The settings module sees that we're reading a number, so it converts
the value to a number. Unit detection needs a string, though, so
convert it back into one in that case.
This is to improve performance when switching to a buffer with a large
number of unread lines.
Fixes#859
Thanks to @lorenzhs for the comments on PR #862.
This introduces ability to pin buffers using localvar storage in
WeeChat. At a later stage some UI should be introduced for this, like
discussed in #604 .
There is a bug in WeeChat when hdata is null, which happens when the
hotlist is empty (meaning all buffers are read). Our websocket callback
machinery expects every command with id to return data, so this code in
current versions of WeeChat will lead to a slow leak. But I think lots
of things in our code will do this already so I'm not entirely sure it's
too problematic to let this patch go by.
We could use infolists instead of hdata, but that is cumbersome to parse
and less performant for WeeChat, and sends more data over the wire.
I propose we make a separate attempt at cleaning up the callbacks. Since
we store time on them we could have a cleanuptask that looks through
them and deletes old callbacks. Maybe @dcormier could have a look?
The Asciinema cast plugin regex was faulty. Casts can have /[a-z]/ characters, not only digits /\d)
Here is a test cast for you to run: https://asciinema.org/a/29qfl1dwsgd25o91nfv3dvvbn
(squashed two commits --@lorenzhs)
Functionality is now replaced by changing
weechat's buffer_time_format. e.g. 24h time:
/set weechat.look.buffer_time_format %H:%M
Supported time formats are 12h and 24h,
with seconds or no seconds.