To get to the bottom of the too-much-fading regression,
it was necessary to clean up the code, which was overly
complicated by multi-purposed functions.
The API for compose_fade now has these functions:
set_focused_recipient
start_compose
clear_compose
update_message_list
update_faded_messages
Internally there is now a notion of "normal display",
so e.g. when you want a normal display, we call
_diplay_messages_normally() internally, which removes the
faded/unfaded classes from all messages.
(imported from commit 7eb2b0a163f29d9ebae26661f432fecc7c331e4c)
This makes fade/unfade start sooner (good), but it might
re-introduce some typing sluggishness (bad).
(imported from commit 4e3112ed1ac931f2931182f91b60567ef2d72695)
When starting a compose, call compose_fade.set_faded_messages,
which will immediately do fade/unfade logic, whereas before
the code path went thru debouncing logic.
(imported from commit 7d0b30435be32a7132dbf05bf064b03b925a2d42)
Move code from compose.update_fade() into
compose_fade.set_focused_recipient(), which makes it
so that we only have to send the msg_type.
(imported from commit c17665d9f34f525bdedcd36d39d3a112fa36a914)
The code in unfade_messages() is O(N) over the number of
messages, but a simple flag allows us to track the fact that
all messages are unfaded, so we can short circuit the O(N)
logic in many cases.
A typical scenario now would be that you start typing a
stream while the topic is still empty. Modulo debouncing,
every keystroke now leads to a call to unfade_messages(),
but this change only does real work the first time.
(imported from commit da07cf408bbdbf5b381ff3ec33a5e05e34eef5b5)
The compose_fade has three public exports:
set_focused_recipient
unfade_messages
update_faded_messages
All code was pulled directly from compose.js, except for the
one-line setter of set_focused_recipient. The focused_recipients
variable that used to be in compose.js was moved to compose_fade.js,
hence the need for the setter.
(imported from commit 462ca5d0d0bd58612d0197f3734a8c78de8c6d30)