Release 11 Enhancements (Release 11 New Features)

Release 11 New Features

MATLAB Language Enhancements

Links to Function Descriptions
If you are reading this in HTML form, clicking on the function name in the tables summarizing functions (or in highlighted links in the text) displays the reference documentation for that function. Use your browser's Back button to return to this document.

Support for Integer Data Types

New Integer Array Classes

Version 5.3 extends the support for integer data types to include several additional array classes that store integer data types. MATLAB 5.3 adds to the existing uint8 class 8-, 16-, and 32-bit signed and unsigned integer array classe; for example, int16 for signed 16-bit integers.

These classes are primarily meant to store integer values. Most operations that manipulate arrays without changing their elements are defined for these data types (examples are reshape, size, the logical and relational operators, subscripted assignment, and subscripted reference). In addition, MATLAB supports the find function for integer arrays, but the returned array is of class double. No math operations except for sum are defined for these classes, since such operations are ambiguous on the boundary of the set (for example, they could wrap or truncate there).

sum Function Now Supports All Integer Types

The sum function can now be used with all of the integer data types supported by MATLAB. Previously, sum only worked with uint8 data type. With this release, sum also supports the int8, int16, and int32 data types and the uint16 and uint32 unsigned data types. When the sum function is used with integer data types, the value returned by sum is of class double.

File I/O Enhancements

User-Extensible File Opening Function

The new open function is a user-extensible function that provides an interface to file open operations. Default behavior is provided for these standard MATLAB file types: Handle Graphics figure files, M-files, model files, and P-files. You can extend the interface to include other file types and to override the default behavior for the standard files.

Reading Data From a Uniformly Formatted File

The new textread function provides easy reading of data from a uniformly formatted file, such as a comma- or tab-delimitered file, into MATLAB variables. The formatted file can contain both numbers and strings. The data is converted using the types and delimiters you specify.

Enhancements to dlmread

The dlmread function was enhanced to significantly improve function performance.

Also, to use the range argument, use this new calling sequence:

M = dlmread (filename,delimiter,range)

Saving MATLAB Figures or Models

The new saveas function saves a MATLAB figure or model to a file using the specified file format.

Support for Single Precision Data

MATLAB now supports single precision data, solely as a storage format. Using the fread and fwrite functions, you can read input files containing single precision data and write data to files in single precision format. To convert data to single precision format, use the single function.

Note
MATLAB does not support operations on single precision data other than conversion. In particular, mathematical operations are not supported. Currently, single precision data is primarily used in MATLAB with the Hierarchical Data Format (HDF) development tools.

String Conversion

The new str2double function converts a character string to a double precision value. The character string should contain the ASCII representation of a scalar value (real or complex). Use of str2double is recommended over the str2num function.

The new texlabel function produces the TeX format from a character string.

Constructing Complex Data

The new complex function constructs complex data from real and imaginary parts.

pause Accepts Fractions of Seconds

The pause function now accepts a fractional number of seconds as an input argument. This means that when you use the calling sequence pause(n), n can be any real number. Previously, n was restricted to integer values.

Enhancements to quit

The quit function has been enhanced to run the script finish.m, if finish.m exists anywhere on the MATLAB path. finish.m is a file you create that contains commands you want to run when MATLAB terminates (i.e., you use quit, exit, or click on the X button to close the window on the PC).

Two sample files illustrating what you could put in finish.m are provided in
/toolbox/local:

finishsav.m - saves the workspace to a MAT-file when MATLAB quits
finishdlg.m - displays a dialog box allowing you to cancel quitting

You can also cancel quitting from within the finish.m file by using quit cancel.

Y2K Support

All versions of The MathWorks, Inc.'s software products have always represented data in 8-byte double-precision floating-point data type form. This ensures that our products will be able to function properly in your environment in the year 2000 and beyond, including leap years, without any adjustments or action required on your part.

Date Functions Calling Sequence Change

With MATLAB 5.3, the date functions datenum, datestr, and datevec include a new calling sequence that allows a pivot year specification to override the default. For example, here's the new calling sequence for datevec:

[...] = datevec(t, pivotyear)

This new call uses the pivot year instead of the current year minus 50 years.

See "Upgrading From MATLAB 5.2 to MATLAB 5.3" in Chapter 4 for details about possible changes you might want to make to existing applications.

Operating on Cell Arrays

The new cellfun function is a multipurpose function that performs common operations on the elements of cell arrays, including: isreal, isempty, islogical, length, ndims, prodofsize, size, and isclass.

Diagonal Concatenation

The new blkdiag function concatenates input arguments diagonally in a matrix.

Enhancements to Sparse Matrix Operations

The iterative methods that operate on sparse matrices have been enhanced to accept a function as an argument. For example,

x = pcg(A,b)

solves the system of linear equations A*x = b for x. When A is not explicitly available as a matrix, you can express A as an operator that accepts vector input x and returns the matrix-vector product A*x. This operator can be the name of an M-file, a string expression, or an inline object.

Here is a simple example.

function y = afun(x)
% AFUN(X) returns A*X, where A = diag(7*ones(n,1)) is diagonal.
y = 7*x; % y = A*x

This example calls pcg using the function afun in place of the matrix A = diag(7*ones(n,1)).

x = pcg('afun',b);

Numerical Analysis

Enhancements to Differential Equation Solvers

Mass Matrix Support. In previous versions of the ODE suite, only the stiff solvers could handle problems of the form M*y'=F(t,y) with a mass matrix M. Since it is often convenient to solve a problem in this mass matrix form, we extended all of the solvers of the ODE suite to solve problems M*y'=F(t,y) with a mass matrix M that is nonsingular and (usually) sparse. In addition, for all but one solver, the mass matrix M can now be both time- and state-dependent, M(t,y). (As before, the ode23s solver allows only constant M.)

