[RFC] Behavior of range syntax in Eshell

[Jim Porter]

Recently in bug#60999, I added support for range syntax in Eshell to let 
you get sub-sequences:

  ~ $ echo $exec-path[1..3]
  ("/usr/local/sbin" "/usr/local/bin")

However, as you can see above, this is a half-open range. Bash, on the 
other hand, uses a closed range:

  ~ $ for i in {1..3}; do echo $i; done
  1
  2
  3

I opted for a half-open range to match the rest of Emacs Lisp, because 
that's how functions like 'substring' work. On the other hand, Eshell is 
a shell-like interface, so maybe using the Bash meaning would be less 
confusing for users. Does anyone on this list have any strong opinions? 
As I see it, there are three options:

----------------------------------------

1. Half-open (the status quo)

I have a soft preference for this, since Eshell lets you call any Emacs 
Lisp function as a command, so by using half-open ranges, we remain 
consistent with Emacs Lisp functions. I also find half-open ranges to be 
a little easier to use. For example, to truncate one list to be the same 
length as another, I could type:

  $long-list[0..$#short-list]
  ;; or...
  $long-list[..$#short-list]

With closed ranges, I'd have to type:

  $long-list[..${1- $#short-list}]

That's not a huge difference, but I think it's less convenient.

----------------------------------------

2. Closed (Bash-style)

This has the benefit of being consistent with Bash (and Zsh, I believe; 
probably other shells too), so users who are used to those shells will 
have less to learn in order to be comfortable with Eshell. However, this 
might set them up for confusion later down the road if they try to treat 
a function like 'substring' as having a closed range.

----------------------------------------

3. Use a different delimiter

To make it more obvious that Eshell ranges have a different meaning from 
Bash ranges, we could use a different delimiter than "..". However, we 
need to be careful here: If the variable 'foo' is a string, you can 
split the string with a regexp delimiter in Eshell via: $foo[my-regexp]. 
That means that we'd want the range syntax to be hard to mistake as a 
regexp; for example, ":" would be a bad range delimiter, since $PATH[:] 
would be a common way for users to split the $PATH env var[1]. ".." is a 
valid regexp, but it's not very useful to use for splitting a string 
with, so it's probably safe in this regard.[2]

----------------------------------------

If anyone has any thoughts or other ideas about this, let me know. Like 
I said, I have a soft preference for the status quo, but if others have 
a strong opinion about changing this, I don't mind. The way I see it 
though, Eshell is a part of Emacs, and you can write commands as 
ordinary Elisp, so following Elisp conventions makes the most sense to me.

If you need any more details on the current syntax, you can check the 
Eshell manual. Hopefully I've described it reasonably thoroughly.

[1] In practice, you'd probably just want to use $exec-path in Eshell, 
which is already a list. But this applies to things like $CPATH, etc too.

[2] If you really wanted to split on "..", you'd just need to surround 
it with quotes so Eshell doesn't interpret it as a range.