Sample Binder Repositories#

Below we list several sample Binder repositories thatdemonstrate how to compose build files in order to createBinders with varying environments. You can find all of therepositories listed on this page at thebinder-examples GitHub organization.

Managing languages#

Python environment with a requirements.txt#

Binder |

A Binder-compatible repo with a requirements.txt file.

Access this Binder at the following URL

https://mybinder.org/v2/gh/binder-examples/requirements/master

Notes#

The requirements.txt file should list all Python libraries that your notebooks depend on, and they will be installed using:

pip install -r requirements.txt

The base Binder image contains no extra dependencies, so be as explicit as possible in defining the packages that you need. This includes specifying explicit versions wherever possible.

If you do specify strict versions, it is important to do so for all your dependencies, not just direct dependencies. Strictly specifying only some dependencies is a recipe for environments breaking over time.

pip-compile is a handy tool for combining loosely specified dependencies with a fully frozen environment. You write a requirements.in with just the dependencies you need and pip-compile will generate a requirements.txt with all the strict packages and versions that would come from installing that package right now. That way, you only need to specify what you actually know you need, but you also get a snapshot of your environment.

In this example we include the library seaborn which will be installed in the environment, and our notebook uses it to plot a figure.

Files in this repository#

.github
index.ipynb
requirements.in
requirements.txt
runtime.txt

Conda environment with environment.yml#

Binder |

A Binder-compatible repo with an environment.yml file.

Access this Binder by clicking the blue badge above or at the following URL:

https://mybinder.org/v2/gh/binder-examples/conda_environment/master?filepath=index.ipynb

Notes#

The environment.yml file should list all Python libraries on which your notebooks depend, specified as though they were created using the following conda commands:

conda activate example-environment
conda env export --from-history -f environment.yml

Note that the only libraries available to you will be the ones specified in the environment.yml, so be sure to include everything that you need!

Also note that if you skip the --from-history, conda may include OS-specific packages in environment.yml, which you would have to manually prune from environment.yml. For example, confirmed macOS-specific packages that should be removed are:

  • libcxxabi=4.0.1

  • appnope=0.1.0

  • libgfortran=3.0.1

  • libcxx=4.0.1

Files in this repository#

.github
environment.yml
index.ipynb

python package with setup.py#

Binder |

A Binder-compatible repo with a python package and a setup.py file.

Notes#

It is convenient to provide an example Jupyter notebook for a new package and add the hooks necessary to run the example with Binder. However, normally the package will not be included in the python path. To do that, one needs a setup.py for the package (see binder docs). Once this is done, it is possible to import the package in a notebook running within Binder.

This setup.py was originally adapted from kennethreitz/setup.py by @cranmer

Files in this repository#

MANIFEST.in
example_notebook
mypackage
setup.py

Julia and Python environments#

Binder |

This example shows how you can install a Julia and Python environment side-by-side. In this repository are both an environment.yml file as well as a REQUIRE file. The former corresponds to an anaconda python environment, and the latter corresponds to a Julia environment. Both kernels will be available to you in a built Binder environment.

Files in this repository#

REQUIRE
environment.yml
julia.ipynb
python-and-julia.ipynb
python.ipynb

Julia Binder demo#

This is a demo of Julia functionality for the Binder project. Simply go to the URL below and it will launch an interactive Julia environment:

Binder |

Files in this repository#

Manifest.toml
Project.toml
demo.ipynb

Specifying an R environment with a runtime.txt file#

Jupyter+R: Binder

RStudio: Binder

RShiny: Binder

Binder supports using R and RStudio, with libraries pinned to a specific snapshot on MRAN.

Note: We recommend to follow r-conda instead. Especially if you want to use a specific version of R or need faster build times.

Note: Another alternative is to use the holepunch package for R.

Requirements and suggestions#

You need to have a runtime.txt file that is formatted like:

r-<YYYY>-<MM>-<DD>

where YYYY-MM-DD is a snapshot at MRAN that will be used for installing libraries. In this line, you can request a specific version of R. To do this list the version between the ‘r’ and the year, as in r-3.6-2019-09-24. Right now the default version of R is 3.6.

