Re: A ZLE widget for calculator

On 2026-02-07 03:51, Bart Schaefer wrote:
>> Changing the reorded history compared with what's actually executed
>> seems a reasonable thing to have, given all the other things you can do
>> with history.
> 
> I'm a little confused by this one:
> 
> On Mon, Feb 2, 2026 at 11:22 PM Vincent Bernat <[email protected]> wrote:
>>
>> _vbe_calc_accept_line() {
>>       if [[ $BUFFER =~ "= *" ]]; then
>>           local expr=${BUFFER#= }
>>           zle -I
>>           command numbat -e "$expr"
>>           print
>>           print -s $BUFFER
>>           BUFFER=""
>>       else
>>           zle .accept-line
>>       fi
>> }
> 
> The original problem statement as I understand it, is: When the buffer
> starts with "= ", form a single quoted word to pass as an argument to
> "numbat -e", but preserve the original buffer in the shell history.
> 
> The first implementation, which modifies $BUFFER and then calls
> .accept-line, un-quotes and re-quotes the tail of the buffer.  You
> said:
>> a bit clunky to detect when something was quoted and undo it
> The implementation above never handles the un-quote part.  If that was
> required in the first implementation, why is that no longer necessary
> there?  Won't unwanted extra quotes that appear in the original
> $BUFFER be passed through to numbat via "$expr"?

On the first implementation, BUFFER was modified then executed. On the 
second implementation, the numbat command is executed directly and 
BUFFER is pushed to history (with print -s) without any alteration. No 
call to accept-line.

> I think the attached does what you want.  It's based on the
> modify-$BUFFER + .accept-line variation because among other things
> that has the advantage of working at the PS2 prompt.  It
> unconditionally removes and re-adds quotes.  The original $BUFFER is
> saved in a global variable which is then added to the history by the
> preexec hook, after suppressing the automatic history with
> zshaddhistory hook.  This could probably use a function instead of an
> alias for "=", but I didn't change that.

Thanks, it works perfectly!