Help to understand a strange behaviour

WTF is the following: (noticed ocationally)

>> clear

>> sind(30)

ans =

   -0.8660

>> whos

Name      Size            Bytes  Class     Attributes
ans       1x1                 8  double

after restart:

>> sind(30)

ans = 0.5000

(R2010b)

[Geoff]

That first value is sind(300).

[Ivan]

How can it be?

[Ivan]

>> sind(300)

ans =

-0.8660

>> sind(30)

ans =

-0.8660

[Geoff]

What do you get when you plot?

x = -720:720; plot(x, sind(x));

[Ivan]

plot looks like normal, but drops to -1 at -720:

>> sind(-720)

ans =

-1

Answer 1

[Jan]

I asked Google for "Matlab bug sind" and found:

See: Answers: bug-in-matlab-2010b-trig-functions.

My conclusion: Ask MathWorks for a bugfix of your version. It is easy to insert a workaround as M-file in your toolbox folder:

function Y = sind(X)
Y = sin(X * 0.017453292519943295769236954843959);

Answer 2

[Jan]

Does

which sind -all

reveal any unexpected details?

[Ivan]

no. i think this is normal:

>> sind(30)

ans =

-0.8660

>> which sind -all

built-in (C:\MATLAB\toolbox\matlab\elfun\@double\sind) % double method

built-in (C:\MATLAB\toolbox\matlab\elfun\@single\sind) % single method

C:\MATLAB\toolbox\distcomp\parallel\@codistributed\sind.m % codistributed method

[Ivan]

The two contiguous commands give:

>> sind(30)

ans =

0.500000000000000

>> sind(30)

ans =

-0.866025403784439

Answer 3

[Ivan]

Thank you, Jan. Yes, i'm now using sin with argument in radians and that's ok. But this issue made me suspicious about all my results ever got with MatLab... Its very sad (

[Jan]

I agree that such bugs reduce the confidence. Therefore it would be great great great, if MathWorks offers a similar service as most other programs: Automatic download of patchs and bugfixes. Even a dynamically created list of known bugs would be helpful, while the bugs list at http://www.mathworks.com/support/bugreports/ is still too hidden for standard users. In addition it is weak: I do not find a report when I'm searching for "sind" on "all releases". Is it possible that this bug has not been reported?

It seems like this bug is either not reproducible or there is no or a not sufficiently powerful unittest for the elementary functions.

Please contact the technical support. The more votes they get the better for the users and the company.