We recommend using https://github.com/binder-examples/r-conda for faster installs than using a install.R

To install R libraries (or packages) you can add an install.R file that specifies one library to install per line.

Both RStudio and IRKernel are installed by default, so you can use either the Jupyter notebook interface or the RStudio interface.

This repository also contains an example of a Shiny app.

URL addresses for RStudio and Shiny environments#

The Binder repository can be used to allow anyone to access an RStudio environment containing our code and data right in their web browser. It also allows hosting a Shiny app. For those purposes, we have to append a bit of text to the URL of our Binder repository, which we can find out at mybinder.org when we enter the URL of our original repository from GitHub or Figshare, etc.

  • For the RStudio environment, we must add the following at the end of the URL: ?urlpath=rstudio

    • Example: https://mybinder.org/v2/gh/binder-examples/r/master?urlpath=rstudio

  • For the Shiny app environment, we must add the following at the end of the URL: ?urlpath=shiny. In this case, we also have to note that if the Shiny app files are located in a folder, this folder should be specified in the URL, after a slash. We would then also have to put in a trailing slash at the end of the URL, and to avoid spaces in the name of the repository, placing instead a hyphen (the reason is that spaces are converted to %20).

    • Example: https://mybinder.org/v2/gh/binder-examples/r/master?urlpath=shiny/bus-dashboard/

Files in this repository#

.github
bus-dashboard
index.ipynb
install.R
runtime.txt

Specifying an R environment by having a DESCRIPTION file#

Jupyter+R: Binder

RStudio: Binder

Binder supports using R and RStudio, with libraries pinned to a specific snapshot on MRAN.

If you specify a runtime.txt file that is formatted like:

r-<YYYY>-<MM>-<DD>

where YYYY-MM-DD it will use the MRAN snapshot of that day for setting up the R runtime.

Without specifying a runtime.txt it will use a 2-day old snapshot of MRAN.

Both RStudio and IRKernel are installed by default, so you can use either the Jupyter notebook interface or the RStudio interface.

Files in this repository#

DESCRIPTION
NAMESPACE
R
test-library.ipynb

Octave on mybinder.org#

Binder |

An example of using Octave on mybinder.org.

This shows you how to make Matlab code that works with Octave run on mybinder.org.

The example notebook is taken from the octave_kernel repository.

Files in this repository#

apt.txt
environment.yml
index.ipynb

User interfaces#

JupyterLab + Binder#

Binder |

JupyterLab is the default UI on Binder.

Classic notebook interface on Binder#

Binder |

JupyterLab is the default UI on Binder, but the classic notebook is available, too. In order to run a JupyterLab session, you have two options:

Start classic notebook after you start your Binder#

Do the following:

  1. Launch a Binder instance (e.g., by clicking the Binder badge)

  2. Replace lab at the end of your URL with tree.

  3. That’s it!

Files in this repository#

.profile
binder
geojson-extension.geojson
index.ipynb

Enabling Jupyter Extensions with post-build commands#

Binder (Jupyter Notebook) |

Binder (Jupyter Lab) |

This example demonstrates how to enable Jupyter extensions with Binder.We currently only cover one example in this repo. Be aware that some are idiosyncratic in how they’re enabled.

We accomplish each step using a requirements.txt file to install the extension, then a postBuild file to enable it.

ipywidgets#

Ipywidgets lets you create interactive widgets in your notebook. Installation is fairly straightforward. You install the python package, then enable the extension.

The postBuild file defines commands (one per line) to be run with bash. In this case, we first enable the ipywidgets extension in the classic notebook interface. We then use it to install a Jupyter Lab extension (by calling jupyter labextension) which allows ipywidgets to be displayed within notebooks.

