Reusing parser modifies previous results about cysimdjson HOT 4 CLOSED

Note that the parser holds the allocated memory. By reusing the parser from document to document, we reuse the allocated memory. The net result is to avoid memory reallocation, an expensive process. On some systems, it is more expensive to allocate memory than to parse JSON !

from cysimdjson.

Ok, after some digging, this is actually correct behaviour - it is linked with the SIMDJSON requirement for the lifecycle of the document, see the remark at https://simdjson.org/api/0.9.1/classsimdjson_1_1dom_1_1parser.html#ab3e5bbb1974a1932aead90ad63883a23

I will try to provide some kind of indication of this as per your suggestion.

from cysimdjson.

Thanks for reporting, we will look at that.

from cysimdjson.

You can see one approach to stopping this in pysimdjson, https://github.com/TkTech/pysimdjson/blob/master/simdjson/csimdjson.pyx#L437. It's likely not our final fix, since it's a bit buggy on pypy whose garbage collector might wait a long, long time to cleanup things pointing into the Parser's document even if it's been explicitly deleted. In @ROpdebee's example above, it would have raised a RuntimeError instead of a potential segfault.

from cysimdjson.