Model patterns: strange behavior with sym and grouped forms

[KevinMGranger]

Perhaps I'm misunderstanding how these model patterns work, or maybe funcparserlib. Does this seem right?

(import hy.model-patterns [pexpr sym FORM whole])

(defn parse-foo-2 [parser] (.parse parser ['(foo 2)]))

(print "Wrong:")
(setv bad-parser (pexpr (sym "foo") FORM))
(try
  (parse-foo-2 bad-parser)
  (except [e TypeError] (print e)))
; Output: 'Integer' object is not iterable

(print "Right:")
(setv good-parser (pexpr (+ (sym "foo") FORM)))
(print (hy.repr (parse-foo-2 good-parser)))
; Output: '(2)

It seems strange to require that additional level of nesting. Easy enough to fix (albeit unergonomically) but still strange.

For this specific case, maybe pvalue from hy.core.result_macros should be exposed in hy.model_patterns to help with that?

[Kodiologist]

This looks like a bug related to whole failing to listify its parsed value if one of the arguments is skipped. I'm working on it.

See #2691.

[KevinMGranger]

I just noticed this bit in the funcparserlib documentation about skip / __neg__:

You should not pass the resulting parser to any combinators other than +. You should have at least one non-skipped value in your p1 + p2 + ... + pN. The parsed value of -p is an internal _Ignored object, not intended for actual use.

It looks like this is actually the expected behavior.