Files in this repository#

index.ipynb
postBuild
requirements.txt

Creating interactive presentations on Binder with RISE#

Binder |

RISE allows you to quickly generate a live, interactive presentation from a Jupyter Notebook that is connected to the underlying Kernel of the notebook. Using a new feature for automatically launching the RISE plugin when a notebook is opened, RISE can be used to share interactive presentations that run in the cloud with Binder. This repository demonstrates how to accomplish this.

To make your RISE presentation automatically-launch with it is open, add an autolaunch=true configuration parameter to a notebook’s livereveal section in the metadata. E.g.:

...
"livereveal": {
        "autolaunch": true
        }
...

When the notebook is launched, your presentation will automatically begin.

See the RISE Documentation for more information.

Files in this repository#

environment.yml
index.ipynb

Interactive dashboard from notebook with Voilà#

Voilà

JupyterLab

voila-binder

jupyterlab-binder

This example demonstrates how to use Voilà on Binder.

screenshot

Configuration#

If you would like to use the same configuration as this repository but for another project, check out the following steps:

  1. Make sure the repository is publicly available (on GitHub, Gitlab or as a GitHub Gist

  2. Define the dependencies in environment.yml. requirements.txt is also supported. In the dependency file, add voila.

  3. Go to mybinder.org and enter the URL of the repository.

  4. In Path to a notebook file, select URL and use the Voilà endpoint: voila/render/path/to/notebook.ipynb

  5. Click Launch.

  6. Binder will trigger a new build if this is the first launch (or if there have been new changes since the last build). This might take a few minutes to complete. If an image is already available, the server will be able to start within a few seconds.

Here is an overview of the Binder configuration on mybinder.org:

image

For more details, check out the Voilà documentation on https://voila.readthedocs.io/en/latest/deploy.html#deployment-on-binder

JupyterLab Preview Extension for Voilà#

The JupyterLab Preview Extension for Voilà should normally be automatically available, since Binder now defaults to JupyterLab 3.0.

For consistency, you can add jupyterlab=3 as a dependency in environment.yml.

lab-preview

Files in this repository#

environment.yml
index.ipynb

Interactive apps from Jupyter Notebooks#

Binder |

This repository demonstrates how to create interactive webapps from a Jupyter Notebook. This is similar to how Shiny apps work in R.

Using the appmode Jupyter plugin, a notebook’s code will be run, and then only the markdown cells and code outputs will be shown.

You can check out the appmode repository here: https://github.com/oschuett/appmode

Files in this repository#

environment.yml
index.ipynb
ipyvolume_demo.ipynb
postBuild

Running a bokeh server with Binder#

Binder |

This repository demonstrates how to run a Bokeh server from within Binder. To do so, we did the following things:

  1. Created a bokeh-app directory in the repo with a main.py file in it. This is the application that will be served. We’ve added the Bokeh weather example as a demo.

  2. Installed bokeh for the viz and jupyter_server_proxy which we’ll use to direct people to the port on which Bokeh runs. See environment.yml.

  3. Added a custom server extension (bokehserverextension.py) that will be run to direct people to the Bokeh app (which is run on a port)

  4. Used postBuild to enable the jupyter_server_proxy extension, then set up and enable our custom server extension for Bokeh.

  5. Created a Binder link that uses urlpath to point users to the port on which the Bokeh server will run:

    https://mybinder.org/v2/gh/binder-examples/bokeh/master?urlpath=/proxy/5006/bokeh-app

When people click on the Binder link, they should be directed to the running Bokeh app.

Files in this repository#

bokeh-app
bokehserverextension.py
environment.yml
postBuild

stencila-py#

Demo for Stencila & DAR on binder with Python code.

Binder |

Learn more about Stencila and its integration with Binder in this blog post.

Files in this repository#

article

System environments#

Specifying a Python 2 environment with runtime.txt#

Binder |

We can specify various runtime parameters with a runtime.txt file. In this repository, we demonstrate how to install python 2 with the environment.

