If you press an incomplete key sequence, previously WeeChat would send
the key_pressed signal again for the same keys on the next key.
E.g. if you press escape and then 1, previously you would get the
key_pressed signal with signal_data `\x01[` when you pressed escape, and
then key_pressed with `\x01[` again when you pressed 1 (plus key_pressed
with `1` for the 1 key). So two signals for the escape key, even though
it was only pressed once.
With this patch, you only get one signal for each key press. So one with
`\x01[` when you press escape and then one with `1` when you press 1.
It seemed strange to me to have word_end_offset point to the last
character in the word, rather than the character after the word,
especially now with the word stopping before a newline character which
meant word_end_offset would be -1 if there was no characters before the
newline character.
The new rendering of multiline lines had some issues with colors at
certain positions not being applied. The color would not be applied if
the color code was at either of these positions:
- At the start of a line after a newline character
- At the end of a line after a space and before a newline character
- At a line by itself before a newline character
The way I had done it by considering newline characters as a word in
gui_chat_get_word_info with a variable specifying that it's newline
characters became messy and didn't really make sense, so rather than
doing this, I changed gui_chat_get_word_info to stop before the first
newline character. That way, we can just check if we are at a newline
character at the start of the loop, and don't need any more special
handling.
Fixes#1928
If you have a chat line with multiple lines that were scrolled so that
part of it is rendered at the bottom of the chat area, and part of it
falls below the chat area, there was a bug where the prefix_suffix would
be displayed at the end of the last line.
Looks like this happens because the gui_chat_display_prefix_suffix is
called for the line below the last line displayed in the chat area. So
we have to add a check for this so we don't add the prefix_suffix. This
check is copied from gui_chat_display_word which checks for the same.
Fixes#1912
If a chat line message contains a newline character (\n) it was
previously rendered as J with reverse video. This commit makes it
render as an actual newline instead, so messages with multiple lines
become supported.
The rendering is fixed in normal mode as well as bare mode both when
scrolled to the bottom and when scrolled up (which is different code
paths). Focus events has also been updated to support this (except for
_chat_line_y which returns -1 for all lines, but the docs says this
variable is only for buffers with free content).
Currently, the only way to include a \n in a chat line message is with
hdata_update because printf splits on \n and creates multiple separate
lines, but hopefully either printf can be changed to not split on \n, or
a new command which doesn't split can be added.
This changes the commands delete_beginning_of_line, delete_end_of_line,
delete_line, move_beginning_of_line and move_end_of_line to operate on
the current line instead of the whole input. The commands
delete_beginning_of_input, delete_end_of_input, delete_input,
move_beginning_of_input and move_end_of_input are added with the
previous implementations that the line commands had.
Additionally, the commands move_previous_line and move_next_line are
added which moves the cursor to the previous/next line and keeps the
horizontal position in the line.
The meta-r key is changed from delete_line to delete_input to keep the
behavior, and because you probably want to delete the whole input more
often than the line. The meta-R key is added for delete_line.
The home, end, ctrl-u and ctrl-k keys are kept to the same commands,
which means that they change behaviour. This is because having them
operate on the line is consistent with other applications (vim, zsh),
and I also think it's more practical.
These new bindings are added:
shift-home: /input move_beginning_of_input
shift-end: /input move_end_of_input
shift-up: /input move_previous_line
shift-down: /input move_next_line
meta-R: /input delete_line
meta-ctrl-u: /input delete_beginning_of_input
meta-ctrl-k: /input delete_end_of_input
Relates to #1498
Instead of replacing newline/tabs when paste is started, do it when the
paste is accepted instead. This makes a difference if you paste again
while the paste confirmation is active, where instead of running it
again for each paste, it will now be run for all the text at the end.
For now this doesn't make a practical difference, but the next commit
will remove the final newline when multiple lines are pasted too, which
we only want to do for the final paste.
This makes pasted text appear in the input bar, instead of each line
being sent. This allows you to edit the text before sending it, and it
makes multiline paste supported in buffers with input_multiline on.
It also replaces \r with \n in pasted text because most terminals (e.g.
xterm and urxvt) print lines separated by \r when pasting as if return
was pressed between each line, even though the copied text uses \n. The
text sent to the buffer should use \n, not \r, so we have to replace it.
Note that this only works when bracketed paste is enabled or the paste
confirmation as shown, because non-bracketed paste with no paste
confirmation is not detected as a paste.
Fixes a part of #1498
When in paste pending mode, the bracketed paste escape sequence should
still be interpreted, so that if you paste while in paste pending the
same things as when pasting in normal mode still happens, i.e. the
escape sequence is removed, ctrl-y/n is not interpreted and newline/tabs
are replaced.
Sébastien Helleu