“I wrote some code for Matlab, and I want to get it running under Octave. Is there anything I should watch out for?”
or alternatively
“I wrote some code in Octave, and want to share it with Matlab users. Is there anything I should watch out for?”
which is not quite the same thing. There are still a number of differences between Octave and Matlab, however in general differences between the two are considered as bugs. Octave might consider that the bug is in Matlab and do nothing about it, but generally functionality is almost identical. If you find a difference between Octave behavior and Matlab, then you should send a description of this difference (with code illustrating the difference, if possible) to bug@octave.org.
Furthermore, Octave adds a few syntactical extensions to Matlab that might cause some issues when exchanging files between Matlab and Octave users. As both Octave and Matlab are under constant development the information in this section is subject to change at anytime.
You should also look at the page http://octave.sourceforge.net/packages.html and http://octave.sourceforge.net/doc/ that has a function reference that is up to date. You can use this function reference to see the number of octave function that are available and their Matlab compatibility.
The major differences between Octave 3.2.N and Matlab R2008a are:
Octave doesn't yet have nested functions. That is
function y = foo (x) y = bar(x) function y = bar (x) y = ...; end end
There was discussion in Octave of having these even prior to Matlab, and the decision was made not to have these in Octave at the time for compatibility. The above written with sub-functions functions would be
function y = foo (x) y = bar(x) end function y = bar (x) y = ...; end
Now that Matlab has recently introduced nested functions, Octave will probably have them soon as well. Until then nested functions in Octave are treated as sub-functions with the same scoping rules as sub-functions.
The authors of Octave consider the nested function scoping rules of Matlab to be more problems than they are worth as they introduce diffiult to find bugs as inadvertantly modifying a variable in a nested function that is also used in the parent is particularly easy.
condest
or to do with specific missing Octave functionality
(gui, dll, java, activex, dde, web, and serial functions). Some of the
core functions have limitations that aren't in the Matlab
version. For example the sprandn
function can not force a
particular condition number for the matrix like Matlab can.
patch
function is currently limited to 2-D patches, due to an underlying
limitation in gnuplot.
Although Octave can load inline abd function handles saved by Matlab, it can not yet save them.
Finally, Some multi-byte unicode characters aren't yet treated in mat-files.
http://www.cae.wisc.edu/pipermail/octave-maintainers/2007-January/001685.html
for more details
&
and |
operators in Matlab short-circuit when
included in an if statemant and not otherwise. In Octave only the
&&
and ||
short circuit. Note that this means that
if (a | b) ... end
and
t = a | b; if t ... end
are different in Matlab. This is really a Matlab bug, but there is too much code out there that relies on this behavior to change it. Prefer the || and && operators in if statements if possible.
Note that the difference is also significant when either argument is a function with side effects or if the first argument is a scalar and the second argument is an empty matrix. For example, note the difference between
t = 1 | []; ## results in [], so... if (t) 1, end ## in if ([]), this is false.
and
if (1 | []) 1, end ## short circuits so condition is true.
Another case that is documented in the Matlab manuals is that
t = [1, 1] | [1, 2, 3]; ## error if ([1, 1] | [1, 2, 3]) 1, end ## OK
Also Matlab requires the operands of && and || to be scalar values but Octave does not (it just applies the rule that for an operand to be considered true, every element of the object must be nonzero or logically true).
Finally, note the inconsistence of thinking of the condition of an if
statement as being equivalent to all(X(:))
when X is a
matrix. This is true for all cases EXCEPT empty matrices:
if ([0, 1]) == if (all ([0, 1])) ==> i.e., condition is false. if ([1, 1]) == if (all ([1, 1])) ==> i.e., condition is true.
However,
if ([]) != if (all ([]))
because samp ([]) == 1
(because, despite the name, it is really
returning true if none of the elements of the matrix are zero, and since
there are no elements, well, none of them are zero). But, somewhere
along the line, someone decided that if ([])
should be false.
Mathworks probably thought it just looks wrong to have []
be true
in this context even if you can use logical gymnastics to convince
yourself that "all" the elements of a matrix that doesn't actually have
any elements are nonzero. Octave however duplicates this behavior for if
statements containing empty matrices.
Matlab's solvers as used by the operators mldivide (\) and mrdivide (/), use a different approach than Octave's in the case of singular, under-, or over-determined matrices. In the case of a singular matrix, Matlab returns the result given by the LU decomposition, even though the underlying solver has flagged the result as erroneous. Octave has made the choice of falling back to a minimum norm solution of matrices that have been flagged as singular which arguably is a better result for these cases.
In the case of under- or over-determined matrices, Octave continues to use a minimum norm solution, whereas Matlab uses an approach that is equivalent to
function x = mldivide (A, b) [Q, R, E] = qr(A); x = [A \ b, E(:, 1:m) * (R(:, 1:m) \ (Q' * b))] end
While this approach is certainly faster and uses less memory than Octave's minimum norm approach, this approach seems to be inferior in other ways.
A numerical question arises: how big can the null space component become, relative to the minimum-norm solution? Can it be nicely bounded, or can it be arbitrarily big? Consider this example:
m = 10; n = 10000; A = ones(m, n) + 1e-6 * randn(m,n); b = ones(m, 1) + 1e-6 * randn(m,1); norm(A \ b)
while Octave's minimum-norm values are around 3e-2, Matlab's results are 50-times larger. For another issue, try this code:
m = 5; n = 100; j = floor(m * rand(1, n)) + 1; b = ones(m, 1); A = zeros(m, n); A(sub2ind(size(A),j,1:n)) = 1; x = A \ b; [dummy,p] = sort(rand(1,n)); y = A(:,p)\b; norm(x(p)-y)
It shows that unlike in Octave, mldivide in Matlab is not invariant with respect to column permutations. If there are multiple columns of the same norm, permuting columns of the matrix gets you different result than permuting the solution vector. This will surprise many users.
Since the mldivide (\) and mrdivide (/) operators are often part of a more complex expression, where there is no room to react to warnings or flags, it should prefer intelligence (robustness) to speed, and so the Octave developers are firmly of the opinion that Octave's approach for singular, under- and over-determined matrices is a better choice that Matlab's
try
/catch
combined with ‘rethrow (lasterror ())’ in
Matlab, however rethrow and lasterror are only available in Octave 2.9.10 and later.
Note that using try
/catch
combined with ‘rethrow
(lasterror ())’ can not guarantee that global variables will be
correctly reset, as it won't catch user interrupts with Ctrl-C. For
example
global a a = 1; try _a = a; a = 2 while true end catch fprintf ('caught interrupt\n'); a = _a; rethrow (lasterror()); end
compared to
global a a = 1; unwind_protect _a = a; a = 2 while true end unwind_protect_cleanup fprintf ('caught interrupt\n'); a = _a; end
Typing Ctrl-C in the first case returns the user directly to the prompt, and the variable "a" is not reset to the saved value. In the second case the variable "a" is reset correctly. Therefore Matlab gives no save way of temporarily changing global variables.
sin(x)(1:10);
for example is perfectly valid
in Octave but not Matlab. To do the same in Matlab you must do
y = sin(x); y = y([1:10]);
\n
(newline), \t
(tab), etc are
interpreted in double quoted strings but not single quoted strings. This
difference is important on Windows platforms where the "\" character is
used in path names, and so single quoted strings should be used in
paths. Matlab doesn't have double quoted strings and so they should
be avoided if the code will be transfered to a Matlab user.