Installation requirements for the Notebook.
The IPython Notebook consists of two related components:
The Notebook can be used by starting the Notebook server with the command:
$ ipython notebook
Note that by default, the notebook doesn’t load pylab, it’s just a normal IPython session like any other. If you want pylab support, you must use:
$ ipython notebook --pylab
which will behave similar to the terminal and Qt console versions, using your default matplotlib backend and providing floating interactive plot windows. If you want inline figures, you must manually select the inline backend:
$ ipython notebook --pylab inline
This server uses the same ZeroMQ-based two process kernel architecture as the QT Console as well Tornado for serving HTTP/S requests. Some of the main features of the Notebook include:
See our installation documentation for directions on how to install the notebook and its dependencies.
You can start more than one notebook server at the same time, if you want to work on notebooks in different directories. By default the first notebook server starts in port 8888, later notebooks search for random ports near that one. You can also manually specify the port with the --port option.
The landing page of the notebook server application, which we call the IPython Notebook dashboard, shows the notebooks currently available in the directory in which the application was started, and allows you to create new notebooks.
A notebook is a combination of two things:
If you have ever used the Mathematica or Sage notebooks (the latter is also web-based) you should feel right at home. If you have not, you should be able to learn how to use it in just a few minutes.
You can create new notebooks from the dashboard with the New Notebook button or open existing ones by clicking on their name. Once in a notebook, your browser tab will reflect the name of that notebook (prefixed with “IPy:”). The URL for that notebook is not meant to be human-readable and is not persistent across invocations of the notebook server.
You can also drag and drop into the area listing files any python file: it will be imported into a notebook with the same name (but .ipynb extension) located in the directory where the notebook server was started. This notebook will consist of a single cell with all the code in the file, which you can later manually partition into individual cells for gradual execution, add text and graphics, etc.
The normal workflow in a notebook is quite similar to a normal IPython session, with the difference that you can edit a cell in-place multiple times until you obtain the desired results rather than having to rerun separate scripts with the %run magic (though magics also work in the notebook). Typically you’ll work on a problem in pieces, organizing related pieces into cells and moving forward as previous parts work correctly. This is much more convenient for interactive exploration than breaking up a computation into scripts that must be executed together, especially if parts of them take a long time to run (In the traditional terminal-based IPython, you can use tricks with namespaces and %run -i to achieve this capability, but we think the notebook is a more natural solution for that kind of problem).
The only significant limitation the notebook currently has, compared to the qt console, is that it can not run any code that expects input from the kernel (such as scripts that call raw_input()). Very importantly, this means that the %debug magic does not work in the notebook! We intend to correct this limitation, but in the meantime, there is a way to debug problems in the notebook: you can attach a Qt console to your existing notebook kernel, and run %debug from the Qt console. If your notebook is running on a local computer (i.e. if you are accessing it via your localhost address at 127.0.0.1), you can just type %qtconsole in the notebook and a Qt console will open up connected to that same kernel.
In general, the notebook server prints the full details of how to connect to each kernel at the terminal, with lines like:
[IPKernelApp] To connect another client to this kernel, use: [IPKernelApp] --existing kernel-3bb93edd-6b5a-455c-99c8-3b658f45dde5.json
This is the name of a JSON file that contains all the port and validation information necessary to connect to the kernel. You can manually start a qt console with:
ipython qtconsole --existing kernel-3bb93edd-6b5a-455c-99c8-3b658f45dde5.json
and if you only have a single kernel running, simply typing:
ipython qtconsole --existing
will automatically find it (it will always find the most recently started kernel if there is more than one). You can also request this connection data by typing %connect_info; this will print the same file information as well as the content of the JSON data structure it contains.
In addition to code cells and the output they produce (such as figures), you can also type text not meant for execution. To type text, change the type of a cell from Code to Markdown by using the button or the Ctrl-m m keybinding (see below). You can then type any text in Markdown syntax, as well as mathematical expressions if you use $...$ for inline math or $$...$$ for displayed math.
If you want to provide others with a static HTML or PDF view of your notebook, use the Print button. This opens a static view of the document, which you can print to PDF using your operating system’s facilities, or save to a file with your web browser’s ‘Save’ option (note that typically, this will create both an html file and a directory called notebook_name_files next to it that contains all the necessary style information, so if you intend to share this, you must send the directory along with the main html file).
The Download button lets you save a notebook file to the Download area configured by your web browser (particularly useful if you are running the notebook server on a remote host and need a file locally). The notebook is saved by default with the .ipynb extension and the files contain JSON data that is not meant for human editing or consumption. But you can always export the input part of a notebook to a plain python script by choosing Python format in the Download drop list. This removes all output and saves the text cells in comment areas. See ref:below <notebook_format> for more details on the notebook format.
The notebook can also import .py files as notebooks, by dragging and dropping the file into the notebook dashboard file list area. By default, the entire contents of the file will be loaded into a single code cell. But if prior to import, you manually add the # <nbformat>2</nbformat> marker at the start and then add separators for text/code cells, you can get a cleaner import with the file broken into individual cells.
While in simple cases you can roundtrip a notebook to Python, edit the python file and import it back without loss of main content, this is in general not guaranteed to work at all. First, there is extra metadata saved in the notebook that may not be saved to the .py format. And as the notebook format evolves in complexity, there will be attributes of the notebook that will not survive a roundtrip through the Python form. You should think of the Python format as a way to output a script version of a notebook and the import capabilities as a way to load existing code to get a notebook started. But the Python version is not an alternate notebook format.
