[WIP] Add support for the Univeral Histogram Interface (UHI)

[LukasBreitwieser]

Current Status

Working implementation for TH1 including basic tests

Access

v = h[2]
v = h[ROOT.underflow]        # return the bin contents of underflow
v = h[ROOT.overflow]4        # return the bin contents of overflow
v = h[ROOT.loc(1.4)]         # return the bin content of the bin that stores the value 1.4 (data coordinates) 
v = h[ROOT.loc(1.4) + 1]     # return the bin content of the next bin that stores the value 1.4 (data coordinates) 
v = h[ROOT.loc(1.4) - 1]     # return the bin content of the previous bin that stores the value 1.4 (data coordinates) 

Setting

h[2]
h[ROOT.underflow] = 3         # set the content of the underflow bin
h[ROOT.overflow] = 4          # set the content of the overflow bin
h[ROOT.loc(1.4)] = 5.0        # set the bin contents in data coordinates 
h[ROOT.loc(1.4) + 1] = 7.0
h[ROOT.loc(1.4) - 1] = 9.0

Slicing, rebinning, reduction

h == h[:]             # Slice over everything
h2 = h[::rebin(2)]    # Modification operations (rebin)
v2 = h[::sum]         # Projection operations # (name may change)

Unsupported by UHI, and therefore raising exceptions:

h[1.0]                # Indexing with a float is disallowed
v2 = h[::sum]         # Projection operations # (name may change)

Currently not supported due to missing slicing funcionality on the cpp side

h2 = h[a:b]           # Slice of histogram (includes flow bins)
h2 = h[:b]            # Leaving out endpoints is okay

This applies also to derivatives, like:

If slicing is implemented, the following list will work too.

h2 = h[loc(v):]       # Slices can be in data coordinates, too
h2 = h[a:b:rebin(2)]  # Modifications can combine with slices
v2 = h[a:b:sum]       # Adding endpoints to projection operations
v2 = h[0:len:sum]     #   removes under or overflow from the calculation
h2 = h[v, a:b]        #   A single value v is like v:v+1:sum
h2 = h[a:b, ...]      # Ellipsis work just like normal numpy

TODO

• Generalize to multiple dimensions
• Implement dictionary of slices on specific axes

h[{0: slice(None, None, bh.rebin(2))}] # rebin axis 0 by two

• Implement Ellipsis

h[...] = array(...)

• Implement

h[:] = np.ones(10)

• Dig deeper to clarify some questions about the API. Open questions like: When is a copy returned, when are the modifications done in-place? ...
• Add more error checks
• Improve testing coverage