Repetition with whitespaces as separator

[sunng87]

This might be related to #271 and it's a real world case in handlebars

consider the rules

WHITESPACE = _{ " " }
name = @{ "_"? ~ ('a'..'z')+ }
param = { name }
hash = { name ~ "=" ~ name }
v = { hash | param }
exp = { "{{" ~ name ~ v+ ~ "}}"}

The rule exp works great to match expression like {{echo a b c d=1}} until there is an invalid name within it, like {{echo a b c_d}}. Note that in our name rule, underscore is only allowed as a leading character. However, in this case it still matches exp, with c and _d parsed as two params. What I expect is a failure of matching.

So I changed exp to below to ensure whitespaces between params.

exp = { "{{" ~ name ~ v ~ (WHITESPACE+ ~ v)* ~ "}}"}

However, it fails to match basic case like {{echo a b}}, and says

 --> 1:8
  |
1 | {{echo a b}}
  |        ^---
  |
  = expected hash

which is confusing. So before #271 landing (seems never to happen) , is there any workaround for my case?

[xelivous]

This is more or less how it works as I see it:

WHITESPACE = _{ " " }

name = _{ (ASCII_ALPHANUMERIC | "_")+ }

command = @{ name }
param = @{ name }
value = @{ name }
hash = { param ~ "=" ~ value }

parameter = _{ hash | param }

exp = { "{{" ~ command ~ parameter+? ~ "}}" }

all = { (exp ~ NEWLINE?)+ }

If you swap the ruleset over to "all" it will match all of these expressions:

{{ echo a b c d=1 }}
{{echo a b=2 c_d = 39 }}
{{tes__t}}

The resulting parse tree is:

- all
  - exp
    - command: "echo"
    - param: "a"
    - param: "b"
    - param: "c"
    - hash
      - param: "d"
      - value: "1"
  - exp
    - command: "echo"
    - param: "a"
    - hash
      - param: "b"
      - value: "2"
    - hash
      - param: "c_d"
      - value: "39"
  - exp > command: "tes__t"

If you add a $ before the brackets for hash (hash = ${ param ~ "=" ~ value }) then you no longer allow whitespace in-between the = signs. AKA

{{echo a b=2 c_d = 39 }} <- this fails
{{echo a b=2 c_d=39 }} <- this is fine

And you can further customize it by using (hash = ${ param ~ WHITESPACE? ~ "=" ~ WHITESPACE? ~ value }), which has optional whitespace on both sides of the = sign. If you only wanted specific kinds of whitespace on a specific kind that is possible as well. Now all of the examples supplied are parsed again.