^W in REPL behaves weird

[bnoordhuis]

Or maybe I should say "unexpectedly" or "non-idiomatically".

Typing o.f === followed by ^W deletes until o. - the expected result is o.f

qjs > var o = {x: 42}
undefined
qjs > o.f === ^W
qjs > o.

(Also, why is there a blank between "qjs" and ">"?)

[bnoordhuis]

Thinking about it more, it should behave like vi does:

function f(x) {  // comment  <cursor>
function f(x) {  // <cursor>
function f(x) { <cursor>
function f(x) {<cursor>
function f(x) <cursor>
function f(x<cursor>
function f(<cursor>
function f<cursor>
function <cursor>
<cursor>

A little more complicated than what we have now but hopefully not too difficult.

[saghul]

I guess we need to delete until the previous space, right?

(Also, why is there a blank between "qjs" and ">"?)

IIRC that was the case from the beginning?

[chqrlie]

(Also, why is there a blank between "qjs" and ">"?)

IIRC that was the case from the beginning?

I admit that is my personal preference, I use PS1="\u@\h \w > " for the shell prompt.

[bnoordhuis]

I guess we need to delete until the previous space, right?

I think the logic to a first approximation is:

1. determine class of character under cursor: blank, alphanumeric, everything else
2. delete all characters of that class until you hit one of a different class

[chqrlie]

Regarding the Ctrl-W behavior, it deletes the last word in the terminal, and words there are just any sequence of non blank characters. In programming environments, a word is more likely a sequence of characters that are part of an identifier, ie: letters, digits, _ and $, but not punctuation marks, so delete-word-backward first skips back over non word characters, then skips back over word characters and deletes both sequences. Meta-Backspace works this way in emacs, I am not familiar with vi bindings. The default terminal bindings are mostly those of emacs, Ctrl-W and a few other ones are exceptions.