concept for GNU BASH syntax highlighing in python using linux procFS by Ari Archer

bash_syntax_highlighting.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""concept for GNU BASH syntax highlighing in python

author : Ari Archer <[email protected]> / <https://ari-web.xyz/>
license : GPLv3

this concept works ... almost, theres a problem, we need to somehow
take ownership of the stdin fd, but we cannot, the reason we need to take
ownership or priority over that fd is to make sure were the only ones reading
it and not giving bash a chance to steal a byte, currently bash sometimes,
especially if youre writing fast, steals a read() from us meaning we miss a byte.
this script use the linux procFS

theres also an issue with ansi, but its not as important as of now, the script is
also quite messy, but i dont have the motivation to fix it rn, so deal with it

if you feel like it -- comment with ideas

stuff i tried so far :

- using LD_PRELOAD to overwrite the read() syscall
    - making always return 0
    - redirecting it to a FIFO
    - closing it
- using os.write / read
- using C++
- using C"""

import distutils.spawn
import os
import signal
import sys
from warnings import filterwarnings as filter_warnings

FDS: list[int] = []


def open_fd(fd: int, mode: int) -> int:
    nfd: int = os.open(
        f"/proc/{sys.argv[1]}/fd/{fd}", os.O_RDWR | os.O_CREAT, mode=mode
    )
    FDS.append(nfd)
    return nfd


def cleanup_fds() -> None:
    for fd in FDS:
        os.close(fd)


def interpret_backspace(string: str) -> str:
    new_string: str = ""

    for char in string:
        if char == "\x7f":
            new_string = new_string[:-1]
        else:
            new_string += char

    return new_string


def highlight_line(line: str) -> str:
    if not line:
        return ""

    line_cmd: str = line.split(maxsplit=1)[0]
    return f"\033[{31 + bool(distutils.spawn.find_executable(line_cmd))}m{line_cmd}\033[0m{line.removeprefix(line_cmd)}"


def logread(fd: int) -> bytes:
    print("reading ...")
    return os.read(fd, 1)


def main() -> int:
    """entry / main function"""

    if len(sys.argv) < 2:
        print(
            "no target bash PID supplied, you can get it by running `echo $$` in your target bash shell instance"
        )
        return 1

    stdin, stderr = open_fd(0, 0o400), open_fd(2, 0o200)
    out = open(f"/tmp/{sys.argv[1]}.bash", "w", buffering=1)

    out.write("")
    out.flush()

    while os.fstat(stdin):
        s: bytes = b""
        c: bytes = b""

        os.write(stderr, b"\033[s")

        while (c := logread(stdin)) not in b"\r\n":
            print(f"{c = }; {s = }")
            s = interpret_backspace((s + c).decode()).encode()

            os.write(stderr, ("\033[u\033[K" + highlight_line(s.decode())).encode())

        print("newline", repr(c))
        os.write(stderr, b"\n")

        if s:
            out.seek(0)
            out.write(s.decode())
            out.flush()

            print("running", s)

        os.kill(int(sys.argv[1]), signal.SIGINT)

    cleanup_fds()
    out.close()

    return 0


if __name__ == "__main__":
    assert main.__annotations__.get("return") is int, "main() should return an integer"

    filter_warnings("error", category=Warning)
    raise SystemExit(main())

• use export PROMPT_COMMAND="source /tmp/$$.bash" if youre using this, if you alrd have it set -- export PROMPT_COMMAND="$PROMPT_COMMAND; source /tmp/$$.bash"