changed 4 years ago
Linked with GitHub

Jupyter Community Call 2/23

Date: February 23, 2021, at 9am Pacific (your timezone)

Video-conference link: https://zoom.us/my/jovyan?pwd=c0JZTHlNdS9Sek9vdzR3aTJ4SzFTQT09

Discourse: https://discourse.jupyter.org/t/jupyter-community-calls/668

Give feedback on the call: https://docs.google.com/forms/d/e/1FAIpQLScwfYswVhafS9PVIoQYepIExq3f-FP7EmsAFULCiTIgc7mRSA/viewform

Please note:

Welcome!

If you are joining the Jupyter Community Call, sign in below so we know who was here.

Name Institution GitHub Handle
Jagane InfinStor, Inc. @jagane-opensource
Loïc Huder ESRF @loichuder
Frédéric Collonval ARIADNEXT @fcollonval
Layne Sadler AIQC @aiqc
Nick Bollweg GTRI, Project Jupyter @nrbgt @bollwyvl
Isabela Presedo-Floyd Quansight Labs @isabela-pf
M. Krassowski (Mike) UOxf (on own behalf) @krassowski
Simon Li University of Dundee @manics
Dan Lester Ideonate @danlester
Wayne Decatur Upstate University @fomightez
Michael Milligan MSI @ U of Minnesota @mbmilligan
Matthew Seal Noteable, Inc @mseal
Jose Ferro. EPO (www.epo.org) @joseberlines
Marvin Kastner Hamburg University of Technology @1kastner
Zach Sailer Apple @Zsailer
A. T. Darian Two Sigma @afshin
W Stein CoCalc @williamstein
Raja Rajendran InfinStor, Inc. @software-artisan
Adhitya Vadivel InfinStor, Inc. @adhityav
Jeremy Tuloup QuantStack @jtpio
Steven Silvester Apple @blink1073
Safia Abdalla Microsoft, nteract @captainsafia
Shelby Sturgis Netflix @stormpython
Pete Blois Google @blois

Purpose

Think of it as a monthly, virtual JupyterCon. It’s a place to announce and share fun things happening in the Jupyter community.

For more discussion on the format of these calls, see the thread here.

Short reports, celebrations, shout-outs

  • Nick B: Are you excited by JS license compliance? Feedback welcome on lab#9779 (screenshot)!
  • Nick B: The latest Lab RTC PR (demo) is 🔥!
  • Isabela Presedo-Floyd: If you are interested in helping run or host a community call, let me know (you won't have to do it alone
    Image Not Showing Possible Reasons
    • The image file may be corrupted
    • The server hosting the image is unavailable
    • The image path is incorrect
    • The image format is not supported
    Learn More →
    )!
  • Mike: pyls fork python-ls in cooperation with the Spyder team; preview of the changes to come in my personal fork (fast and more clever autocompletion for jupyterlab-lsp!)
  • Matt S: testbook had a talk at PyCascades around unittesting Jupyter Notebooks.

Agenda Items

  • Loic Huder: short presentation of jupyter-h5web, a JLab extension to explore/visualize HDF5 files
  • Jagane Sundar: short presentation of InfinStor's free hosted Jupyterhub service (https://infinstor.com), a Cloud hosted free Jupyterhub - our Jupyterhub service is free, you pay for Jupyterlab instance resources consumed
  • Frédéric Collonval: short presentation of papermill_report. JupyterHub service to generate report from notebooks combining papermill and nbconvert.
  • Nick B: demo of ipydrawio (🍴 with 💖 of @quantstack/jupyterlab-drawio), the full diagrams.net UI (with all plugins, themes, shapes) for Lab 3. Supports *.{ipynb,drawio,png,svg} (plus .pdf, sorta)

This is a space to store links shared during community call discussions related to or separate from the agenda items.

  • agenda from 1/26 (last time)

  • h5web

    • based on same backend as other hd5f extension, will share more code in the future
      • other one's based on lumino datagrid
    • keras: export neural network weights as hdf5, potential use case
    • xray community currently
      • viewer should work for all formats
    • drug discovery
    • preventing jlab from sending all the data
      • quick fix: filetype base64 to just send a little data (custom contentsmanager?)
      • would prefer basically no data
  • infinistor product

    • hub
      • custom spawner/authenticator (FOSS tbd)
        • didn't want existing spawner (container/k8s)
        • vm interface granular enough
        • stoppable machines for lower idle cost
        • requires infinistor
        • aws auth
        • token refreshing
      • hub runs on infinistor
      • spawns in your own aws, shutdown on 15m of idle
      • stock Lab, BYO extensions
      • custom tool for extensions
        • snapshots
      • dashboard
      • s3 storage
      • users
      • multicloud (aws, azure)
      • customize the python kernel?
        • in backend, have transforms, use dockerfile, modify dockerfile
        • stock open source distributions (pip, conda)
  • show and tell on discourse

  • papermill_report

    • speaks python (other libraries could be added)
    • install as a service on hub
    • pick notebook
    • fill out form from papermill
    • click button, get report
    • shareable url
    • if break, get full traceback
    • click to get annotated notebook report
    • next time: integration testing jupyterlab, screenshots w/ playwright, etc.
    • papermill inspect added for this purpose, first example of using in FOSS
    • multi-step wizard?
      • schedulers
        • cylc
        • dagster
        • prefect
    • hiding code?
      • nbconvert (use tags?)
      • preprocessors
      • all standard stuff
  • ipydrawio

    • nb: most important takeaway: write the mimerenderer first, then it's easier to write a widget/document

    • More draw.io features available in notebooks
Select a repo