The Mass property of odeset has been enhanced to include new possible values; the old values are still available. For examples, see the M-file help for fem1ode, fem2ode, or batonode.

Singular Mass Matrices and Differential-Algebraic Equations. If the mass matrix is singular, then M*y' = F(t,y) is a differential-algebraic equation (DAE). DAEs have solutions only when y0 is consistent, that is, when there is a vector yp0 such that M(t0)*yp0 = F(t0,y0). The two solvers ode15s and ode23t can solve DAEs of index 1 provided that M is not state-dependent and y0 is sufficiently close to being consistent.

If there is a mass matrix, you can use odeset to set MassSingular to 'yes', 'no', or 'maybe'. The default of 'maybe' causes the solver to test whether the problem is a DAE. If it is, the solver treats y0 as an initial estimate, attempts to compute consistent initial conditions that are close to y0, and proceeds to solve the problem. When solving DAEs, it is advantageous to formulate the problem so that M is diagonal (a semi-explicit DAE). For examples, see the M-file help for hb1dae or amp1dae.

Changes to Function Functions

These MATLAB function functions have new names and calling sequences to support new functionality.


Old Function Name	New Function Name
`fmin`	`fminbnd`
`fmins`	`fminsearch`

Note that if you have older M-files that use the old names and calling sequences, these calls will generally continue to work. However, the older functions may be removed from MATLAB in future releases, so it is a good idea to revise your code now to use the new names and calling sequences.

Changes to Least Squares Equation Solver

The name of the nnls (nonnegative least squares) function was changed to lsqnonneg, and its calling sequences have changed as well. These changes have been made to support new functionality.

As noted above, if you have older M-files that use the old names and calling sequences, these calls will generally continue to work.

New Mechanism for Setting Optimization Parameters

MATLAB now has a new mechanism for setting parameters used by the optimization functions. The options argument for these functions now takes a structure containing the parameters to set, rather than a vector. This structure is created and modified by a new function, optimset. The new optimget function extracts the value of a parameter from an options structure.

This change affects the new optimization functions only (fminbnd, fminsearch, and lsqnonneg). The older functions (fmin, fmins, and nnls) still use the vector returned by the foptions function.

Note that not all of the parameters set with optimset apply to every function. Many of the parameters apply only to functions in the Optimization Toolbox.

Changes to cholinc Function

In the cholinc function's handling of the Cholesky-Infinity factorization, the input matrix is assumed to be positive semi-definite, so negative pivots are treated as if they were 0. Because of this change, the functional form that includes a second output argument p (shown below) is obsolete:

[R,p] = cholinc(X,'inf')

Programming Enhancements

New evalc Function

The new evalc function is an extension of eval. Calling evalc executes a MATLAB expression and captures any output that would be written to the MATLAB command window in a character array output argument.

New symvar Function

The new symvar function searches a MATLAB expression for symbolic variables and returns the names of the variables in a cell array of strings. The identifiers i, j, pi, and other MATLAB constants and function names are ignored.

Enhancements to inline

The inline function was improved to better recognize symbolic variable names in a function expression. In addition, multiple variable names are now located.

Enhancements to MATLAB Object-Oriented Programming

Several functions have been added or enhanced to provide additional support for MATLAB object-oriented programming.

Loading and Saving Objects. The new loadobj function is an overloadable function called by the load command when reading objects from a MAT file into the MATLAB workspace. To define load behavior for user objects, create a loadobj function in the associated class directory.

The new saveobj function is an overloadable function called by the save command when writing objects from the MATLAB workspace to a MAT file. To define save behavior for user objects, create a saveobj function in the associated class directory.

Enhancements to end Statement. You can overload the end statement for indexing a user object. To do this, write a method end.m in the class directory. The end method must have the following calling sequence

end(myobj,K,N)

where myobj is the object, K is the index for which you are using the end syntax, and N is the number of indices in the indexing expression.

Use clear classes to Clear the Class Definition Table. To clear the class definition table, use

clear classes

This is useful when, during a MATLAB session, you change the way a class is defined.

You should no longer use

clear all

to clear the class definition table.

Application Program Interface (API) Enhancements

ActiveX Support Enhanced

The ActiveX support enhancements for MATLAB 5.3 include

Support for interactively using get to return a list of properties and send to get a list of all events for an interface
Enhanced data conversion
Improved event/callback management

The support for ActiveX controls is described in the Application Program Interface Guide, in Chapter 7.

MATLAB 5.0 Data Types Supported in the MATLAB 5.3 API Engine

All MATLAB 5.3 data types, including those introduced in MATLAB 5.0 (cell arrays, multidimensional arrays, and structures) are now supported in the MATLAB 5.3 API Engine.

Exploratory MATLAB Java Interface

Intended as a Prototype for Soliciting Your Feedback

Release 11 includes an exploratory MATLAB Java interface. Based on your feedback and additional development and testing efforts, this MATLAB Java interface may be refined and expanded in future releases.

Caution Do not use this exploratory version of the MATLAB Java interface for production-level code. This interface will almost certainly change in future releases, and The MathWorks does not commit to ensuring that code written using this exploratory version of the interface will work with future versions of the interface or MATLAB.

What You Can Do with the MATLAB Java Interface

You can use the MATLAB Java interface to:

Create Java objects in the MATLAB workspace
Invoke Java methods on objects in the MATLAB workspace
Invoke Java static methods within MATLAB
Pass Java objects to MATLAB M-file functions and built-in functions
Add Java code to toolboxes
List the methods defined by a Java class

How to Start the MATLAB Java Interface

To start the MATLAB Java interface, at the command line type:

java on

[ Previous | Next ]