed
future.cancel()

Common Workflows

Using Python Libraries in MATLAB

% Use scikit-learn from MATLAB
sklearn = py.importlib.import_module('sklearn.linear_model');

% Prepare data
X = rand(100, 5);
y = X * [1; 2; 3; 4; 5] + randn(100, 1) * 0.1;

% Convert to Python/numpy
X_py = py.numpy.array(X);
y_py = py.numpy.array(y);

% Train model
model = sklearn.LinearRegression();
model.fit(X_py, y_py);

% Get coefficients
coefs = double(model.coef_);
intercept = double(model.intercept_);

% Predict
y_pred = double(model.predict(X_py));

Using MATLAB in Python Scripts

import matlab.engine
import numpy as np

# Start MATLAB
eng = matlab.engine.start_matlab()

# Use MATLAB's optimization
def matlab_fmincon(objective, x0, A, b, Aeq, beq, lb, ub):
    """Wrapper for MATLAB's fmincon."""
    # Convert to MATLAB types
    x0_m = matlab.double(x0.tolist())
    A_m = matlab.double(A.tolist()) if A is not None else matlab.double([])
    b_m = matlab.double(b.tolist()) if b is not None else matlab.double([])

    # Call MATLAB (assuming objective is a MATLAB function)
    x, fval = eng.fmincon(objective, x0_m, A_m, b_m, nargout=2)

    return np.array(x).flatten(), fval

# Use MATLAB's plotting
def matlab_plot(x, y, title_str):
    """Create plot using MATLAB."""
    eng.figure(nargout=0)
    eng.plot(matlab.double(x.tolist()), matlab.double(y.tolist()), nargout=0)
    eng.title(title_str, nargout=0)
    eng.saveas(eng.gcf(), 'plot.png', nargout=0)

eng.quit()

Sharing Data Between MATLAB and Python

% Save data for Python
data = rand(100, 10);
labels = randi([0 1], 100, 1);
save('data_for_python.mat', 'data', 'labels');

% In Python:
% import scipy.io
% mat = scipy.io.loadmat('data_for_python.mat')
% data = mat['data']
% labels = mat['labels']

% Load data from Python (saved with scipy.io.savemat)
loaded = load('data_from_python.mat');
data = loaded.data;
labels = loaded.labels;

% Alternative: use CSV for simple data exchange
writematrix(data, 'data.csv');
% Python: pd.read_csv('data.csv')

% Python writes: df.to_csv('results.csv')
results = readmatrix('results.csv');

Using Python Packages Not Available in MATLAB

% Example: Use Python's requests library
requests = py.importlib.import_module('requests');

% Make HTTP request
response = requests.get('https://api.example.com/data');
status = int64(response.status_code);

if status == 200
    data = response.json();
    % Convert to MATLAB structure
    dataStruct = struct(data);
end

% Example: Use Python's PIL/Pillow for advanced image processing
PIL = py.importlib.import_module('PIL.Image');

% Open image
img = PIL.open('image.png');

% Resize
img_resized = img.resize(py.tuple({int64(256), int64(256)}));

% Save
img_resized.save('image_resized.png');