Make Minuit.matrix and Minuit.print_matrix consistent about iminuit HOT 10 CLOSED

I think I did matrix that way to keep it compatible with pyminuit which I don't think it's a good reason anymore though...

from iminuit.

@piti118 How would you like to handle this?

I think my preference would be to be explicit as written above and to either remove directly or deprecate matrix and print_matrix. If we do e.g. #124 we could bump the iminuit version number to 2.0 and backwards-incompatible API changes might be OK?

(I plan to work on iminuit in the coming weeks ... reading the Minuit papers and Cython code at the moment ... )

from iminuit.

@piti118 At #86 (comment) you said it might be OK to break pyminuit compatibility to clean this up two years ago, but then I didn't make the change. Now in #153 (comment) you said it's better not to break compatibility.

What's your decision on this?
(I'm +1 to break compatibility.)

The two options are:

• Implement these improvements / cleanups now and break compatibility a little, call the next release 2.0.
• Keep compatibility for now and label this issue 2.0 which might happen at some point in the future.

from iminuit.

Hahahahahaha. I disagree with my younger self.

If this function ever got in the way of making change, then we will remove it. But right now, it doesn't get in the way of anything. So we could just mark it deprecated and keep it as is.

from iminuit.

Oh, it does get in the way..... so let's get rid of it.

from iminuit.

I disagree with my younger self.

It's normal to get more conservative as time goes by. 😄

from iminuit.

I think we can change print_matrix() to show the covariance matrix without breaking any user code. Then it would be consistent with matrix(). This seems like the simplest solution. Any objections?

from iminuit.

OK to change the print if this gives more consistency.

from iminuit.

I gave this more thought: maybe we should live with the inconsistency and keep things the way they are. When you do "print_matrix()", you most likely want to see the correlation. The correlation is easy to mentally process, the covariance matrix is not easy to understand. The covariance matrix is useful when you do further calculations (error propagation), you don't want to print in an interactive session.

from iminuit.

ok

from iminuit.