TinyDB Storage Proxy performance benchmark

storage_proxy_bench.py

from tinydb import TinyDB
from tinydb.database import Table, StorageProxy, Document
from tinydb.storages import MemoryStorage
from uuid import uuid4


class UUIDTable(Table):
    def _init_last_id(self, data):
        pass

    def _get_next_id(self):
        return str(uuid4())


class UUIDStorageProxy(StorageProxy):
    def _new_document(self, key, val):
        doc_id = key
        return Document(val, doc_id)


# TinyDB.storage_proxy_class = UUIDStorageProxy
# TinyDB.table_class = UUIDTable


## DB
db = TinyDB(storage=MemoryStorage)
table = db.table("cubes")


"""
Single insert is extremely slow because of repeated reads/writes:

for x in range(100000):
    db.insert({"key": x, "value": x ** 2})
    table.insert({"key": x, "value": x ** 2})
"""

size = 1_000_000

squares = [{"key": x, "value": x ** 2} for x in range(size)]
cubes = [{"key": x, "value": x ** 3} for x in range(size)]

db.insert_multiple(squares)
table.insert_multiple(cubes)


"""
Benchmarking

Repeated reads from:
table
db
dict
"""


def benchmark():
    from timeit import timeit
    from random import randint

    repeats = 50

    def db_bench():
        x = randint(1, size)
        return db.get(doc_id=x)

    def table_bench():
        x = randint(1, size)
        return table.get(doc_id=x)

    def dict_bench():
        x = randint(1, size)
        return db._storage.memory["cubes"][x]

    print(f"Reads on db: {timeit(db_bench, number=repeats)}")
    print(f"Reads on table: {timeit(table_bench, number=repeats)}")
    print(f"Reads on dict: {timeit(dict_bench, number=repeats)}")

Turns out the db/table comparison isn't really worth doing since db is just defaulting to _default which in turn is another table object.