More convenience when constructing TGraphs

[hageboeck]

Explain what you would like to see improved

I find this kind of plotting code too cumbersome, it only takes double* and sizes:

  const double x[] = {0, 1, 2};
  const double y[] = {1, 2, 3};
  auto graph = new TGraph(3, x, y);

Optional: share how it could be improved

Provide an overload that constructs from iterable collections and/or std::initializer_list, maybe.

  auto graph = new TGraph({0, 1, 2}, {1, 2, 3);

When the collections don't have the same length, throw std::invalid_argument.

[eguiraud]

Assigning both @couet and @Axel-Naumann because the ticket is about ROOT6 graphics but the improvement might be added to ROOT7 graphics.

[couet]

Nice idea. Except I do not have any idea how to make a such constructor. That should be of course extended to TGraphErrors, TGraphAssymError .etc .. TGraph2D etc .. I am happy to do it .. can you point some example/code I can take inspiration from ?

[Axel-Naumann]

Can't we simply take vector<double>s instead of int size, double*?

[couet]

@hageboeck is Axel suggestion fine for you ?

[hageboeck]

I think that is a good compromise.

Unfortunately, you cannot use the initializer_list syntax as I was hoping: https://godbolt.org/z/4szszTv4v

It works if you call it a bit more explicitly, though:

TGraph gr(std::vector<double>{0., 1.}, {0., 1.});

@Axel-Naumann good enough or can we do better?

[Axel-Naumann]

Instead of modifying TGraph, throw in

   TVectorT(std::initializer_list<Element> &&il);

and it will work.

[hageboeck]

It doesn't work, unfortunately, because overload resolution cannot disambiguate between the TGraph(TVectorD and TGraph(TVectorF cases. Nevertheless, with #9968 merged, one can write:

TGraph gr(TVectorD{1., 2., 3.}, {1., 2., 3.});

It probably won't get much better, so closing this here.

[github-actions[bot]]

Hi @hageboeck, @couet,

It appears this issue is closed, but wasn't yet added to a project. Please add upcoming versions that will include the fix, or 'not applicable' otherwise.

Sincerely, :robot:

[guitargeek]

Re-opening, since #9968 was not merged after all. This issue should not have been closed before actually merging the PR.

[guitargeek]

Talking again about the proposed interface:

  auto graph = new TGraph({0, 1, 2}, {1, 2, 3);

I mean it looks nice and is probably useful for unit tests, but is it really helping our users?

If you have data that you want to plot, when do you hardcode it in initializer lists instead of reading it from files or getting it programmatically as the output of your analysis?