Client Meeting 2

30 May 2018 17:00-18:00

Progress since last meeting

• Went through all implemented features on alpha version via screen share on Skype
• Fixed a white-space preservation issue regarding line spacing on python card which was working but stopped after a fix. An Issue was raised on GitHub and a unit test was created in response.

Suggestions from Microsoft

• add search output option for users
• add button for export to jupyter
• asked about infer card names algorithm; the choice of algorithm was a question included in user survey
• add 'restart kernel' button if it crashes
• add 'import jupyter notebook' feature into editor, inject individual cards into a python file
• load notebook and re-export option - ask the users, could be unprompted in most cases. The general concept - make it easy to edit a .ipynb file in editor and go between jupyter and the extension
• add ability to have spreadsheets and grid control, not just data frame viewing but manipulations
• add options for user to disable or enable features

Summary of Q&A with Microsoft

• How is the output rendered? - Microsoft
  • Objects are taken from integrated jupyter kernel and we render the html
• How efficient is the extension? - Microsoft
  • We Injected loads of cells to stress test, and pushed extension to its limits until it crashes and monitored memory usage. The plotly library slows down on stress test, but we can have up to 500 card instances before the slow down.
• Why was a export to browser function requested? - The Team
  • VS Code guidelines prohibit dual windows which can help compare multiple plots and maximize screen real estate. They don't care if it is in jupyter or a standard browser, its more about using multiple screen real estate and comparing things side by side.
• What do we call the extension? - The Team
  • VS Code Interactive Programming Experience is the internal high level Microsoft name, we can come up with a code-name for the extension
• How tightly should we couple to the Jupyter Kernel?
  • The extension should be agnostic of back-end to the user. The extension has loose coupling to a particular language which is good because it can switch between languages easily. Being tightly coupled to jupyter is fine, because Azure Notebooks uses Jupyter Kernel.

General Comments

• Extension is 'fantastic' and lots of potential to add more useful functionality
• Set up weekly meetings as deadlines approach
• Send e-mail when the next extension pre-realease is ready on GitHub
• Contacted Lee Stott about presentation, he provided some guidance about Microsoft branding and presentation