Arithmetic expansion

[phy1729]

Needs docs and a changlog update (and perhaps a few more tests if any come to mind). One outstanding question is should "$(( ... ))" break the double quoted style as "$( ... )" does?

[phy1729]

Added docs, changelog entry, and some more tests. Ready for review.

Outstanding questions:

• Should "$(( ... ))" break the double quoted style as "$( ... )" does?
• Which style should command substitution inside arithmetic substitution use: -quoted, -unquoted, or no suffix?

[danielshahaf]

Needs docs and a changlog update (and perhaps a few more tests if any come to mind).

Thinking out loud:

• Things that look like commands, but aren't:

  • $((telnet * 42)) (multiplication)
  • $((ls /etc)) (division)
  • $((diff - foo)) (subtraction)
  • $((zsh +1 -f /etc/passwd)) (addition, subtraction, and division)
  • $((ls & ls)) (bitwise and)

• Things that look like arithmetic, but aren't:

  • $((ls ls)) (syntax error)
  • $(() function ()) (anonymous function)
  • $((echo case foo in bar\)); echo ';;') (subshell)
  • $((echo case foo in bar\)) (prefix of previous)
  • $((: ")) )(((" )) (what do you expect this to do, without running it? Arithmetic, subshell, or syntax error?)
  • $((: ") ))(((" )) (same question)

(I'm not saying that all of these be added; I'm just brainstorming.)

The rule is in lex.c.

Should "$(( ... ))" break the double quoted style as "$( ... )" does?

The criteria I would consider are:

• Consistency across different types of quoting.

• Forward compatibility with highlighting other forms of arithmetic expressions: e.g., subscripts, repeat loop counts, arithmetic for loops' control expressions.

• Which approach would give theme authors more options. (E.g., if the possibilities with approach A are a strict superset of the possibilities with approach B, that'd be an argument in favour of the former.)

Which style should command substitution inside arithmetic substitution use: -quoted, -unquoted, or no suffix?

What kind of processing (word-splitting, etc) will it undergo? If it's the same processing in "$(foo)" and in this case, then it could use the same highlighting.

The same goes for $foo, by the way. In $(($foo + $bar)), both variables are double-quoted, right? So it'd be nice to use the double-quoted variables style (never mind that we don't have have a style for non-double-quoted variables).

Ready for review.

Ship it :-)

[danielshahaf]

Triage: Ready to merge, and there's still enough time prior to tagging to find any issues: let's merge it! @phy1729, any reason not to? If not, would you do the honours?

[danielshahaf]

Noting here that #709, once merged, would benefit from integration to this. (Though it's not clear yet which PR will be merged first.)

[phy1729]

Thanks for the reminder. Merged.