Hi Mia, On Tue, Feb 03, 2026 at 08:00:56PM +0200, Mia Kanashi wrote: > Hiiii, replying to the archive, hopefully it works > > > It could be helpful to understand why this lib needs to allocate RAM > > when others do not. > > It is due to the fact that the API is DOM based, while mjson is SAX like, > in other words yyjson allocates structure that represents document tree, > while mjson is just a glorified tokenizer. > > It is in their plans to support StAX API, but nothing ready yet.
OK! > See: > > https://github.com/ibireme/yyjson/issues/33 > > https://github.com/ibireme/yyjson/issues/219 Thanks for the links. > Note that it already supports incremental (chunked) reader with DOM: > > https://github.com/ibireme/yyjson/pull/199 > > > Allocation was my concern too, they allow custom allocators though. > StAX API would be worth looking for. > > > Overall I just wanted to tell that that library exists, > to integrate it if mjson parsing speed ever becomes a concern, > integrating it right now wouldn't be productive in my opinion. I agree, it's always interesting to know about alternate implementations. Even if not needed right now, it could be needed later and being aware of it would help. > One place I think it could be useful is to provide it as an API for Lua, > json parsing is important enough to have in the core, > and json documents can be sometimes quite big, Maybe but we would need to be extremely careful not to waste all CPU cycles parsing a doc at once, otherwise during this time, other connections are stalled and QUIC packets are lost. But still something to keep an eye on anyway. > but integrating it just for that isn't worth it IMO. thanks! Willy
