Getting Started

You’ll need:

  • The Java 8 JDK.

  • Spark 2.0.2. Hail should work with other versions of Spark 2, see below.

  • Python 2.7 and IPython. We recommend the free Anaconda distribution.

  • CMake and a C++ compiler that supports -std=c++11 (we recommend at least GCC 4.7 or Clang 3.3).

    On a Debian-based Linux OS like Ubuntu, run:

    $ sudo apt-get install g++ cmake

    On Mac OS X, install Xcode, available through the App Store, for the C++ compiler. CMake can be downloaded from the CMake website or through Homebrew. To install with Homebrew, run

    $ brew install cmake
  • The Hail source code. To clone the Hail repository using Git, run

    $ git clone
    $ cd hail

    You can also download the source code directly from Github.

    You may also want to install Seaborn, a Python library for statistical data visualization, using conda install seaborn or pip install seaborn. While not technically necessary, Seaborn is used in the tutorial to make prettier plots.

To install all dependencies for running locally on a fresh Ubuntu installation, use this script.

The following commands are relative to the hail directory.

Building and running Hail

Hail may be built to run locally or on a Spark cluster. Running locally is useful for getting started, analyzing or experimenting with small datasets, and Hail development.

Running locally

The single command

$ ./gradlew shadowJar

creates a Hail JAR file at build/libs/hail-all-spark.jar. The initial build takes time as Gradle installs all Hail dependencies.

Add the following environmental variables by filling in the paths to SPARK_HOME and HAIL_HOME below and exporting all four of them (consider adding them to your .bashrc):

$ export SPARK_HOME=/path/to/spark
$ export HAIL_HOME=/path/to/hail
$ export PYTHONPATH="$PYTHONPATH:$HAIL_HOME/python:$SPARK_HOME/python:`echo $SPARK_HOME/python/lib/py4j*`"
$ export SPARK_CLASSPATH=$HAIL_HOME/build/libs/hail-all-spark.jar

Running ipython on the command line will open an interactive Python shell.

Here are a few simple things to try in order. To import the hail module and start a HailContext, run:

>>> from hail import *
>>> hc = HailContext()

To import the included sample.vcf into Hail’s .vds format, run:

>>> hc.import_vcf('src/test/resources/sample.vcf').write('sample.vds')

To split multi-allelic variants, compute a panel of sample and variant quality control statistics, write these statistics to files, and save an annotated version of the vds, run:

>>> vds = ('sample.vds')
...     .split_multi()
...     .sample_qc()
...     .variant_qc())
>>> vds.export_variants('variantqc.tsv', 'Variant = v, va.qc.*')
>>> vds.write('sample.qc.vds')

To count the number of samples, variants, and genotypes, run:

>>> vds.count(genotypes=True)

Now let’s get a feel for Hail’s powerful objects, annotation system, and expression language. To print the current annotation schema and use these annotations to filter variants, samples, and genotypes, run:

>>> print('sample annotation schema:')
>>> print(vds.sample_schema)
>>> print('\nvariant annotation schema:')
>>> print(vds.variant_schema)
>>> (vds.filter_variants_expr('v.altAllele().isSNP() && va.qc.gqMean >= 20')
...     .filter_samples_expr('sa.qc.callRate >= 0.97 && sa.qc.dpMean >= 15')
...     .filter_genotypes('let ab =[1] / in '
...                       '((g.isHomRef() && ab <= 0.1) || '
...                       ' (g.isHet() && ab >= 0.25 && ab <= 0.75) || '
...                       ' (g.isHomVar() && ab >= 0.9))')
...     .write('sample.filtered.vds'))

Try running count() on sample.filtered.vds to see how the numbers have changed. For further background and examples, continue to the Overview and API reference.

Note that during each run Hail writes a hail.log file in the current directory; this is useful to developers for debugging.

Running on a Spark cluster

Hail can run on any cluster that has Spark 2 installed. For instructions specific to Google Cloud Dataproc clusters and Cloudera clusters, see below.

