histogrammar is a Python package for creating histograms. histogrammar has multiple histogram types, supports numeric and categorical features, and works with Numpy arrays and Pandas and Spark dataframes. Once a histogram is filled, it's easy to plot it, store it in JSON format (and retrieve it), or convert it to Numpy arrays for further analysis.
At its core histogrammar is a suite of data aggregation primitives designed for use in parallel processing. In the simplest case, you can use this to compute histograms, but the generality of the primitives allows much more.
Several common histogram types can be plotted in Matplotlib, Bokeh and PyROOT with a single method call. If Numpy or Pandas is available, histograms and other aggregators can be filled from arrays ten to a hundred times more quickly via Numpy commands, rather than Python for loops. If PyROOT is available, histograms and other aggregators can be filled from ROOT TTrees hundreds of times more quickly by JIT-compiling a specialized C++ filler. Histograms and other aggregators may also be converted into CUDA code for inclusion in a GPU workflow. And if PyCUDA is available, they can also be filled from Numpy arrays by JIT-compiling the CUDA code.
This Python implementation of histogrammar been tested to guarantee compatibility with its Scala implementation.
Latest Python release: v1.0.33 (Dec 2022). Latest update: Dec 2023.
See Changes log here.
With Spark 3.X, based on Scala 2.12 or 2.13, make sure to pick up the correct histogrammar jar files:
spark = SparkSession.builder.config("spark.jars.packages", "io.github.histogrammar:histogrammar_2.12:1.0.30,io.github.histogrammar:histogrammar-sparksql_2.12:1.0.30").getOrCreate()For Scala 2.13, in the string above simply replace "2.12" with "2.13".
December, 2023
| Tutorial | Colab link |
|---|---|
| Basic tutorial | 
|
| Detailed example (featuring configuration, Apache Spark and more) |
|
| Exercises |
|
See histogrammar-docs for a complete introduction to histogrammar. (A bit old but still good.) There you can also find documentation about the Scala implementation of histogrammar.
The historgrammar library requires Python 3.8+ and is pip friendly. To get started, simply do:
$ pip install histogrammaror check out the code from our GitHub repository:
$ git clone https://github.com/histogrammar/histogrammar-python
$ pip install -e histogrammar-pythonwhere in this example the code is installed in edit mode (option -e).
You can now use the package in Python with:
import histogrammarCongratulations, you are now ready to use the histogrammar library!
As a quick example, you can do:
import pandas as pd
import histogrammar as hg
from histogrammar import resources
# open synthetic data
df = pd.read_csv(resources.data('test.csv.gz'), parse_dates=['date'])
df.head()
# create a histogram, tell it to look for column 'age'
# fill the histogram with column 'age' and plot it
hist = hg.Histogram(num=100, low=0, high=100, quantity='age')
hist.fill.numpy(df)
hist.plot.matplotlib()
# generate histograms of all features in the dataframe using automatic binning
# (importing histogrammar automatically adds this functionality to a pandas or spark dataframe)
hists = df.hg_make_histograms()
print(hists.keys())
# multi-dimensional histograms are also supported. e.g. features longitude vs latitude
hists = df.hg_make_histograms(features=['longitude:latitude'])
ll = hists['longitude:latitude']
ll.plot.matplotlib()
# store histogram and retrieve it again
ll.toJsonFile('longitude_latitude.json')
ll2 = hg.Factory().fromJsonFile('longitude_latitude.json')These examples also work with Spark dataframes (sdf):
from pyspark.sql.functions import col
hist = hg.Histogram(num=100, low=0, high=100, quantity=col('age'))
hist.fill.sparksql(sdf)For more examples please see the example notebooks and tutorials.
This package was originally authored by DIANA-HEP and is now maintained by volunteers.
- Issues & Ideas & Support: https://github.com/histogrammar/histogrammar-python/issues
Please note that histogrammar is supported only on a best-effort basis.
histogrammar is completely free, open-source and licensed under the Apache-2.0 license.