simdjson
simdjson copied to clipboard
Parsing gigabytes of JSON per second : used by Facebook/Meta Velox, the Node.js runtime, ClickHouse, WatermelonDB, Apache Doris, Milvus, StarRocks
This is related to https://github.com/simdjson/simdjson/issues/94 The idea is that we do very few memory allocations as it is, but we could, optionally, allow the user to provide us with a...
Sorry about opening 2 tickets in a day, but I'm having some trouble started. With release 0.3.1, for `#include "simdjson.h"`, I get: ``` error LNK2019: unresolved external symbol "private: class...
Current we fuzz simdjson for crashes and internal checks, but not for correctness. We have a framework for testing for correctness by generating random JSON and testing against different JSON...
The API differs depending on whether exceptions are enabled or disabled. We currently publish and test mostly just the version with exceptions.
This makes it possible to do stuff like `std::distance(element.begin(), element.end())`
E.g., https://www.sphinx-doc.org/
Allow precise error reporting, so that we can narrow it down to the offending character when needed. cc @TkTech
We divide documents into small (e.g., 16kB) chunks. The document is parsed (in units of say 16kB) and possibly copied so it may be padded as needed. We run stage...
The padded string approach simplifies the code logic somewhat. However, it is not genuinely required. In time, it should be removed.
Now when oss-fuzz provides pretty code coverage, it is easy to see which parts of the code are not fuzzed properly. Identify what can be done, and add fuzzers targeting...