Fixes#5932
This updates our core terminal internals to track [semantic prompts
(OSC133)](https://gitlab.freedesktop.org/Per_Bothner/specifications/blob/master/proposals/semantic-prompts.md)
in a way that can enable us to fully implement the specification. This
PR makes our terminal state fully handle all specified sequences
properly, but we still ignore some noted options (`aid`, `cl`, etc.)
that are not critical to functionality.
I took a look at our shell integrations and they should still be
accurate and correct with regards to their OSC133 usage.
This doesn't introduce any new functionality, but should result in
existing functionality being more correct, more robust, and gives us the
groundwork to do more with semantic prompts in the future.
> [!NOTE]
>
> This PR description is **fully human written.** I realize its layout
and use of headers may trigger AI smells. But it's all me, I promise. :)
## Extensions
I kept support for the Kitty `redraw=0` extension: if an OSC133A command
sets `redraw=0` then we will NOT clear the prompt on resize (assuming
the shell can NOT redraw the prompt). I don't actually know any scripts
that utilize this, so we may want to just throw it away, but we had
support before and it's cheap to keep around and it's tested.
## Limitations, Missing Features
* We don't currently respect `aid` so nested semantic zones can break
things. This would require significant rework since aids are arbitrary
strings. It'd require managed memory on pages to store the aids. Fwiw,
in my brief survey of popular mainstream terminals, I wasn't able to
find any that respect this. Maybe Warp does since semantic prompts are
so core to their product but its not OSS for me to verify.
## Internal Changes
### `terminal.Page`
The internal state now tracks the current _semantic content type_ via a
2-bit flag on each `Cell`:
* `output` (zero value) - Command output
* `input` - Prompt input, typed by the user
* `prompt` - Prompt itself, e.g. `user@host $`
In addition to this, each `Row` also has a 2-bit flag. This flag _only
tracks if the row might contain a prompt_:
* `no_prompt` (zero value) - No prompt cells
* `prompt` - Prompt cells exist and this started the prompt (OSC133 A)
* `prompt_continuation` - Prompt cells exist and this continued a prior
prompt (OSC 133 A `k=s` or `k=c`)
The row flags server two functions:
1. An optimization to make prompt scanning faster
2. The only place the existence of continuation vs primary prompt is
stored since cells themselves only store `prompt`
### `terminal.PageList`
The PageList has two new functions:
* `promptIterator` - An iterator that yields `Pin` values that map to
the _first line_ of a prompt.
* `highlightSemanticContent` - Returns a highlight (pin range) to map to
specific semantic content for a given prompt pin. This lets you quickly
grab stuff like the prompt lines, content lines, etc.
### `terminal.Screen`
* The Screen now has a flag that tracks whether we've seen any row
trigger a semantic prompt. This can be used as a fast path to avoid
expensive semantic prompt operations if we know our screen can't
possibly contain any. We don't currently use this but I plan to.
* **Big one: clearing the prompt lines is now built-in to resize.** This
makes our resize aware of semantic prompts which makes it work _so much
better_ already.
### `terminal.Terminal`
* A new function `semanticPrompt` for handling all the semantic prompt
operations.
**AI Disclosure:** AI was used in various ways throughout this. It is by
no means 100% AI-written, it's mostly human written. I reviewed
everything. AI was used mostly for some assistance on rote tasks.
Nu properly marks input areas with OSC 133 B, but if it spans multiple
lines it doesn't mark the continuation lines with the k=s sequence.
Our prompt redraw logic before only cleared explicitly designated prompt
lines. But if the input line is multi-line and the continuation lines
are not marked, those lines would not be cleared, leading to visual
issues on resize.
To workaround this, we assume that if the current cursor semantic
content is anything other than "command output" (default), then we're
probably at a prompt line and should clear from there all the way up.
This works around Fish (at least v4.2) having a non-compliant OSC133
implementation paired with not having the hooks to fix this via shell
integration. We have to instead resort to heuristics in the terminal
emulator. Womp, womp.
The issue is that Fish does not emit OSC133 secondary prompt (`k=s`)
markers at the beginning of continuation lines. And, since Fish doesn't
provide a PS2-equivalent, we can't do this via shell integration.
We fix this by assuming on newline (`\n`) that a cursor that is already
painting prompt cells is continuing a prior prompt line, and
pre-emptively mark it as a prompt line. But this has two further issues
we have to work around:
1. Newline/index (`\n`) is one of the _hottest path_ functions in
terminal emulation. It sucks to add any new conditional logic here.
We do our best to gate this on unlikely conditions that the branch
predictor can easily optimize away.
2. Fish also emits these for auto-complete hints that may be deleted
later. So, we also have to handle the scenario where a prompt is
continued, then replaced by command output, and fix up the prompt
continuation flag to go back to output mode.
Point 2 is ALMOST automatically handled, because Fish does emit a `CSI J`
(erase display below) to erase the auto-complete hint. This resets all
our rows back to output rows. **Unfortunately**, Fish emits `\n` before
triggering the preexec hooks which set OSC133C. So we get the newline
logic FIRST (sets the prompt line), THEN sets the output cursor. If they
switched ordering here everything would just work (with the one
heuristic). But now, we need two!
To address this, I put some extra heuristic logic in the OSC133C
(output starting) handler: if our row is marked as a prompt AND our
cursor is at x=0, we assume that the prompt continuation was deleted
and we unmark it.
I put the heuristic logic dependent on OSC133C because that's way colder
of a path than putting something in `printCell` (which is the actual
hottest path in Ghostty).
We could get more rigorous here by also checking if every cell is empty
but that doesn't seem to be necessary at this time for any Fish version
I've tested. I hope thats correct.
I'd really love for Fish to improve their OSC133 implementation to
conform more closely to the terminal-wg spec, but we're going to need
these workarounds indefinitely to handle older Fish versions anyway.
Insert OSC 133 A k=s marks after each newline in PS1, so that all lines
following the first are marked as secondary prompts. This prevents ghostty
from erasing leading lines during terminal resize.
I'm going to keep this PR description brief, because there's a lot more
work I want to do here. And when I do it I'll talk more about it. But
this basically rewrites the inspector. It has almost all the same
functionality (missing one from the old inspector -- cell picking). But
it is organized in a much cleaner way and the memory management is also
a lot more clean. We also expose a LOT more details about things like
PageList and screens.
There are plenty of bugs and polish issues here. But going to merge this
because the inspector isn't a hot path item for people and a review of
this is not really possible. Plus it paves the way to more debug
overlays.
