Error in eigs when using Matlab 2017b?

Hi,

When I'm using eigs(A,B,K) in Matlab 2017b and receiving the following error:

Error: File: eigs.m Line: 1442 Column: 5

"c" previously appeared to be used as a function or command, conflicting with its use here as the name of a variable.

A possible cause of this error is that you forgot to initialize the variable, or you have initialized it implicitly using load or eval.

The code works with Matlab 2016b. What could be the problem when using Matlab 2017b?

Best regards, Emil

[Walter Roberson]

I do not get that error when I test.

I suggest you put in

dbstop if error

and run the code. When it stops, ask

which -all c

to find out where c is being found. Perhaps you have a c.m of your own that is somehow interfering.

[the cyclist]

I agree with Walter that it is likely a function in file c.m, in the directory from which you are calling eigs. (I have found that just being in your path is ok, as long as it is not in your current working directory.)

Strangely, using "dbstop if error" (which is an excellent suggestion in general!) did not seem to help here. When I created a function file c.m in my current directory, and then called

eigs(rand(3),rand(3),2)

execution was not halted, and I was returned to the command line with the same error as Emil.

[the cyclist]

This is a bug, right? There should be no problem with having such a file.

I'm finding some version-dependent behavior here, too. I do not see this behavior in R2017b.

[Emil Aggestam]

I think it is a bug. I see no reason why you shall not be able to have a file namned c.m.

[Christine Tobler]

In general, I'd say it's a good idea to not have one-letter function names. However, I don't see why this should stop EIGS from working here. I'll take a look, this indeed sounds like a bug.

[Walter Roberson]

Possibly

dbstop if caught error

is needed to locate the problem.

[the cyclist]

That didn't halt execution either (at least for me).

[the cyclist]

Addendum: I do see the bug in R2017b. I did the wrong test earlier.

Answer 1

[Emil Aggestam]

I had a file named c.m that caused the problem. Thank you!

Answer 2

[Christine Tobler]

Thank you for pointing out this bug. There is now a bug report available where this can be tracked, and it will be fixed in a future version.

To address a discussion in the comments, the reason that the debugger wasn't helping is that this error is coming from the initial parsing of eigs.m, before the function is entered (similar to what happens for an "if" without an "end": the whole file is seen as inconsistent).