jupyterlab / jupyterlab

JupyterLab computational environment.

Home Page:https://jupyterlab.readthedocs.io/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

.csv file viewer very slow

asteppke opened this issue · comments

Description

I vividly remember that during the transition from Jupyter Notebook to Jupyter Lab that there was a presentation from the core team where a .csv file was opened and scrolling through a huge table happened smooth and instantaneous.

Recently I need to work more and more with .csv files and even scrolling from top to bottom in a very moderately sized table (15 rows, 2000 columns) takes roughly 1-2 minutes.

No other parts of the interface (menus, notebooks, side-panels) show any sign of sluggishness.

Reproduce

  1. Open a .csv file of moderate size (< 1 MB) with mixed contents (numbers and strings)
  2. Scroll to the bottom and observe the update rate of the output (roughly every 2 seconds)

Expected behavior

Fast scrolling in both dimensions.

Context

Linux (RHEL 8)
Jupyterlab 4.1.8
Chrome browser

Several extensions that should not influence .csv file behavior as far as I know:

  • jupyterlab-matplotlib
  • jupyterlab-jupytext
  • jupyterlab-code-formatter
  • jupyterlab-pygments
  • jupyterlab-collaboration-extension

@krassowski Thank you for looking into this instantly!

In the lumino examples the "Streaming Rows", "Random Ticks 2" and "Trillion Rows" are very slow (2s updates) here.

In the examples on my computer (macOS Ventura, M1 Pro chip, Firefox), "Random Ticks 2" is well below 60 fps when I scroll on either axis. Other examples on that page render smoothly. With Chrome on the same computer, only "Streaming Rows" displays very slowly.

Which version of Chrome are you using? Do you have hardware acceleration enabled? If so or if not, how does the behavior change when you disable/enable hardware acceleration in Chrome settings?

@JasonWeill:
This occurs with Chrome 121.0.6167.160 and hardware acceleration enabled.

Meanwhile I tried to reproduce this and with a fresh Chrome and only loading the lumino example and cannot get it the sluggishness back. At least in Chrome it seems to require some other ingredient (maybe lack of GPU resources from other browser tabs) to really trigger this.

I will close this for now because until I find a clear way how to reproduce this.