[math] Fix inconsistent behaviour of TFormula parameter functions

[parth-07]

This PR aims to fix the problem of inconsistency in TFormula::SetParNames() and TFormula::SetParameters() functions as described in #7805.

I have added warning if more values are provided than the actual number of parameters of TFormula object, original behaviour of the function is maintained apart from the added warning.

Some test will fail due to the added warning, please tell if I should remove the warning or resolve the root cause of the warnings ( For instance, TF1 using TFormula::SetParameters() and always passing 11 arguments is one cause of failing tests)

I would also like to slightly modify behaviour of these functions such that value and names of only specified parameters should be changed, current behaviour is to reset the value and names of parameters which aren't specified, but this change may negatively impact programs using these functions. But still, I think this change will make these functions more intuitive. Please tell if I should add this change.

This PR fixes #7805

[lmoneta]

Very nice improvements and I agree it is consistent now.
I think we should then fixes also TF1::SetParNames and TF1::SetParameters to be consistent and since TF1 is probably more used than TFormula itself

[parth-07]

Very nice improvements and I agree it is consistent now.
I think we should then fixes also TF1::SetParNames and TF1::SetParameters to be consistent and since TF1 is probably more used than TFormula itself

Yes agreed, I will make these changes in this PR.

[parth-07]

I think we should then fixes also TF1::SetParNames and TF1::SetParameters to be consistent and since TF1 is probably more used than TFormula itself

TF1::SetParNames and TF1::SetParameters member functions are virtual, and C++ currently do not support virtual function templates. Can you suggest some other way to make these functions generic ? Normal variadic functions can be used, but they are not very easy to work with and maintain, and also do not provide static type checking.

[lmoneta]

I see, I understand the issue with TF1::SetParameters. There is probably no need to have those functions virtual, in ROOT we do not have a derived class re-implementing those, but we cannot say this for user defined classes.
These improvements in TF1 could be then done later in a separate PR

[guitargeek]

Since I still think this is a nice improvement, I resurrected this PR by rebasing the changes on top of ROOT master. Let's see what the CI says now!

[guitargeek]

@phsft-bot build

[phsft-bot]

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac11/noimt, mac12arm/cxx20, windows10/default
How to customize builds

[phsft-bot]

Build failed on mac11/noimt.
See console output.

[github-actions]

Test Results

         5 files           5 suites   1d 5h 46m 49s ⏱️
  2 476 tests   2 476 ✔️ 0 💤 0 ❌
11 647 runs  11 647 ✔️ 0 💤 0 ❌

Results for commit 7a06e35.

[guitargeek]

Hi @parth-07! I have resurrected this PR in order to close the underlying issue.

I simplified your implementation a bit, and also did the same treatment with the variadic template methods to the TF1Parameters class.

I didn't do this for the TF1 though, because variadic templates would break the support for the context menu in the TBrowser. In addition, there is the problem that it's a virtual function as @lmoneta pointed out.

But what is consistent now is that both SetParNames and SetParameters have default argument values that correspond to the parameters or names not changing. For the Parameters, this is encoded by NaN, and for the names it is encoded with empty strings.

[parth-07]

Hi @guitargeek, thank you for resurrecting the pull request.