Missing output for optional nodes about canopy HOT 1 OPEN

The key thing to remember is that the structure of the output mirrors the
structure of the grammar. In your first grammar you have these rules:

query <- (expression @" "? expr_op? @" "?)+
expression <- negation? @" "? lvalue @" " operator @" " rvalue

You try to parse this string:

'not distance is 20 or not date is today and not date is tomorrow'

And you get this sequence of elements from the root node of the tree:

not distance is 20 or
not date is today and
not date is tomorrow

This mirrors the structure of the query rule whose expression is (expression @" "? expr_op? @" "?)+ -- a repetition of one or more instances of the
expression expression @" "? expr_op? @" "?. So each element of a match for
query will contain both an expression and a expr_op.

In order to get this output:

not distance is 20
or
not date is today
and
not date is tomorrow

You would need a grammar where the query rule was a repetion of either
expression or expr_op, i.e. (expression / expr_op)+ (with spaces added
where necessary). This is probably not what you want, because it means the
grammar would match nonsense queries like or or and or.

If I change the grammar such that all nodes are required I get the expected
output.

I'm not sure what you mean by "required" here, and the grammar presented after
this paragraph is identical to the first one.

Then you give another version of the query rule:

query <- expression (@" " expr_op @" " expression)*

This rule is a sequence of two elements: it requires one expression followed
by zero or more instances of (" " expr_op " " expression). Here is the output
showing which part of the rule is responsible for each part:

not distance is 20                                  expression
or not date is today and not date is tomorrow       (@" " expr_op @" " expression)*

If you look inside the second element here you'll see multiple elements for each
match of the (...)* expression:

or not date is today
and not date is tomorrow

This rule looks to me like the correct grammar, as you seem to want your query
language to allow one or more expressions joined together by and/or
operators. Every other version of query you've shown will also match other
strings that won't make any sense, e.g. allowing the string to start or end with
an operator, or be composed entirely of operators.

from canopy.