1. Forum
  2. Usenet
  3. COMP.LANG.AWK

  • getchar implementation without GNUishms

    From anthk@21:1/5 to All on Tue Mar 25 09:51:54 2025
    Hello the guy from https://luxferre.top and gopher://hoi.st
    has pretty interesting stuff, such as

    git://git.luxferre.top/awk-gold-collection.git

    (Run
    git clone --recursive git://git.luxferre.top/awk-gold-collection.git
    to get them all)

    The issue is 'subleq.awk' uses read from GNU Bash I'd guess; thus,
    is not portable to sh/ksh. The flags are missing:

    function getchar(c, cmd) { # POSIX-compatible getchar emulation with sh read
    (cmd="c='';IFS= read -r -n 1 -d $'\\0' c;printf '%u' \"'$c\"") | getline c
    close(cmd)
    return int(c)
    }

    There's another one implemented at tgl.awk, some 'universal' library
    with mission functions for POSIX awk's. Not so universal,
    as it falls in the same traps: it depends on 'od' from
    GNU coreutils, with flags equallly missing:

    function getchar(c) {
    if(!TGL_GCH_CMD) TGL_GCH_CMD = "od -tu1 -w1 -N1 -An -v" # first time usage
    TGL_GCH_CMD | getline c
    close(TGL_GCH_CMD)
    return int(c)
    }


    Then I tried to it under C, both with a leading space and a
    newline and with just the char:

    int main()
    {
    int c;
    if ((c = getchar()) != EOF) {
    printf(" %d\n", c);
    /* printf("%d", c); */
    }
    }

    Then I edited both commands to be piped into getline to use
    my custom C program, but I had no luck.
    Subleq programs with no input work fine.
    Once I try a Forth implemented in subleq, no input
    is parsed right, as Forth doesn't eval a simple
    "2 2 + .s" instruction.

    subleq.fth and forth:

    https://github.com/howerj/subleq/

    The one in C works fine:

    ./subleq sublec.dec sublec.fth

    2 3 + .s
    5 ok

    Not the case with awk (OpenBSD, one true awk), mawk and gawk:

    awk -f subleq.awk ./subleq/subleq.dec ./sublec.fth

    sublec.fth has extra Forth words, thus it can be slower to
    parse under subleq.awk. But the core subleq.dec has a
    minimal implementation enough to do basic arithmetic.
    It works under the C implementation of subleq, again,
    but not under subleq.awk (outside GNU oses).

    Could it be possible to implement a true portable getchar?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Janis Papanagnou@21:1/5 to anthk on Tue Mar 25 14:10:49 2025
    On 25.03.2025 10:51, anthk wrote:
    Hello the guy from https://luxferre.top and gopher://hoi.st
    has pretty interesting stuff, such as

    [ shell specific or other tries to emulate some getchar function ]

    Could it be possible to implement a true portable getchar?

    Those who think that getchar is a useful function may implement that
    natively in Awk. (It avoids external dependencies and all the issues
    that the posted/quoted code has made obvious.)

    Janis

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Arti F. Idiot@21:1/5 to Janis Papanagnou on Tue Mar 25 22:16:32 2025
    On 3/25/25 7:10 AM, Janis Papanagnou wrote:
    On 25.03.2025 10:51, anthk wrote:
    Hello the guy from https://luxferre.top and gopher://hoi.st
    has pretty interesting stuff, such as

    [ shell specific or other tries to emulate some getchar function ]

    Could it be possible to implement a true portable getchar?

    Those who think that getchar is a useful function may implement that
    natively in Awk. (It avoids external dependencies and all the issues
    that the posted/quoted code has made obvious.)


    I know you can iterate over input strings a character at a time in AWK
    but I don't think you can read a single character from stdin without
    also providing a newline via ENTER, which is perhaps what the OP was
    actually wanting to do?

    I kind of like making CLI interactive stuff and like not having to
    press ENTER after entering single character menu choices. Tried a
    few options using bash and lua before landing on stty(1) and dd(1),
    both POSIX tools:

    --
    # getchar_posix.awk - read exactly one char from stdin and return it
    # without waiting for ENTER to be pressed.
    #

    BEGIN {
    printf "enter a char: "
    Char = getchar()
    printf "\n you entered: %s\n", Char
    }

    function getchar( ,Cmd, Chr) {
    # put TTY in "raw" mode..
    system ("stty -icanon")
    #
    # read Chr via dd(1)..
    Cmd = "dd bs=1 count=1 2>/dev/null"
    Cmd | getline Chr
    close (Cmd)
    #
    # put TTY in "normal" mode..
    system ("stty icanon")
    #
    return Chr
    }
    --

    Saving the TTY state via 'stty -g' beforehand would probably be a good
    addition so if things go sideways the TTY isn't a mess.

