Should filters pushed to the bottom of the query work?

[NiklasRosenstein]

Hi! I'm not entirely sure if this is supposed to work (though I think it would be intuitive and ergonomic if it would):

data = [
    {"payload": {"type": "A"}},
    {"payload": {"type": "B"}},
    {"payload": {"type": "C"}},
]

print(jsonpath.findall('$[*].payload[?(@.type == "B")]', data))

This gives you an empty list, when instead I would expect it to be equivalent to the result of this (working) version:

print(jsonpath.findall('$[?(@payload.type == "B")].payload', data))

[jg-rp]

Hi @NiklasRosenstein,

I believe this behaviour to be correct, according to RFC 9535, albeit not immediately intuitive.

If we apply $[*].payload to your example data without the filter, we get a node list containing these values:

[
  {
    "type": "A"
  },
  {
    "type": "B"
  },
  {
    "type": "C"
  }
]

If we then apply the filter, [?(@.type == "B")], to each of those values in turn, it selects nothing, because filters are applied to object/dict values without referencing their keys/properties.

Changing the filter to [?(@ == "B")] will give us a single node with the value "B".

import jsonpath

data = [
    {"payload": {"type": "A"}},
    {"payload": {"type": "B"}},
    {"payload": {"type": "C"}},
]

print(jsonpath.findall('$[*].payload[?(@ == "B")]', data))  # ['B']