Main Content

Program a MATLAB Function in a Chart

A MATLAB® function in a Stateflow® chart is a graphical element that you use to write algorithms that are easier to implement by calling built-in MATLAB functions. This type of function is useful for coding algorithms that are more easily expressed by using MATLAB instead of the graphical Stateflow constructs. For more information, see Reuse MATLAB Code by Defining MATLAB Functions.

Inside a MATLAB function, you can call these types of functions:

  • Local functions defined in the body of the MATLAB function.

  • Graphical, Simulink®, truth table, and other MATLAB functions in the chart.

  • Built-in MATLAB functions that support code generation. These functions generate C code for building targets that conform to the memory and data type requirements of embedded environments.

  • Extrinsic MATLAB functions that do not support code generation. These functions execute only in the MATLAB workspace during simulation of the model. For more information, see Call Extrinsic MATLAB Functions in Stateflow Charts.

  • Simulink Design Verifier™ functions for property proving and test generation. These functions include:

This example shows how to create a model with a Stateflow chart that calls two MATLAB functions, meanstats and stdevstats:

  • meanstats calculates the mean of the values in vals.

  • stdevstats calculates a standard deviation for the values in vals.

Build Model

Follow these steps:

  1. Create a new model with the following blocks:

  2. Save the model as call_stats_function_stateflow.

  3. In the model, double-click the Chart block.

  4. Drag two MATLAB functions into the empty chart using this icon from the toolbar:

    A text field with a flashing cursor appears in the middle of each MATLAB function.

  5. Label each function as shown:

    You must label a MATLAB function with its signature. Use the following syntax:

    [return_val1, return_val2,...] = function_name(arg1, arg2,...)
    

    You can specify multiple return values and multiple input arguments, as shown in the syntax. Each return value and input argument can be a scalar, vector, or matrix of values.

    Tip

    For MATLAB functions with only one return value, you can omit the brackets in the signature label.

  6. In the chart, draw a default transition into a terminating junction with this condition action:

    {
    mean = meanstats(invals);
    stdev = stdevstats(invals);
    }
    

    The chart should look something like this:

    Tip

    If the formal arguments of a function signature are scalars, verify that inputs and outputs of function calls follow the rules of scalar expansion. For more information, see Assign Values to All Elements of a Matrix.

  7. In the Modeling tab, under Design Data, select Model Explorer.

  8. In the Model Hierarchy pane of the Model Explorer, select the function meanstats.

    The Contents pane displays the input argument vals and output argument meanout. Both are scalars of type double by default.

  9. Double-click the vals row under the Size column to set the size of vals to 4.

  10. In the Model Hierarchy pane of the Model Explorer, select the function stdevstats and repeat the previous step.

  11. In the Model Hierarchy pane of the Model Explorer, select Chart and add the following data:

    Name

    Scope

    Size

    invals

    Input

    4

    mean

    Output

    Scalar (no change)

    stdev

    Output

    Scalar (no change)

    You should now see the following data in the Model Explorer.

    After you add the data invals, mean, and stdev to the chart, the corresponding input and output ports appear on the Stateflow block in the model.

  12. Connect the Constant and Display blocks to the ports of the Chart block and save the model.

Program MATLAB Functions

To program the functions meanstats and stdevstats, follow these steps:

  1. Open the chart in the model call_stats_function_stateflow.

  2. In the chart, open the function meanstats.

    The function editor appears with the header:

    function meanout = meanstats(vals)

    This header is taken from the function label in the chart. You can edit the header directly in the editor, and your changes appear in the chart after you close the editor.

  3. On the line after the function header, enter:

    %#codegen

    The %#codegen compilation directive helps detect compile-time violations of syntax and semantics in MATLAB functions supported for code generation.

  4. Enter a line space and this comment:

    % Calculates the statistical mean for vals
    
  5. Add the line:

    len = length(vals);
    

    The function length is an example of a built-in MATLAB function that is supported for code generation. You can call this function directly to return the vector length of its argument vals. When you build a simulation target, the function length is implemented with generated C code. Functions supported for code generation appear in Functions and Objects Supported for C/C++ Code Generation (MATLAB Coder).

    The variable len is an example of implicitly declared local data. It has the same size and type as the value assigned to it — the value returned by the function length, a scalar double. To learn more about declaring variables, see Data Definition Basics (MATLAB Coder).

    The MATLAB function treats implicitly declared local data as temporary data, which exists only when the function is called and disappears when the function exits. You can declare local data for a MATLAB function in a chart to be persistent by using the persistent construct.

  6. Enter this line to calculate the value of meanout:

    meanout = avg(vals,len);
    

    The function meanstats stores the mean of vals in the Stateflow data meanout. Because these data are defined for the parent Stateflow chart, you can use them directly in the MATLAB function.

    Two-dimensional arrays with a single row or column of elements are treated as vectors or matrices in MATLAB functions. For example, in meanstats, the argument vals is a four-element vector. You can access the fourth element of this vector with the matrix notation vals(4,1) or the vector notation vals(4).

    The MATLAB function uses the functions avg and sum to compute the value of mean. sum is a function supported for code generation. avg is a local function that you define later. When resolving function names, MATLAB functions in your chart look for local functions first, followed by functions supported for code generation.

    Note

    If you call a function that the MATLAB function cannot resolve as a local function or a function for code generation, you must declare the function to be extrinsic.

  7. Now enter this statement:

    coder.extrinsic('plot');
  8. Enter this line to plot the input values in vals against their vector index.

    plot(vals,'-+');
    

    Recall that you declared plot to be an extrinsic function because it is not supported for code generation. When the MATLAB function encounters an extrinsic function, it sends the call to the MATLAB workspace for execution during simulation.

  9. Now, define the local function avg, as follows:

    function mean = avg(array,size)
    mean = sum(array)/size;
    

    The header for avg defines two arguments, array and size, and a single return value, mean. The local function avg calculates the average of the elements in array by dividing their sum by the value of argument size.

    The complete code for the function meanstats looks like this:

    function meanout = meanstats(vals)
    %#codegen
    
    % Calculates the statistical mean for vals
    
    len = length(vals);
    meanout = avg(vals,len);
    
    coder.extrinsic('plot');
    plot(vals,'-+');
    
    function mean = avg(array,size)
    mean = sum(array)/size;
    
  10. Save the model.

  11. Back in the chart, open the function stdevstats and add code to compute the standard deviation of the values in vals. The complete code should look like this:

    function stdevout = stdevstats(vals)
    %#codegen
    
    % Calculates the standard deviation for vals
    
    len = length(vals);
    stdevout = sqrt(sum(((vals-avg(vals,len)).^2))/len);
    
    function mean = avg(array,size)
    mean = sum(array)/size;
  12. Save the model again.