elastic/elasticsearch

XContentParser Behaves Differently for Stream and Byte Array

Geschlossen

#61.489 geöffnet am 24.08.2020

 (7 Kommentare) (0 Reaktionen) (0 zugewiesene Personen)Java (25.882 Forks)batch import
:Core/Infra/Core>bugTeam:Core/Infrahelp wantedtriaged

Repository-Metriken

Stars
 (76.700 Sterne)
PR-Merge-Metriken
 (Durchschn. Merge 2T) (1.000 gemergte PRs in 30 T)

Beschreibung

Optimizing BytesReference parsing in #61447 revealed a subtle bug in x-content parsing.

The following test:

    public void testIncompatibleStreamTypes() throws Exception {
        final Map<String, Object> map = Map.of("foo", "bar");
        final BytesReference jsonBytes = BytesReference.bytes(JsonXContent.contentBuilder().map(map));
        final Map<String, Object> deserialized = SmileXContent.smileXContent.createParser(
                NamedXContentRegistry.EMPTY, DeprecationHandler.IGNORE_DEPRECATIONS, jsonBytes.streamInput()).map();
        assertEquals(deserialized, Collections.emptyMap());

        final BytesRef bytes = jsonBytes.toBytesRef();
        final Map<String, Object> deserialized2 = SmileXContent.smileXContent.createParser(
                NamedXContentRegistry.EMPTY, DeprecationHandler.IGNORE_DEPRECATIONS, bytes.bytes, bytes.offset, bytes.length).map(); // this throws
    }

will actually pass until the last line which will throw:

com.fasterxml.jackson.core.JsonParseException: Input does not start with Smile format header (first byte = 0x7b) -- rather, it starts with '{' (plain JSON input?) -- can not parse
 at [Source: (byte[])"{"foo":"bar"}"; line: -1, column: 0]

So for reading from streams, we simply read empty map when the type is off but when reading from a byte array we (in my opinion correctly) throw an exception. I think we should throw that same exception when reading from a stream of the wrong content type shouldn't we?

Contributor Guide