    -A

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Kenny McCormack@21:1/5 to Arti F. Idiot on Wed Mar 26 05:08:56 2025
    In article <vrvv30$spj$[email protected]>,
    Arti F. Idiot <[email protected]d> wrote:
    On 3/25/25 7:10 AM, Janis Papanagnou wrote:
    On 25.03.2025 10:51, anthk wrote:
    Hello the guy from https://luxferre.top and gopher://hoi.st
    has pretty interesting stuff, such as

    [ shell specific or other tries to emulate some getchar function ]

    Could it be possible to implement a true portable getchar?

    Those who think that getchar is a useful function may implement that
    natively in Awk. (It avoids external dependencies and all the issues
    that the posted/quoted code has made obvious.)


    I know you can iterate over input strings a character at a time in AWK
    but I don't think you can read a single character from stdin without
    also providing a newline via ENTER, which is perhaps what the OP was
    actually wanting to do?

    This works for me:

    --- Cut Here ---
    @load "call_any"
    BEGIN {
    system("stty raw -echo");
    s = " "
    n = call_any("iisi","read",0,s,1)
    system("stty sane");
    print "s =","|"s"|","n =",n
    }
    --- Cut Here ---

    Note: I could probably change the system() calls to use call_any() as well.

    It looks like tcgetattr(3) and tcsetattr(3) are the currently accepted ways
    of doing this. I've used those calls in my C programs, so would have to
    adapt that to work with call_any().

    --
    Liberals live in a fantasy world where (street) criminals are good people.

    Conservatives live in a fantasy world where businessmen are good people.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Kenny McCormack@21:1/5 to All on Wed Mar 26 05:48:06 2025
    In article <vs0258$ncqh$[email protected]>,
    Kenny McCormack <[email protected]> wrote:
    ...
    This also works for me:

    --- Cut Here ---
    BEGIN {
    cmd = "exec bash -c 'while :;do read -rsn1 < /dev/tty;echo $REPLY;done'"
    while (cmd |& getline)
    print "Result:",$0
    }
    --- Cut Here ---

    --
    The randomly chosen signature file that would have appeared here is more than 4 lines long. As such, it violates one or more Usenet RFCs. In order to remain in compliance with said RFCs, the actual sig can be found at the following URL:
    http://user.xmission.com/~gazelle/Sigs/Voltaire

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Janis Papanagnou@21:1/5 to Arti F. Idiot on Wed Mar 26 06:27:52 2025
    On 26.03.2025 05:16, Arti F. Idiot wrote:
    On 3/25/25 7:10 AM, Janis Papanagnou wrote:
    On 25.03.2025 10:51, anthk wrote:
    Hello the guy from https://luxferre.top and gopher://hoi.st
    has pretty interesting stuff, such as

    [ shell specific or other tries to emulate some getchar function ]

    Could it be possible to implement a true portable getchar?

    Those who think that getchar is a useful function may implement that
    natively in Awk. (It avoids external dependencies and all the issues
    that the posted/quoted code has made obvious.)

    I know you can iterate over input strings a character at a time in AWK
    but I don't think you can read a single character from stdin without
    also providing a newline via ENTER, which is perhaps what the OP was
    actually wanting to do?

    I don't know. - When I read 'getchar' I associated a character oriented function like awk's 'getline'. That would mean not reading from stdin
    with some I/O buffered return-terminated input but just processing the
    data as Awk would read its input from stdin or using 'getline' would.
    You need to maintain some state, though; here the actual read in line
    which acts like a buffer in buffered OS reads.

    What I associated was actually something like

    function getchar ()
    {
    if (_pos >= _len) {
    do {
    if ((getline _line) <=0)
    return ""
    }
    while (!(_len = length (_line)))
    _pos = 0
    }
    return substr (_line, ++_pos, 1)
    }

    used in contexts like

    BEGIN {
    # RS = "$^"
    while (c = getchar())
    print ">" c "<"
    }

    where the commented RS assignment could be activated in case you wanted
    to also read the newline characters, or, without it, to just read in
    the payload data of a line (or record). The necessary state information
    is stored in those variables that are named with a leading underscore.

    As said it's native Awk code without those dependencies on OS, on tools,
    or on specific non-standard flags of tools.

    (But as indicated in my previous response, I don't have any need for a
    function like that. Maybe others do, don't know.)

    Janis

    [...]

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)