The native format of the notebook, a file with a .ipynb extension, is a JSON container of all the input and output of the notebook, and therefore not valid Python by itself. This means that by default, you can not import a notebook or execute it as a normal python script. But if you want use notebooks as regular Python files, you can start the notebook server with:
ipython notebook --script
or you can set this option permanently in your configuration file with:
This will instruct the notebook server to save the .py export of each notebook adjacent to the .ipynb at every save. These files can be %run, imported from regular IPython sessions or other notebooks, or executed at the command-line as normal Python files. Since we export the raw code you have typed, for these files to be importable from other code you will have to avoid using syntax such as %magics and other IPython-specific extensions to the language.
In regular practice, the standard way to differentiate importable code from the ‘executable’ part of a script is to put at the bottom:
if __name__ == '__main__': # rest of the code...
Since all cells in the notebook are run as top-level code, you’ll need to similarly protect all cells that you do not want executed when other scripts try to import your notebook. A convenient shortand for this is to define early on:
script = __name__ == '__main__':
and then on any cell that you need to protect, use:
if script: # rest of the cell...
All actions in the notebook can be achieved with the mouse, but we have also added keyboard shortcuts for the most common ones, so that productive use of the notebook can be achieved with minimal mouse intervention. The main key bindings you need to remember are:
You can protect your notebook server with a simple single-password by setting the NotebookApp.password configurable. You can prepare a hashed password using the function IPython.lib.security.passwd():
In : from IPython.lib import passwd In : passwd() Enter password: Verify password: Out: 'sha1:67c9e60bb8b6:9ffede0825894254b2e042ea597d771089e11aed'
passwd() can also take the password as a string argument. Do not pass it as an argument inside an IPython session, as it will be saved in your input history.
You can then add this to your ipython_notebook_config.py, e.g.:
# Password to use for web authentication c.NotebookApp.password = u'sha1:67c9e60bb8b6:9ffede0825894254b2e042ea597d771089e11aed'
When using a password, it is a good idea to also use SSL, so that your password is not sent unencrypted by your browser. You can start the notebook to communicate via a secure protocol mode using a self-signed certificate by typing:
$ ipython notebook --certfile=mycert.pem
A self-signed certificate can be generated with openssl. For example, the following command will create a certificate valid for 365 days with both the key and certificate data written to the same file:
$ openssl req -x509 -nodes -days 365 -newkey rsa:1024 -keyout mycert.pem -out mycert.pem
Your browser will warn you of a dangerous certificate because it is self-signed. If you want to have a fully compliant certificate that will not raise warnings, it is possible (but rather involved) to obtain one for free, as explained in detailed in this tutorial.
Keep in mind that when you enable SSL support, you’ll need to access the notebook server over https://, not over plain http://. The startup message from the server prints this, but it’s easy to overlook and think the server is for some reason non-responsive.
If you want to access your notebook server remotely with just a web browser, here is a quick set of instructions. Start by creating a certificate file and a hashed password as explained above. Then, create a custom profile for the notebook. At the command line, type:
ipython profile create nbserver
In the profile directory, edit the file ipython_notebook_config.py. By default the file has all fields commented, the minimum set you need to uncomment and edit is here:
c = get_config() # Kernel config c.IPKernelApp.pylab = 'inline' # if you want plotting support always # Notebook config c.NotebookApp.certfile = u'/absolute/path/to/your/certificate/mycert.pem' c.NotebookApp.ip = '*' c.NotebookApp.open_browser = False c.NotebookApp.password = u'sha1:bcd259ccf...your hashed password here' # It's a good idea to put it on a known, fixed port c.NotebookApp.port = 9999
You can then start the notebook and access it later by pointing your browser to https://your.host.com:9999.
The notebooks themselves are JSON files with an ipynb extension, formatted as legibly as possible with minimal extra indentation and cell content broken across lines to make them reasonably friendly to use in version-control workflows. You should be very careful if you ever edit manually this JSON data, as it is extremely easy to corrupt its internal structure and make the file impossible to load. In general, you should consider the notebook as a file meant only to be edited by IPython itself, not for hand-editing.
Binary data such as figures are directly saved in the JSON file. This provides convenient single-file portability but means the files can be large and diffs of binary data aren’t very meaningful. Since the binary blobs are encoded in a single line they only affect one line of the diff output, but they are typically very long lines. You can use the ‘ClearAll’ button to remove all output from a notebook prior to committing it to version control, if this is a concern.
The notebook server can also generate a pure-python version of your notebook, by clicking on the ‘Download’ button and selecting py as the format. This file will contain all the code cells from your notebook verbatim, and all text cells prepended with a comment marker. The separation between code and text cells is indicated with special comments and there is a header indicating the format version. All output is stripped out when exporting to python.
Here is an example of a simple notebook with one text cell and one code input cell, when exported to python format:
# <nbformat>2</nbformat> # <markdowncell> # A text cell # <codecell> print "hello IPython"
When behind a proxy, especially if your system or browser is set to autodetect the proxy, the html notebook might fail to connect to the server’s websockets, and present you with a warning at startup. In this case, you need to configure your system not to use the proxy for the server’s address.
In Firefox, for example, go to the Preferences panel, Advanced section, Network tab, click ‘Settings...’, and add the address of the notebook server to the ‘No proxy for’ field.