Process filters — line at a time?

Process filters — line at a time?

[Christopher Howard]

Hi, I maintain a legacy system that communicates over a 9600 7E1 RS-232 connection. I want to use Emacs to automate some things that normally involving hours of sitting in front of the terminal and a lot of keystrokes. Currently 29.4 is installed on the computer attached to the serial port.

I've made a good start using serial-term and serial-process-configure to setup my terminal, with the idea of using process-send-string to send commands, and a process filter to analyze the output. However, I need to analyze the output a line a time, to know what commands to send, and my understanding is that process filters do not necessarily give you a whole line at a time. I was wondering if there was some Emacs abstraction or approach that might be helpful here, so that I could focusing on matching lines of data, without getting all caught up in the intricacies of buffering and character processing. Or maybe some way to configure the process so that the process filter always delivers a line at a time.

-- 
📛 Christopher Howard
🚀 gemini://gem.librehacker.com
🌐 http://gem.librehacker.com

בראשית ברא אלהים את השמים ואת הארץ

Re: Process filters — line at a time?

[Eli Zaretskii]

> From: Christopher Howard <christopher@librehacker.com>
> Date: Wed, 28 Aug 2024 14:53:07 -0800
> 
> Hi, I maintain a legacy system that communicates over a 9600 7E1 RS-232 connection. I want to use Emacs to automate some things that normally involving hours of sitting in front of the terminal and a lot of keystrokes. Currently 29.4 is installed on the computer attached to the serial port.
> 
> I've made a good start using serial-term and serial-process-configure to setup my terminal, with the idea of using process-send-string to send commands, and a process filter to analyze the output. However, I need to analyze the output a line a time, to know what commands to send, and my understanding is that process filters do not necessarily give you a whole line at a time. I was wondering if there was some Emacs abstraction or approach that might be helpful here, so that I could focusing on matching lines of data, without getting all caught up in the intricacies of buffering and character processing. Or maybe some way to configure the process so that the process filter always delivers a line at a time.

One idea is to use string-lines to split the process output into
individual lines.  You will need to write custom code to identify
incomplete lines, though.  You will find examples of the usage in
shell.el and in compile.el.

Re: Process filters — line at a time?

[Christopher Howard]

> One idea is to use string-lines to split the process output into
> individual lines.  You will need to write custom code to identify
> incomplete lines, though.  You will find examples of the usage in
> shell.el and in compile.el.

Thank you, string-lines sounds helpful. In the case that there an incomplete line at the end, is there some way to give that back to the process output stream, or would I need to buffer that myself?

-- 
Christopher Howard

Re: Process filters — line at a time?

[Eli Zaretskii]

> From: Christopher Howard <christopher@librehacker.com>
> Cc: help-gnu-emacs@gnu.org
> Date: Thu, 29 Aug 2024 07:38:37 -0800
> 
> > One idea is to use string-lines to split the process output into
> > individual lines.  You will need to write custom code to identify
> > incomplete lines, though.  You will find examples of the usage in
> > shell.el and in compile.el.
> 
> Thank you, string-lines sounds helpful. In the case that there an incomplete line at the end, is there some way to give that back to the process output stream, or would I need to buffer that myself?

The latter, I believe.

Re: Process filters — line at a time?

[Stefan Monnier via Users list for the GNU Emacs text editor]

> Thank you, string-lines sounds helpful. In the case that there an incomplete
> line at the end, is there some way to give that back to the process output
> stream, or would I need to buffer that myself?

The usual way to do that is to keep the "unprocessed output" in
a process property (see `process-put` and `process-get`) or in a global
var, e.g.:

    (lambda (proc output)
      (setq output (concat (process-get proc 'foo--leftover) output))
      (let ((end (string-match ".*\\'" output)))
        (process-put proc 'foo--leftover (substring output end))
        (setq output (substring output 0 end)))
      ;; Now `output` contains only wholelines.
      ...)


- Stefan