Problem: Current window is checked to determine whether "more" window
is open. Making it the current window is scheduled in case the
cmdwin is open so this can be too late.
"cmdline_hide" may be emitted when the topline is
temporarily invalid (after incsearch->restore_viewstate()).
Solution: Use the window visibility to determine an active "more"
window instead.
Don't nvim__redraw->flush the "cmdline_hide" event (a normal
will already happen).
Problem: When opening a new tabpage, extui windows are initialized with
their default config. Window visiblity/dimensions on other
tabpages may get out of sync with their buffer content.
Solution: Copy the config of the window to the new tabpage.
No longer keep track of the various windows on each tabpage.
Close windows on inactive tabpages instead (moving them could
be more efficient but is currently not supported in the API).
Problem: Buffer setup after moving the message buffer to the "more"
window after ff95d7ff is incomplete.
Solution: Adjust and invoke tab_check_wins() to do the setup instead.
Problem: Closing the "more" window for an entered cmdwin in ff95d7ff9
still requires special casing to properly handle "more" visibility.
Solution: Reposition the "more" window instead; anchoring it to the
cmdwin.
Problem: Message lines from multiple message events that end up
spilling 'cmdheight' end up spread out over the cmdline
and "more" window.
Messages emitted as feedback to a typed :command (rather than
its sole purpose like :echo/:=) are routed to the more window.
The more window isn't closed when entering the cmdwin, and
doesn't allow `vim.hl.on_yank()`.
Solution: When first opening the "more" window for spilled messages,
move the message buffer to the more window.
Restrict routing of typed commands to echo kinds.
Ignore all events but WinLeave and TextYankPost.
Problem: An already open "more" window is the most convenient place
to route messages to with `cfg.msg.pos == 'cmd'`. Instead we
route to the cmdline (unless that message also exceeds
'cmdheight').
Solution: Route to "more" window while it is open and scroll to bottom
to show the newest message. This is more convenient and more
efficient due to not writing to the target buffer first (which
is done to calculate the height of the message).
(Ensure message highlights are deleted when text is replaced.)
Problem: Bulking message lines to write in a single API call is
complicated and still not correct w.r.t. overwriting
highlights.
Solution: Write each chunk pattern separately with it's highlight
such that it will be spliced correctly for message chunks
that contain a carriage return. Go with correctness over
performance until this proves to be too inefficient.
Also add an identifying name to the various extui buffers.
Problem: Using tracked message column as column in cleared buffer.
Solution: Ensure column does not exceed current line length.
Further work to ensure carriage return and cmdline block mode
work properly.
Problem: A "search_cmd" message alters the message state (since 8e8f4523),
unnecessarily affecting logic for messages that follow.
Solution: Reset the appropriate variables after a "search_cmd" message.
Don't show "search_cmd" message with zero 'cmdheight'.
Problem: <Tab> is not translated on the cmdline, and exposes a wrong
assumption in search messages that may contain multiple chunks.
Solution: Translate unprintable characters in the cmdline content.
Extract the 'search_count' from the last chunk and route
'search_cmd' to cmdline to handle multiple chunks.
Problem: Incorrect message highlighting for highlighted message
chunks containing carriage returns.
Solution: Add a highlight for each substring ending in a newline or
carriage return separately.
Problem: Hint message to press "g<" for message that spills
'cmdheight' is too intrusive.
Solution: Remove the hint message. Document the meaning of the spill
indicator instead.
Problem: Carriage return should place the "write cursor" at the start of a line.
Solution: Construct the to be printed line by iterating over carriage returns.
Problem: Decision whether message is sent to "more" window is based on
the number of newlines present in a message, rather than the
actual text height.
With the "box" target, messages that come from a cmdline
entered command are not always routed to the more window.
Solution: Still write the message to the target buffer, and calculate
the actual text height. Postpone updating several state
variables until after the decision to re-route is made.
With the "box" target, only consider the text height of the
message if it is not a message from a cmdline entered command.
Problem: Extui shows a spill indicator to hint to the user to press "g<"
to show the output of the last command. The user may be
unaware of this mapping.
Solution: Route a hint message to the message "box" window.
Problem: Cannot enter "more" window while cmdwin is open since it is a
regular window and changing windows is disallowed.
Solution: Close cmdwin to enter the "more" window. Minor regression
w.r.t. the current message grid. Resolving that will require
somehow bypassing textlock for the "more" window.
Problem: Extui does not route messages emitted as a result of a typed
command to the "more" window.
Command message leading shell messages is missing a kind.
Messages not routed to 'cmdheight' area after it was 0.
Solution: Route messages that were emitted in the same event loop as an
entered command to the "more" window. Also append multiple
messages in an already open more-window.
Assign it the `shell_cmd` kind.
Change message position when 'cmdheight' changes from 0.
Problem: The "box" window width is calculated on the last line when
applying "last" virt_text. There may be longer lines
part of the same message.
Solution: Don't decrease box width if the calculated width is smaller.
Minor unrelated changes.
Co-authored-by: Eike <eike.rackwitz@mail.de>
Problem: More-window is left visible after leaving it. Cursor may be
hidden behind it and it is hard to re-enter afterwards.
Solution: Close the more-window when another window is entered.
Problem: The cmdline popupmenu is hidden behind extui windows.
'showmode' message is drawn over the cmdline.
Changing the 'cmdheight' to accommodate space for the text in
the cmdline may change the current cursor position.
Solution: Ensure kZIndexMessages < zindex < kZIndexCmdlinePopupMenu.
Clear the 'showmode' message when the cmdline level is negative.
Temporarily set 'splitkeep' = "screen" when changing the 'cmdheight'.
Problem: We have an unmaintained Vimscript parser and cmdline
highlighting mechanism, with which it is hard to leverage the
treesitter highlighter. Long messages result in a hit-enter-prompt.
Solution: Implement a vim.ui_attach() UI, that replaces the message
grid (orphaning some 3000+ LOC core C code). Introduce an experimental
vim._extui module, because removing the message grid at the same time is
too risky. The new UI leverages the bundled treesitter highlighter and
parser for Vimscript, as well as the matchparen plugin, to highlight the
cmdline. Messages are truncated in the cmdline area, or placed in a
floating message box in the bottom right corner. Special ("list_cmd")
messages and the message history are shown in a, "more prompt" (now a
fully interactive regular window). Various default UI elements ('showcmd',
'ruler') are also placed in the cmdline area, as virtual text.
`require('vim._extui').enable({})` enables the experimental UI.
`{ msg.pos = 'box' }` or `:set cmdheight=0` enables the message
box variant.
Followup:
- Come to a consensus for how best to represent messages (by default).
- Start removing message grid when this is deemed a successful replacement.
When that is finished, make this new UI the default and update a lot of tests.
