libervia-backend/libervia/cli/call_tui.py

#!/usr/bin/env python3

# Libervia CLI
# Copyright (C) 2009-2024 Jérôme Poisson (goffi@goffi.org)

# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.

# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU Affero General Public License for more details.

# You should have received a copy of the GNU Affero General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.


import asyncio
from functools import partial
import sys

from PIL import Image
import gi
from rich.padding import Padding
from term_image import image as t_image

from libervia.cli.constants import Const as C
from libervia.frontends.tools import aio, webrtc
from libervia.frontends.tools.webrtc import CallData, WebRTCCall

from .call_simple import BaseAVTUI

gi.require_versions({"Gst": "1.0", "GstWebRTC": "1.0"})

from gi.repository import Gst


aio.install_glib_asyncio_iteration()


class AVCallUI(BaseAVTUI):
    def __init__(self, parent):
        super().__init__(parent.host, align="center")
        self.parent = parent
        self.image_rows = 0
        self.buffer = None
        self._processing = False
        self.render_class: t_image.ImageMeta | None = None
        self.target_size
        self.fps = 25
        for oo in parent.args.output_opts:
            if oo.startswith("renderer="):
                renderer = oo[9:].lower().strip()
                match renderer:
                    case "auto":
                        # we let None as it will auto-detect best option later
                        pass
                    case "block":
                        self.render_class = t_image.BlockImage
                    case "iterm2":
                        self.render_class = t_image.ITerm2Image
                    case "kitty":
                        self.render_class = t_image.KittyImage
                    case _:
                        parent.parser.error(f"Invalid renderer: {renderer!r}")
            elif oo.startswith("fps="):
                try:
                    self.fps = int(oo[4:])
                except ValueError:
                    parent.parser.error(f"Invalid FPS: {oo[4:]!r}")

    async def init_call(self, call_data):
        kwargs = self.parse_output_opts(self.parent)
        if "target_size" not in kwargs:
            # we use low res by default for performance reason
            kwargs["target_size"] = (640, 380)
        webrtc_call = await WebRTCCall.make_webrtc_call(
            self.parent.host.bridge,
            self.parent.profile,
            call_data,
            sinks=webrtc.SINKS_APP,
            appsink_data=webrtc.AppSinkData(
                local_video_cb=partial(self.on_new_sample, video_stream="local"),
                remote_video_cb=None,
            ),
            merge_pip=True,
            # we want to be sure that call is ended if user presses `Ctrl + c` or anything
            # else stops the session.
            on_call_setup_cb=lambda sid, profile: self.parent.host.add_on_quit_callback(
                self.parent.host.bridge.call_end, sid, "", profile
            ),
            on_call_ended_cb=lambda sid, profile: self.parent.host.a_quit(),
            **kwargs,
        )
        self.webrtc = webrtc_call.webrtc

    async def start(self, call_data):
        term_rows = self.host.console.size.height
        if term_rows < 20:
            self.host.disp(
                "Your terminal must have a height of a least 20 rows.", error=True
            )
            self.host.a_quit(C.EXIT_ERROR)
        self.image_rows = term_rows - 6
        self.image_cols = self.host.console.size.width
        await self.init_call(call_data)
        assert self.webrtc is not None

        idx = 0
        self.buffer = ""

        # we detect render
        if self.render_class is None:
            self.render_class = t_image.auto_image_class()

        loop_sleep = 1/self.fps

        with self.input.raw_mode():
            # for whatever reason, using self.input.attach is breaking KittyImage and uses
            # a BlockImage style rendering instead. So we don't use it and we call
            # ``self.keys_ready()`` ourself in the loop, below.
            # cursor is not restored despite the ``screen`` context if peer is hanging up,
            # so we reactivate cursor here
            self.host.add_on_quit_callback(self.host.console.show_cursor, True)
            with self.host.console.screen():
                while True:
                    idx += 1
                    if self.buffer is not None:
                        sys.stdout.write(self.buffer)
                        sys.stdout.write("\n")
                        self.parent.console.print(
                            # the padding avoid artifact when toggling buttons
                            Padding(self.generate_control_bar(), (0, 2, 0, 0))
                        )
                        sys.stdout.flush()
                        rendered = True
                    else:
                        rendered = False
                    await asyncio.sleep(loop_sleep)
                    self.keys_ready()
                    if self.done.is_set():
                        break
                    if rendered:
                        # we put cursor back at the top of image to print the next frame
                        # FIXME: we use +4 for the controls because we know the height of the
                        #   renderable, but it would be better to measure it dynamically.
                        sys.stdout.write(f"\033[{self.image_rows+4}A")

        await self.webrtc.end_call()
        await self.host.a_quit()

    @classmethod
    async def run(cls, parent, call_data: CallData) -> None:
        ui = cls(parent)
        await ui.start(call_data)

    def on_new_sample(self, video_sink, video_stream: str) -> bool:
        if self._processing:
            # image encoding for terminal is slow, if one is already processing, we don't
            # bother going further
            return False
        sample = video_sink.emit("pull-sample")
        if sample is None:
            return False

        video_pad = video_sink.get_static_pad("sink")
        assert video_pad is not None
        s = video_pad.get_current_caps().get_structure(0)
        stream_size = (s.get_value("width"), s.get_value("height"))
        buf = sample.get_buffer()
        result, mapinfo = buf.map(Gst.MapFlags.READ)
        if result and self.render_class is not None:
            self._processing = True
            image_data = mapinfo.data
            image = Image.frombuffer("RGB", stream_size, image_data, "raw", "RGB", 0, 1)
            img_renderer = self.render_class(image, height=self.image_rows)
            img_fmt = f"<{self.image_cols}.^1"
            if self.render_class == t_image.KittyImage:
                # we don't do compression to speed up things
                img_fmt += "+Wc0"
            self.host.loop.loop.call_soon_threadsafe(
                self.update_sample,
                sample,
                stream_size,
                video_stream,
                format(img_renderer, img_fmt),
            )
            self._processing = False

        buf.unmap(mapinfo)

        return False

    def update_sample(self, sample, stream_size, video_stream: str, buffer) -> None:
        if sample is None:
            return

        if video_stream == "remote":
            return

        self.buffer = buffer