Introducing JupyterLab_IPython Interactive Computing and Visualization Cookbook（Second Edition）-QQ阅读男生科幻网

上QQ阅读APP看书，第一时间看更新

Introducing JupyterLab

JupyterLab is the next generation of the Jupyter Notebook. It aims at fixing many Notebook usability issues and it greatly expands its scope. JupyterLab offers a general framework for interactive computing and data science in the browser, using Python, Julia, R, or one of many other languages.

In addition to providing an improved interface to existing notebooks, JupyterLab also brings, within the same interface, a file browser, consoles, terminals, text editors, Markdown editors, CSV editors, JSON editors, interactive maps, widgets, and so on. The architecture is completely extensible and open to developers. In a word, JupyterLab is a web-based, hackable IDE for data science and interactive computing.

JupyterLab uses the exact same Notebook server and file format as the classic Jupyter Notebook, so that it is fully compatible with existing notebooks and kernels. Notebook and JupyterLab can run side to side on the same computer. You can easily switch between the two interfaces.

Note

At the time of writing, JupyterLab is still in an early stage of development. However, it is already fairly usable. The interface may change until the production release. The developer API used to customize JupyterLab is still not stable. There is no user documentation yet.

Getting ready

To install JupyterLab, type conda install -c conda-forge jupyterlab in a Terminal.

To be able to render GeoJSON files in an interactive map, install the GeoJSON JupyterLab extension with: jupyter labextension install @jupyterlab/geojson-extension.

How to do it...

We can launch JupyterLab by typing jupyter lab in a Terminal. Then we go to http://localhost:8888/lab in the web browser.
The dashboard shows, on the left, a list of files and subdirectories in the current working directory. On the right, the launcher lets us create notebooks and text files, or open a Jupyter console or a Terminal. Available Jupyter kernels are automatically displayed (here, IPython, but also IR and IJulia).

JupyterLab home
On the left panel, we can also see a list of open tabs, a list of running sessions, or a list of available commands:

Running sessions

Available commands
If we open a Jupyter notebook, we get an interface that closely resembles the classic Notebook interface:

A notebook

There are a few improvements compared to the classic Notebook. For example, we can drag and drop one or several cells:

Drag and drop in the notebook

We can also collapse cells.
If we right-click in the notebook, a contextual menu appears:

Contextual menu in the notebook

If we click on Create Console for Notebook, a new tab appears with a standard IPython console. We can drag and drop the tab anywhere in the screen, for example below the notebook panel:

Notebook and console

The IPython console is connected to the same kernel as the Notebook, so they share the same namespace. We can also open a new IPython console from the launcher, running in a separate kernel.
We can also open a system shell directly in the browser, using the term.js library:

Notebook and shell
JupyterLab includes a text editor. We can create a new text file from the launcher, rename it by giving it the .md extension, and edit it:

Markdown document

Let's right-click on the Markdown file. A contextual menu appears:

Contextual menu in a Markdown file

We can add a new panel that renders the Markdown file in real time:

Markdown rendered

We can also attach an IPython console to our Markdown file. By clicking within a code block and pressing Shift + Enter, we send the code directly to the console:

Markdown and console
We can also create and open CSV files in JupyterLab:

CSV file

The CSV viewer is highly efficient. It can smoothly display huge tables with millions or even billions of values:

Viewing a CSV file
GeoJSON files (files that contain geographic information) can also be edited or viewed with the Leaflet mapping library: