I'm trying to select logs with mouse -- only the logs that are currently in viewport are selected

[Felix-neko]

Apache Airflow version

3.1.0

If "Other Airflow 2 version" selected, which one?

No response

What happened?

Hi folks! I'm trying to select some logs in my airflow 3.1.0 (a bit more that could be placed in one screen). I'm pressing left mouse button and start selecting text. I'm moving mouse cursor down to the bottom of the viewport, the text is scrolled -- everything looks okay.

But.

The logs that are not currently on the screen are not selected. Can you please fix it?

What you think should happen instead?

No response

How to reproduce

I'm using Ubuntu 24.04 + Google Chrome + Python 3.11

Here's some screencast: https://www.sendspace.com/file/yddkb0

Operating System

Ubuntu 24.04

Versions of Apache Airflow Providers

No response

Deployment

Virtualenv installation

Deployment details

No response

Anything else?

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[RoyLee1224]

I can also reproduce this issue.

CleanShot.2025-10-08.at.11.13.50.mp4

I think the root cause is that useVirtualizer renders only the visible subset of parsedLogs.
When you select text across virtual boundaries, invisible logs simply don’t exist in the DOM, so the selection can’t continue.

[TheEZIC]

I think workaround possible in that case with virtualization. You can use Selection API to manage selection manually

[viveknanda]

I would like to work on this issue. Can it be assigned to me?

[NilsJPWerner]

@bbovenzi we run into a lot of issues like this one with the virtualized log viewer. Would it be feasible to have a toggle that just loads the whole log like airflow 2 did?

[Felix-neko]

Hi folks! Do you have any news about it?

[viveknanda]

@Felix-neko I am working on it, and should have something ready by this weekend

[viveknanda]

I have worked on implementing this approach, but load testing with > 10,000 lines of logs is causing jank and freezing up the tab when logs exceed 25,000 in my test step (obviously due to the no. of DOM elements). Showing such an option selectively, say when when the number of logs are much less may cause some confusion, if the option appears/disappears or is (dis)enabled conditionally.

A simpler approach is to increase the overscan count from the existing value 10 to maybe 250. In that ways, a little over 500 lines of logs will always be available in the virtualizer context view which can be copied to the clipboard. Anyone wanting to copy and analyze more than 500 lines of logs, is better off downloading the log file rather than debugging from the browser.

@bbovenzi , @jscheffl would you be able to suggest another approach, if at all this issue is important to be addressed?

[viveknanda]

In my further testing, setting a very high overscan value (like 250) is going to add to some performance while initial render. However, the bigger user experience problem it brings is long stutters/flashes when the user scrolls to a part of the view beyond overscan, as now the virtualizer has to process (create DOM nodes for) and bring in say ~50 lines of code in view and 250 above and under it too, which eats noticeable time.

Here's are gifs of scrolling through 25,000 lines of log with different overscan values

Overscan: 250

Overscan: 100

Overscan: 10

Keeping overscan to a not so high value like 50 will not really help with a lot of log copying.

Open to hear out to any ideas