Element-web: Navigation in the text editor is inconvenient

[rkfg]

Description

Long text navigation with arrow keys is mixed with the history and chat navigation. For example, I copy a big piece of text from somewhere else and I want to get to the very start of it to insert '>' and make it look like a quote or ``` if it's a piece of code.

Steps to reproduce

• press Home but it only moves the cursor to the beginning of the current line
• press Ctrl+Home but instead the chat window scrolls to top
• PageUp/PageDown don't work in the editor and scroll the chat window instead.
• press and hold Up but after the cursor gets to the first line the previous messages pop up (I expect this to happen on Ctrl+Up/Down as it's implemented in Gajim and Psi, for example)

So it's quite inconvenient to work with long text. I usually don't need to scroll the chat window from keyboard and expect all the usual shortcuts to work in the text editor only. One way I found to get to the beginning of the text is to press Ctrl+Shift+Home and then Left to deselect the text. I think it's better to either make such mixed navigation optional or apply it to the chat window only if it's explicitly focused.

Version information

• Platform: web (in-browser), same applies to the desktop version
• Browser: Chrome 60.0.3112.78
• OS: Debian testing amd64
• URL: https://riot.im/develop/ 4b8764c07074-react-5752345b8082-js-bdb90b4b334f

[lampholder]

Optimal composer behaviour will always be difficult, since different people have different habits and conventions.

The best we can do is:

• honour established conventions where they exist
• where there's ambiguity:
  
  
  
  • start off by providing whatever is expected by the majority of our users
  
  
  • consider adding config options
  
  

Regarding this specific issue:

• Home - I believe this universally takes you to the start of the line, so we should keep doing that
• Ctrl+Home - This usually takes you to the start of the document - I find our current behaviour surprising, so I'd like us to consider changing it
• PageUp/PageDown - I think this makes sense to be attached to timeline - if people are often typing such large messages that they need to page up/down through them then we might consider supporting the composer's expanding to a 'large message composer' which is more optimal for this use case (potentially arresting control of PageUp/PageDown when open)
• ↑ - I'm used to bash/irssi/weechat so I expect ↑ to take me to the previous message. I'm not familiar with Gajim/Psi - if lots of people only expect to move through history on Ctrl+↑/↓ then this might be worth a config option?

[lampholder]

This issue is a good place for some discussion, though if we decide we do want to tackle any of these let's split out a separate issue.

[rkfg]

I'm used to bash/irssi/weechat so I expect ↑ to take me to the previous message.

It's fine if the messages you usually type are short. However, if they're multiline it's pretty easy to accidentally jump to the previous message while you only wanted to skim through the current message. IRC doesn't generally allow multiline messages so it's not much of a problem there but there's no such limitation in any modern messenger. I agree, this should be configurable.

As for the other notes I mostly agree with you, I just tried every possible way to do what I wanted. Basically, making Ctrl+Home/End to work inside the text editor only and allowing to scroll through the history with Ctrl is all I want.

[jcsteh]

I get bitten on a daily basis by control+home/control+end not working as expected in the editor (i.e. scrolling instead of jumping to start/end of editor message). I'm a screen reader user (and thus keyboard only user), and while this isn't screen reader specific, I wonder if it impacts keyboard only users more; i.e. perhaps most users just use the mouse to jump to the start/end of multi-line messages.

[jryans]

I have extracted the control+home/end portion to a separate issue https://github.com/vector-im/riot-web/issues/12438, as that seems like the most critical problem here.