JSON parser: error data path loses ancestors and keys when list keys follow other members of a list entry

[pepa-cz]

When JSON data contain a list entry whose key members come after other members (valid JSON – object members
are unordered and RFC 7951 does not require keys to be first), errors found in the members parsed before the keys
are reported with a truncated data path. The path misses not only the predicate of the list entry, but also all
its ancestors, so it is not a valid absolute data path at all (e.g. /test:l/e, although l is not a top-level
node).

The same data with the keys placed first produce the correct path (/test:top/l[name='k']/e).

Reproducer

test.yang:

module test {
  yang-version 1.1;
  namespace "urn:test";
  prefix t;

  container top {
    list l {
      key "name";
      leaf name {
        type string;
      }
      leaf e {
        type enumeration {
          enum a;
        }
      }
      container c {
        leaf-list ll {
          type string;
        }
      }
    }
  }
}

Data files and results of yanglint -f json -t config test.yang <file>:

file	data	reported path
enum-key-first.json	{"test:top": {"l": [{"name": "k", "e": "bogus"}]}}	/test:top/l[name='k']/e ✔
enum-key-last.json	{"test:top": {"l": [{"e": "bogus", "name": "k"}]}}	/test:l/e ✘
dup-key-first.json	{"test:top": {"l": [{"name": "k", "c": {"ll": ["x", "x"]}}]}}	/test:top/l[name='k']/c/ll[.='x'] ✔
dup-key-last.json	{"test:top": {"l": [{"c": {"ll": ["x", "x"]}, "name": "k"}]}}	/test:l/c/ll[.='x'] ✘

Full output:

$ yanglint -f json -t config test.yang enum-key-first.json
libyang err : Invalid enumeration value "bogus". (/test:top/l[name='k']/e) (line 1)
YANGLINT[E]: Failed to parse input data file "enum-key-first.json".
$ yanglint -f json -t config test.yang enum-key-last.json
libyang err : Invalid enumeration value "bogus". (/test:l/e) (line 1)
YANGLINT[E]: Failed to parse input data file "enum-key-last.json".
$ yanglint -f json -t config test.yang dup-key-first.json
libyang err : Duplicate instance of "ll". (/test:top/l[name='k']/c/ll[.='x']) (line 1)
libyang err : Duplicate instance of "ll". (/test:top/l[name='k']/c/ll[.='x']) (line 1)
YANGLINT[E]: Failed to parse input data file "dup-key-first.json".
$ yanglint -f json -t config test.yang dup-key-last.json
libyang err : Duplicate instance of "ll". (/test:l/c/ll[.='x']) (line 1)
libyang err : Duplicate instance of "ll". (/test:l/c/ll[.='x']) (line 1)
YANGLINT[E]: Failed to parse input data file "dup-key-last.json".

The same data without the error are accepted, i.e. a key placed after other members is valid input
(ok-key-last.json: {"test:top": {"l": [{"c": {"ll": ["x"]}, "name": "k"}]}}):

$ yanglint -f json -t config test.yang ok-key-last.json
{
  "test:top": {
    "l": [
      {
        "name": "k",
        "c": {
          "ll": [
            "x"
          ]
        }
      }
    ]
  }
}

Note that the duplicate leaf-list is reported correctly when it is a direct child of the list entry (it is validated
only after the whole entry, including its keys, is parsed). It is truncated when it is nested deeper (c/ll),
because the nested subtree is validated before the keys of the enclosing list entry are parsed.

Expected behavior

The reported path should be a valid absolute data path. Ideally the full path with the list keys
(/test:top/l[name='k']/e), because the keys are present in the data. If the keys cannot be known at the time the
error is detected, at least the ancestors should be kept (/test:top/l/e), so that the error can be located.