If you specify python-2.7 in runtime.txt, then:

  • A python3 environment is created & installed (this is what the notebook runs from)

  • A python2 environment is created and registered

  • The contents of requirements.txt are installed into the python2 environment

important: Make sure that you save your notebooks with a python 2 kernel activated, as this defines which kernel Binder will use when a notebook is opened.

note: If you also wish to install python 3 dependencies, you may do so by including a file called requirements3.txt. The packages inside will be installed into the python 3 environment.

Files in this repository#

index.ipynb
requirements.txt
runtime.txt

Mixing Python 2 and 3 kernels with runtime.txt#

Binder |

Sometimes you want both Python 2 and Python 3 (e.g., if you have a mixture of notebooks that use each version of the language). This repository demonstrates how to handle these cases with repo2docker. You can specify a Python 2.7 environment with the runtime.txt file. In this case, repo2docker will install Python 2 alongside Python 3 (though all commands will default to Python 2). In this case, you can install Python 3 dependencies with requirements3.txt, while a file called requirements.txt alone will install to the Python 2 environment.

Files in this repository#

index2.ipynb
index3.ipynb
requirements.txt
requirements3.txt
runtime.txt

Using latex with Binder#

Binder |

This repository demonstrates how to install latex alongside matplotlib for Binder. This repository also makes use of JupyterLab Latex to render latex files in Jupyter lab. This requires a few different build components:

  • apt.txt for apt-installing the latex components

  • environment.yml for installing the python dependencies

  • postBuild for forcing matplotlib to build the font cache and for installing JupyterLab Latex.

Thanks to m-weigand for giving inspiration for this repo!

Files in this repository#

apt.txt
environment.yml
images
index.ipynb
postBuild
sample.tex

Installing packages from apt repositories#

Binder |

Sometimes you want packages that exist outside of the language-specific packaging ecosystems of Python/R/Julia. Binder makes it possible to apt-install packages from the ubuntu apt repository. This repository demonstrates how to do this by specifying names in an apt.txt file.

Notes#

The apt.txt file should list all packages that your notebooks depend on, and they will be installed using:

apt-get install --yes --no-install-recommends

The base Binder image contains no extra dependencies, so be as explicit as possible in defining the packages that you need.

In this example we include the tools cowsay and lolcat which will be installed in the environment, and our notebook uses them to show a colorful message.

Files in this repository#

apt.txt
index.ipynb
postBuild

Multi-language demo.#

This is a demo showing how you can intermingle Python, R, Rust, Fortran, Cython, C.

You can try it :

Binder |

And read the accompanying blog post.

Files in this repository#

23-Cross-Language-Integration.ipynb
Manifest.toml
Project.toml
apt.txt
data.csv
environment.yml
julia.ipynb
polyglot-ds-prep.ipynb
polyglot-ds.ipynb
postBuild

Using conda with pip in the same build#

Binder |

If you use environment.yml, then Binder will use a Miniconda distribution to install your packages. However, you may still want to use pip. In this case, you should not use a requirements.txt file, but instead use a - pip section in environment.yml. This repository is an example of how to construct your environment.yml file to accomplish this.

Files in this repository#

environment.yml
index.ipynb

Data and reproducibility#

Remote Storage with Binder#

Binder |

A Binder-compatible repo that shows accessing data from remote sources.

Access this Binder at the following URL:

https://mybinder.org/v2/gh/binder-examples/remote_storage/master

Notes#

The notebooks use boto and requests to load both images and tables from S3. The image loading makes use of PIL and the table loading makes use of pandas.

Files in this repository#

index.ipynb
requirements.txt

Run this example#

Binder |

Pull data into Binder notebooks#

This example uses Quilt to inject data packages into a Jupyter notebook.

Data packages are versioned, immutable snapshots of data. Data packages may contain data of any size. Here is an example of data package: uciml/iris.

