Problems with axis sizing for mplhep v0.2.10+

Cross post of https://github.com/matthewfeickert/heputils/issues/41

---

In PR #239 the ATLAS style figure defaults were updated and this change was released in mplhep v0.2.10. This is apparently having some extreme effects (c.f. https://github.com/matthewfeickert/heputils/issues/41 for more examples)

Reproducible Example

In a clean Python 3 virtual environment the following

# example.py
import uproot4 as uproot
from heputils import convert
import mplhep
import matplotlib.pyplot as plt


def main():
    root_file = uproot.open("example.root")
    root_hists = [convert.uproot_to_hist(root_file[key]) for key in root_file.keys()]
    data_hist = root_hists[-1]
    simulation_hists = root_hists[:-1]

    mplhep.set_style("ATLAS")
    fig, ax = plt.subplots()
    mplhep.histplot(simulation_hists[0], histtype="fill", ax=ax)
    ax.set_xlabel(r"$X$ Mass [$\mathrm{GeV}$]")
    ax.set_ylabel("Count")

    fig.savefig(f"mplhep_v{mplhep.__version__}_example.png")


if __name__ == "__main__":
    main()

$ python -m pip install --quiet heputils
$ pip list
Package         Version
--------------- ---------
awkward         1.0.2
awkward0        0.15.2
awkward1        1.0.0
boost-histogram 0.11.1
cachetools      4.2.0
certifi         2020.12.5
chardet         4.0.0
click           7.1.2
cycler          0.10.0
future          0.18.2
heputils        0.0.4
hist            2.0.1
histoprint      1.5.2
idna            2.10
kiwisolver      1.3.1
matplotlib      3.3.3
mplhep          0.2.10
numpy           1.19.5
packaging       20.8
Pillow          8.1.0
pip             20.3.3
pyparsing       2.4.7
python-dateutil 2.8.1
requests        2.25.1
scipy           1.6.0
setuptools      51.1.2
six             1.15.0
uncertainties   3.1.5
uproot          4.0.0
uproot3         3.14.2
uproot3-methods 0.10.0
uproot4         4.0.0
urllib3         1.26.2
wheel           0.36.2
$ curl -sL https://raw.githubusercontent.com/matthewfeickert/heputils/master/tests/example_files.py | python
$ file example.root 
example.root: ROOT file Version 1061800 (Compression: 256)
$ python example.py
$ pip install --upgrade "mplhep<0.2.10"
$ python -m pip list | grep mplhep
mplhep                 0.2.9
$ python example.py

produces

mplhep_v0.2.9_example.png	mplhep_v0.2.10_example.png

cc @andrzejnovak and @kratsg

[andrzejnovak]

Does this leave anything unresolved on mplhep side? The sci notation effect is unfortunate, but I don’t see what we could do about it. In fact, I am not even sure what the correct solution is, move the xlabel toward the center? Either way I think this needs to be resolved on the user side.

[matthewfeickert]

what changed wrt the horiztonal axis buffers

It seems the answer is

https://github.com/scikit-hep/mplhep/blob/e8d592e2e52a974f0730d2b645cb95b8a86c8398/src/mplhep/styles/atlas.py#L45

but even if setting "axes.autolimit_mode": "data" (matplotlib’s default) things aren’t looking like the ATLAS style used in JHEP 06 (2018) 107.

The closest I can get if I write some very crappy PyROOT (and feel embarrassed at how bad my ROOT skills are now) with AtlasStyle.C and AtlasStyle.h in the directory

# example_pyroot.py
from ROOT import TCanvas, TColor, TFile, gROOT, kFALSE, kTRUE

gROOT.LoadMacro("AtlasStyle.C")
gROOT.SetBatch(kTRUE)


def main():
    read_file = TFile.Open("example.root", "READ")
    hist_ttbar = read_file.Get("ttbar")

    canvas = TCanvas("")
    canvas.cd()

    hist_ttbar.SetStats(kFALSE)
    fill_color = TColor.GetColor("#1f77b4")
    hist_ttbar.SetFillColor(fill_color)
    hist_ttbar.GetXaxis().SetTitle("X Mass [GeV]")
    hist_ttbar.GetYaxis().SetTitle("Count")

    hist_ttbar.Draw("HIST")
    canvas.SaveAs("pyroot_example.png")


if __name__ == "__main__":
    main()

is if I amend ATLAS style to have "axes.xmargin": 0.0

# example.py
import matplotlib.pyplot as plt
import mplhep
import uproot
from matplotlib.pyplot import style


def main():
    root_file = uproot.open("example.root")
    root_hists = [root_file[key].to_hist() for key in root_file.keys()]
    data_hist = root_hists[-1]
    simulation_hists = root_hists[:-1]

    mplhep.set_style("ATLAS")
    style.use({"axes.xmargin": 0.0})

    fig, ax = plt.subplots()
    mplhep.histplot(simulation_hists[0], histtype="fill", ax=ax)
    ax.set_xlabel(r"$X$ Mass [$\mathrm{GeV}$]")
    ax.set_ylabel("Count")

    fig.savefig(f"mplhep_v{mplhep.__version__}_example_revised.png")


if __name__ == "__main__":
    main()

pyroot_example.png	mplhep_v0.2.17_example_revised.png

Thoughts? Am I doing something wrong?