GitHub - scikit-hep/hist: Histogramming for analysis powered by boost-histogram (original) (raw)

Hist

Hist is an analyst-friendly front-end forboost-histogram, designed for Python 3.8+ (3.6-3.7 users get older versions). See what's new.

Installation

You can install this library from PyPI with pip:

python3 -m pip install "hist[plot,fit]"

If you do not need the plotting features, you can skip the [plot] and/or[fit] extras. [fit] is not currently supported in WebAssembly.

Features

Hist currently provides everything boost-histogram provides, and the following enhancements:

• Hist augments axes with names:
  • name= is a unique label describing each axis.
  • label= is an optional string that is used in plotting (defaults to nameif not provided).
  • Indexing, projection, and more support named axes.
  • Experimental NamedHist is a Hist that disables most forms of positional access, forcing users to use only names.
• The Hist class augments bh.Histogram with simpler construction:
  • flow=False is a fast way to turn off flow for the axes on construction.
  • Storages can be given by string.
  • storage= can be omitted, strings and storages can be positional.
  • data= can initialize a histogram with existing data.
  • Hist.from_columns can be used to initialize with a DataFrame or dict.
  • You can cast back and forth with boost-histogram (or any other extensions).
• Hist support QuickConstruct, an import-free construction system that does not require extra imports:
  • Use Hist.new.<axis>().<axis>().<storage>().
  • Axes names can be full (Regular) or short (Reg).
  • Histogram arguments (like data=) can go in the storage.
• Extended Histogram features:
  • Direct support for .name and .label, like axes.
  • .density() computes the density as an array.
  • .profile(remove_ax) can convert a ND COUNT histogram into a (N-1)D MEAN histogram.
  • .sort(axis) supports sorting a histogram by a categorical axis. Optionally takes a function to sort by.
  • .fill_flattened(...) will flatten and fill, including support for AwkwardArray.
  • .integrate(...), which takes the opposite arguments as .project.
• Hist implements UHI+; an extension to the UHI (Unified Histogram Indexing) system designed for import-free interactivity:
  • Uses j suffix to switch to data coordinates in access or slices.
  • Uses j suffix on slices to rebin.
  • Strings can be used directly to index into string category axes.
• Quick plotting routines encourage exploration:
  • .plot() provides 1D and 2D plots (or use plot1d(), plot2d())
  • .plot2d_full() shows 1D projects around a 2D plot.
  • .plot_ratio(...) make a ratio plot between the histogram and another histogram or callable.
  • .plot_pull(...) performs a pull plot.
  • .plot_pie() makes a pie plot.
  • .show() provides a nice str printout using Histoprint.
• Stacks: work with groups of histograms with identical axes
  • Stacks can be created with h.stack(axis), using index or name of an axis (StrCategory axes ideal).
  • You can also create with hist.stacks.Stack(h1, h2, ...), or use from_iter or from_dict.
  • You can index a stack, and set an entry with a matching histogram.
  • Stacks support .plot() and .show(), with names (plot labels default to original axes info).
  • Stacks pass through .project, *, +, and -.
• New modules
  • intervals supports frequentist coverage intervals.
• Notebook ready: Hist has gorgeous in-notebook representation.
  • No dependencies required

Usage

from hist import Hist

Quick construction, no other imports needed:

h = ( Hist.new.Reg(10, 0, 1, name="x", label="x-axis") .Var(range(10), name="y", label="y-axis") .Int64() )

Filling by names is allowed:

h.fill(y=[1, 4, 6], x=[3, 5, 2])

Names can be used to manipulate the histogram:

h.project("x") h[{"y": 0.5j + 3, "x": 5j}]

You can access data coordinates or rebin with a j suffix:

h[0.3j:, ::2j] # x from .3 to the end, y is rebinned by 2

Elegant plotting functions:

h.plot() h.plot2d_full() h.plot_pull(Callable)

Development

From a git checkout, either use nox, or run:

python -m pip install -e .[dev]

See Contributing guidelines for information on setting up a development environment.

Contributors

We would like to acknowledge the contributors that made this project possible (emoji key):

Henry Schreiner🚧 💻 📖	Nino Lau🚧 💻 📖	Chris Burr💻	Nick Amin💻	Eduardo Rodrigues💻	Andrzej Novak💻	Matthew Feickert💻
Kyle Cranmer📖	Daniel Antrim💻	Nicholas Smith💻	Michael Eliachevitch💻	Jonas Eschle📖

This project follows the all-contributors specification.

Talks

• 2021-07-07 PyHEP 2021 -- High-Performance Histogramming for HEP Analysis 🎥
• 2020-09-08 IRIS-HEP/GSOC -- Hist: histogramming for analysis powered by boost-histogram 🎥
• 2020-07-07 SciPy Proceedings 🎥
• 2020-07-17 PyHEP 2020 🎥

Acknowledgements

This library was primarily developed by Henry Schreiner and Nino Lau.

Support for this work was provided by the National Science Foundation cooperative agreement OAC-1836650 (IRIS-HEP) and OAC-1450377 (DIANA/HEP). Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.