I would like someone to confirm whether the following behaviour is expected. Given:

{
  "a": [ 
    {"b": 1, "c": [1, 2, 3], "d": "foo"},
    {"b": 2, "d": "bar"},
    {"c": [0]}
  ]
}

• a.exists(x, x.b == 1) returns true
• a.exists(x, x.b == 3 ) returns this error: failed to evaluate: no such key: b

Is this asymmetry expected? It seems that exists is positively safe but negatively unsafe, which feels weird. I would expect the evaluation to fail in both cases. For what it's worth, that is indeed what happens with filter. You can check the example online here.

At first I thought that in the affirmative case the computation was being short-circuited after evaluating the first (matching) element but that is not the case. You can confirm by rearranging the array; the order does not matter.

Originally posted here:

• Exists: positively safe but negatively unsafe (???) projectnessie/cel-java#631