Which parts of this library are threadsafe?

[laian]

Is this library in general threadsafe?

Assuming that I preload all schemas for the RefResolver:

Can I share validators across threads?

If not, can I have 1 validator per thread, but have all validators across threads share the same RefResolver?

[Julian]

Hey -- in general no objects are explicitly guaranteed to be threadsafe (and no tests currently deal with threads).

If all you're trying to do is share some large dicts, adding them to a bunch of different RefResolver stores doesn't copy them.

[Julian]

#223 might also interest you.

[laian]

thanks!

[stkrp]

Hello.

Sorry to bring up the old closed topic, but is it safe to pass the same object (dict) to different validators (one validator per thread)?

Example

I doubt this code works, but I hope that's enough for an example.

import threading
from multiprocessing.pool import ThreadPool

from jsonschema import Draft7Validator


heavy_schema = {
    # Includes several `$ref` like "#/definitions/some_struct"
    # ...
}


class ThreadSafeValidator:
    def __init__(self, schema):
        self._schema = schema
        self._tl = threading.local()
    
    def validate(self, value):
        if not hasattr(self._tl, "validator"):
            # Create independent `Draft7Validator` for each thread.
            # Can I use `_schema` without deepcopy?
            self._tl.validator = Draft7Validator(_schema)
        
        self._tl.validator.validate(value)


validator = ThreadSafeValidator(heavy_schema)


def validate_value(value):
    validator.validate(value)


pool = ThreadPool(4)
pool.map(validate_value, range(1000))
pool.close()
pool.join()

[Julian]

Validators don't modify the schemas they're given so that looks like it should be fine yeah

…

On Wed, Oct 23, 2019, 12:48 Stanislav Karpov ***@***.***> wrote: Hello. Sorry to bring up the old closed topic, but is it safe to pass the same object (dict) to different validators (one validator per thread)? Example I doubt this code works, but I hope that's enough for an example. import threadingfrom multiprocessing.pool import ThreadPool from jsonschema import Draft7Validator heavy_schema = { # Includes several `$ref` like "#/definitions/some_struct" # ... } class ThreadSafeValidator: def __init__(self, schema): self._schema = schema self._tl = threading.local() def validate(self, value): if not hasattr(self._tl, "validator"): # Create independent `Draft7Validator` for each thread. # Can I use `_schema` without deepcopy? self._tl.validator = Draft7Validator(_schema) self._tl.validator.validate(value) validator = ThreadSafeValidator(heavy_schema) def validate_value(value): validator.validate(value) pool = ThreadPool(4) pool.map(validate_value, range(1000)) pool.close() pool.join() — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#464>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACQQXXISRD3RREL22ALQBLQQB57HANCNFSM4FXVNBSA> .

[stkrp]

Great! Thanks.