For this use case it makes sense. cJSON is a general-purpose library, but Redis only needs fast access to a few top-level fields for filters. A tiny purpose-built parser that avoids allocations and skips building a full tree will naturally beat a generic JSON parser here. It’s not “better than cJSON” in general, just better for this very narrow hot path.