JupyterHub

Once you login, a new instance/server will be allocated to you, where you can store your data files and code files. The file structure is similar to what you see on your own system (laptop/pc). 

To create a new Notebook for Python:

1. Navigate to ‘Files’ tab.  
2. Click ‘New▼’ drop-down button. 
3. Select ‘Python 3 (ipykernel).
   
4. A new Python (iPython notebook) file will be created in the current directory and opened in a new tab. 
5. Click on the ‘Untitled’ in the title bar to rename the file.
   The body of the page contains a rectangular bar-like structure called Cell. This is where the code/markdown text is written. The toolbar has various buttons to add a new cell, copy a cell, run a cell, change the type of the cell (Code, Markdown, Raw NBConvert and Heading), and so on.

Markdown Cell

The markdown cell is generally used for providing explanations or descriptions of code within Jupyter notebooks. It enables users to incorporate various formatting elements such as headings, titles, bullet points, and numbering, enhancing the readability and organization of their content. The cell mode can be changed to Markdown type by selecting Markdown from the dropdown button in the tool bar.

An example of a Markdown cell is illustrated below, showcasing how the output is rendered when the cell is executed.  

To learn more about Markdown refer to The Ultimate Markdown Guide (for Jupyter Notebook). 

The U of T’s Educational JupyterHub is setup with Anaconda Python distribution and comes with common packages preinstalled. The packages can be imported into the current notebook as usual by typing “import “.  

You can install your own packages or packages that do not exist in the U of T JupyterHub directly from the code cell by using the Jupyter line magic command, %.  

To install a new package, run “%pip install <package name>” in a new cell. By doing this, you do not need to open the terminal to install new packages. This is done from within the notebook itself.  

The manually installed packages do not persist in the system after logging out. This means that if you had installed a package from within the notebook, the next time you log in and try to run the code, you will have to re-install the package. To avoid this, it is better to implement the try and except method offered by Python, as shown below: 

Another advantage of a Jupyter notebook is that you can easily read the descriptions of packages or Python objects such as class, methods, and functions, if the description is provided in the code by developer. Just type ‘?’ after the function/package name and run the code cell.

The ‘%’ sign used while installing new packages and ‘?’ used to view the documentation of certain packages or functions are examples of magic commands.

These commands can be used in Jupyter notebooks for enhancing the functionality and understanding of the code. Magics are specific to and provided by the IPython kernel.

While there are numerous magic functions available with the Python kernel, two prominent examples are showcased below. To learn more about other available commands, please refer to the Built-in magic commands documentation.

Examples of such magic commands 

1. %%time or %time: These commands are used to measure the time taken by a function or a line of code to execute. To use this command in line mode prefix with single % sign. To use it in cell mode, prefix with %%.  
2. %who_ls: This command shows the sorted list of all the interactive variables/aliases used in the notebook. 
   

There are various ways to distribute files for use in JupyterHub:  

• One-click download (nbgitpuller):

  This method requires files to be uploaded in a public GitHub repository. A sharable link can be created using nbgitpuller. When clicked on this link, users will be taken directly to the UofT JupyterHub and files will be fetched automatically into user file explorer inside JupyterHub. This can be used as a one-to-many file distribution mechanism.

  1. Create public repository on GitHub and upload files to the repository.
  2. Navigate to https://jupyterhub.github.io/nbgitpuller/link and fill out all the required details.
     

  Nbgitpuller will always respect and retain user modified files when a git repo is retrieved more than once. Learn more about Automatic Merging Behavior and how various possible cases are handled where the student (end user) and instructor (author of the repo) modify the repository.

• Upload/download files manually

  Instructors can use this method to post all the files required for the assignment on Quercus from where the students can download and upload manually to their JupyterHub directory. Clicking on the upload button will open the file picker to select and upload files from student’s computer to JupyterHub.

  

• Read files from URL using supported Python modules

  Example for one such module, Pandas, is shown below which reads a csv file into a Pandas DataFrame object.

  

1. Open a notebook you want to export. 
2. Click on the ‘File’ menu button.
3. Click on ‘Download as’ to see file options. 
4. Select the appropriate file extension. 
5. A download of exported file will begin shortly. 
6. The file will be saved at a location specified or the default download location. 

1. Navigate to the ‘File’ tab on the top-left.
2. Select the ‘New file’ button.
3. Select the type of file such as (R Script, R Markdown).
4. A new (untitled) file will be created in the current directory and opened in a new tab.
5. Select the save current document button to save and change your file name. 

1. Navigate to the ‘File’ tab on the top-left.
2. Click the ‘New Project’ button.
3. Choose to create a project in existing directory or new directory.
4. Choose the folder you want to store the project.
5. A new project will be created in the current directory. 

1. Navigate to the “Files” tab in the bottom-right panel.
2. Select a file by clicking the checkbox on the left.
3. Click the “More” button.
4. Select “Move”.
5. Select the destination folder. 
6. Click “Yes” to finish move or “No” to cancel this action. 

1. Navigate to the “Files” tab in the bottom-right panel.
2. Select a file by clicking the checkbox on the left.
3. Click the “More” button.
4. Select “Copy”.
5. By default, the copied file is named as “Copyof{filename}”, you can change the file name.
6. Click “OK” to finish move or “Cancel” to cancel this action. 

1. Click the ‘Knit’ button on the top.
2. Select the file type (Knit to HTML, Knit to PDF, Knit to Word).
3. Wait until a new window shows you the exported file. 
4. Click the download button on the top-left to download the exported file. 

1. Navigate to the ‘Files’ tab on the bottom-right.
2. Click “Upload” button on the bottom-right panel.  
3. Click the “Choose File” button.
4. Select the file you want to upload from your computer.
5. Click “OK” to finish or “No” to cancel this action.

1. Navigate to the ‘Files’ tab on the bottom-right.
2. Select a file by clicking the checkbox on the left. 
3. Click the “More” button.
4. Select “Export”.
5. Click “Download” to download the file to your computer. 

Given the JupyterHub limited resources, it is important to be mindful of certain factors to ensure optimal performance and efficiency, and prioritize tasks and workflows that are well-suited to the available system resources, optimizing productivity while minimizing the risk of performance issues.

A default instance of JupyterHub is allocated 2GB of RAM and shared access to several CPU cores. 

Things to avoid when working within these constraints: 

• Needlessly Complicated Dataset Manipulations: While JupyterHub provides a versatile platform for data manipulation and analysis, it is essential to avoid overly complex operations that may strain the system resources. Attempting to perform intricate dataset manipulations or computations can lead to slow performance and potential system instability. 
• Working with Very Large Datasets: It is advisable to avoid working with very large datasets that exceed the system’s capacity. Processing large datasets can quickly consume available memory and processing power, resulting in performance degradation and potential system crashes. If you are considering an exercise or assignment to be distributed to students – ensure your data set size is well matched to the exercise and not needlessly large. 
• Installing using ‘conda install’: pip install has considerably lower memory usage, so if you experience issues installing packages using ‘conda’, try ‘pip’ instead.