How to specify data dependencies in your own Binder#

  1. Add quilt to requirements.txt

  2. Specify data package dependencies in quilt.yml (docs). For example:

packages:
  - vgauthier/DynamicPopEstimate   # get the latest version
  - danWebster/sgRNAs:a972d92      # get a specific hash (short hash)
  - akarve/sales:tag:latest        # get a specific tag
  - asah/snli:v:1.0                # get a specific version

  1. Include the following lines at the top of postBuild. (postBuild should be executable: chmod +x postBuild on UNIX, git update-index --chmod=+x postBuild for Windows).

###!/bin/bash
quilt install

If you are adopting the binder folder pattern for your repo2docker configuration files, and including quilt.yml, your postBuild file should look like this:

###!/bin/bash
quilt install @./binder/quilt.yml

Now you can access the package data in your Jupyter notebooks:

In [1]: from quilt.data.akarve import sales
In [2]: sales.transactions()
Out[2]:
      Row ID  Order ID Order Date Order Priority  Order Quantity       Sales  \
0          1         3 2010-10-13            Low               6    261.5400
1         49       293 2012-10-01           High              49  10123.0200
2         50       293 2012-10-01           High              27    244.5700
...

Developer#

Files in this repository#

binder
index.ipynb

Nixpkgs BinderHub example#

Binder |

Why Nix?#

Nix would be a great addition to reproducible data science. It is a unique package manager. Some notable features:

  • 100% reproducible environments (pin to exact commit in repository)

  • both a source and binary package repository

  • allows customized compilation and version of every package

  • can run identical environment outside of docker (all linux distros + dawin)

  • as of now 45,000+ packages

  • fully declarative environments

  • packages: python, javascript, julia, R, haskell, perl, and many other languages (some better than others).

Assuming that you have nix installed (compatible with all linux distributions and darwin (Mac OS)) you can run this repository locally (no need for binderhub). It will be identical assuming you have pinned repositories. Nix can coexist fine with other package managers.

This derivation installs python37, numpy, and scipy.

For a more detailed example see the detailed binderhub example costrouc/nix-binder-example

Files in this repository#

LICENSE.md
default.nix
nix-introduction.ipynb

Dockerfile environments#

Minimal Dockerfiles for Binder#

Binder |

Binder needs only one thing for images to work:

  • to be able to launch jupyter notebook as a specified user (passed via docker build args as NB_UID/NB_USER)

What this means in practice is that the notebook package must be installed and on PATH:

RUN pip install --no-cache notebook

That’s almost everything.

The remaining piece is that the specified user must be able to start the notebook, which requires certain permissions like being able to write to the home directory.

The absolute bare minimum for this is to set HOME to /tmp so that it’s writable, which would make the shortest, smallest Dockerfile that works on Binder:

FROM python:3.7-slim
RUN pip install --no-cache notebook
ENV HOME=/tmp

which you can try out: Binder

However, it would be better to consume the NB_UID/NB_USER arguments and create a real user:

### create user with a home directory
ARG NB_USER
ARG NB_UID
ENV USER ${NB_USER}
ENV HOME /home/${NB_USER}

RUN adduser --disabled-password \
    --gecos "Default user" \
    --uid ${NB_UID} \
    ${NB_USER}
WORKDIR ${HOME}

From this point, you can start adding files, installing packages, etc.

Files in this repository#

Dockerfile

Using a Docker image from the Jupyter docker-stacks repository#

Binder |

Sometimes you just want to inherit from one of the pre-built images maintained by the Jupyter Project’s Docker Stacks, and perhaps add just an extra library or two. This example shows you how to do that - check out the Dockerfile.

Note that in this case we are using a docker image that already satisfies the criteria for use on binder, we don’t need to install notebook or anything manually.

Files in this repository#

Dockerfile

Specifying an R environment with a runtime.txt file#

Jupyter+R: Binder

RStudio: Binder

RShiny: Binder

Both RStudio and IRKernel are installed by default, so you can use either the Jupyter notebook interface or the RStudio interface.

This repository also contains an example of a Shiny app.

Files in this repository#

Dockerfile
bus-dashboard
index.ipynb
install.R