At work they recently moved me to a new project. It is a Matlab-based
data analysis thing. I haven't really touched Matlab in over a year
(the last time I used Matlab at work), and, instead, use GNU Octave at
home when the language is appropriate. I got so used to Octave that I
found a pretty critical feature missing from Matlab's implementation:
treat an expression as if it were of the type of its output.
Let's say we want to index into the result of a function. Take, for
example, the magic square function,
magic(). This spits
of the given size. In Octave we can generate a 4x4 magic square and
chop out the middle 2x2 portion in one line.
Or more possibly clearly,
Try this in Matlab and you will get a big, fat error. You have to
assign the magic square to a temporary variable to do the same
thing. I kept trying to do this sort of thing in Matlab and was
thinking to myself, "I know I can do this somehow!". Nope, I
was just used to having Octave.
Where this really shows is when you want to reshape a matrix into a
nice, simple vector. If you have a matrix
M and want to
count the number of NaN's it has, you can't just apply
sum() function over
isnan() because it
only does sums of columns. You can get around this with a special
So, to sum all elements in
In Octave, to count NaN's with
Again, Matlab won't let you index the result of
directly. Stupid. I guess the Matlab way to do this is to
Every language I can think of handles this properly. C, C++, Perl,
Ruby, etc. It is strange that Matlab itself doesn't have it. Score one
more for Octave.