TL;DR you have to read the entire blob binary and create the appropriate data structure before filtering whereas you can skip parts with JSON, and thus make it faster.

JSON, however, allows stream parsing. Consider for example, the Jackson library’s [nextToken()](https://fasterxml.github.io/jackson-core/javadoc/2.8/com/fasterxml/jackson/core/JsonParser.html#nextToken()) or Go’s json.Tokenizer.

Sadly the code produced for Amazon Web Services is not open-source yet.