For all other Spark clusters, you will need to build Hail from the source code. First, log onto the master node of the Spark cluster, and build a Hail JAR and a zipfile of the Python code by running:

$ ./gradlew shadowJar archiveZip

Then set the following environment variables with values peculiar to your cluster for SPARK_HOME and HAIL_HOME:

$ export SPARK_HOME=/path/to/spark/
$ export HAIL_HOME=/path/to/hail
$ export PYTHONPATH="$PYTHONPATH:$HAIL_HOME/build/distributions/$SPARK_HOME/python:$SPARK_HOME/python/lib/py4j-*"

You can open an IPython shell with the ipython command:

$ ipython

Within the interactive shell, check that you can successfully create a HailContext by running the following commands. Note that you have to pass in the existing SparkContext instance sc to the HailContext constructor.

>>> import hail
>>> hc = hail.HailContext(sc)

Files can be accessed from both Hadoop and Google Storage. If you’re running on Google’s Dataproc, you’ll want to store your files in Google Storage. In most on premises clusters, you’ll want to store your files in Hadoop.

To convert sample.vcf stored in Google Storage into Hail’s .vds format, run:

>>> hc.import_vcf('gs:///path/to/sample.vcf').write('gs:///output/path/sample.vds')

To convert sample.vcf stored in Hadoop into Hail’s .vds format, run:

>>> hc.import_vcf('/path/to/sample.vcf').write('/output/path/sample.vds')

It is also possible to run Hail non-interactively, by passing a Python script to spark-submit. In this case, it is not necessary to set any environment variables.

For example,

$ spark-submit --jars build/libs/hail-all-spark.jar \
               --py-files build/distributions/ \

runs the script (which reads and writes files from Hadoop):

import hail
hc = hail.HailContext()

Running on a Cloudera Cluster

These instructions explain how to install Spark 2 on a Cloudera cluster. You should work on a gateway node on the cluster that has the Hadoop and Spark packages installed on it.

Once Spark is installed, running Hail on a Cloudera cluster is exactly the same as above, except:

  • On a Cloudera cluster, SPARK_HOME should be set as: SPARK_HOME=/opt/cloudera/parcels/SPARK2/lib/spark2,

  • On Cloudera, you can create an interactive Python shell using pyspark2:

    $ pyspark2 –jars build/libs/hail-all-spark.jar

    –py-files build/distributions/ –conf,, –conf spark.sql.files.openCostInBytes=1099511627776 –conf spark.sql.files.maxPartitionBytes=1099511627776 –conf spark.hadoop.mapreduce.input.fileinputformat.split.minsize=1099511627776 –conf spark.hadoop.parquet.block.size=1099511627776

  • Cloudera’s version of spark-submit is called spark2-submit.

Running in the cloud

Google and Amazon offer optimized Spark performance and exceptional scalability to tens of thousands of cores without the overhead of installing and managing an on-prem cluster.

Hail publishes pre-built JARs for Google Cloud Platform’s Dataproc Spark clusters. If you would prefer to avoid building Hail from source, learn how to get started on Google Cloud Platform by reading this forum post.

Building with other versions of Spark 2

Hail should work with other versions of Spark 2. To build against a different version, such as Spark 2.1.0, modify the above instructions as follows:

  • Set the Spark version in the gradle command

    $ ./gradlew -Dspark.version=2.1.0 shadowJar
  • SPARK_HOME should point to an installation of the desired version of Spark, such as spark-2.1.0-bin-hadoop2.7

  • The version of the Py4J ZIP file in the hail alias must match the version in $SPARK_HOME/python/lib in your version of Spark.


Hail uses BLAS and LAPACK optimized linear algebra libraries. These should load automatically on recent versions of Mac OS X and Google Dataproc. On Linux, these must be explicitly installed; on Ubuntu 14.04, run

$ apt-get install libatlas-base-dev

If natives are not found, hail.log will contain the warnings

Failed to load implementation from: com.github.fommil.netlib.NativeSystemLAPACK
Failed to load implementation from: com.github.fommil.netlib.NativeSystemBLAS

See netlib-java for more information.

Running the tests

Several Hail tests have additional dependencies:

Other recent versions of QCTOOL and R should suffice, but PLINK 1.7 will not.

To execute all Hail tests, run

$ ./gradlew -Dspark.home=$SPARK_HOME test