Mongojet

Mongojet

Async (asyncio) MongoDB client for Python. It uses Rust MongoDB driver and tokio under the hood. Mongojet is 2-4x faster than Motor (and 1.5-3.5x faster than PyMongo AsyncMongoClient) in high concurrency scenarios (see benchmarks below).

Requirements

• Python >= 3.9
• pymongo>=4.6.2 (only bson package is required)

Installation

pip install mongojet

Usage

Mongojet has an API similar to PyMongo/Motor (but not exactly the same)

Creating a Client

Typically, you should create a single instance of Client per application/process. All client options should be passed via MongoDB connection string.

from mongojet import create_client, ReadPreference

client = await create_client('mongodb://localhost:27017/test_database?maxPoolSize=16')

Getting a Database

default database

db = client.get_default_database()

database with specific name

db = client.get_database('test_database')

database with specific name and options

db = client.get_database('test_database', read_preference=ReadPreference(mode='secondaryPreferred'))

Getting a Collection

collection = db['test_collection']

with options

collection = db.get_collection('test_collection', read_preference=ReadPreference(mode='secondary'))

Inserting documents

insert_one

document = {'key': 'value'}
result = await collection.insert_one(document)
print(result)
#> {'inserted_id': ObjectId('...')}

insert_many

documents = [{'i': i} for i in range(1000)]
result = await collection.insert_many(documents)
print(len(result['inserted_ids']))
#> 1000

Find documents

find_one (to get a single document)

document = await collection.find_one({'i': 1})
print(document)
#> {'_id': ObjectId('...'), 'i': 1}

find (to get cursor which is an async iterator)

cursor = await collection.find({'i': {'$gt': 5}}, sort={'i': -1}, limit=10)

you can iterate over the cursor using the async for loop

async for document in cursor:
    print(document)

or collect cursor to list of documents using to_list method

documents = await cursor.to_list()

find_many (to get list of documents in single batch)

documents = await collection.find_many({'i': {'$gt': 5}}, sort={'i': -1}, limit=10)

Counting documents

n = await collection.count_documents({'i': {'$gte': 500}})
print(n)
#> 500

Aggregating documents

cursor = await collection.aggregate(pipeline=[
    {'$match': {'i': {'$gte': 10}}},
    {'$sort': {'i': 1}},
    {'$limit': 10},
])
documents = await cursor.to_list()
print(documents)

Updating documents

replace_one

result = await collection.replace_one(filter={'i': 5}, replacement={'i': 5000})
print(result)
#> {'matched_count': 1, 'modified_count': 1, 'upserted_id': None}

update_one

result = await collection.update_one(filter={'i': 5}, update={'$set': {'i': 5000}}, upsert=True)
print(result)
#> {'matched_count': 0, 'modified_count': 0, 'upserted_id': ObjectId('...')}

update_many

result = await collection.update_many(filter={'i': {'$gte': 100}}, update={'$set': {'i': 0}})
print(result)
#> {'matched_count': 900, 'modified_count': 900, 'upserted_id': None}

Deleting documents

delete_one

result = await collection.delete_one(filter={'i': 5})
print(result)
#> {'deleted_count': 1}

delete_many

result = await collection.delete_many(filter={'i': {'$gt': 5}})
print(result)
#> {'deleted_count': 94}

Working with GridFS

bucket = db.gridfs_bucket(bucket_name="images")

with open('/path/to/my/awesome/image.png', mode='rb') as file:
    data = file.read()
    result = await bucket.put(data, filename='image.png', content_type='image/png')
    file_id = result['file_id']

with open('/path/to/my/awesome/image_copy.png', mode='wb') as file:
    data = await bucket.get_by_id(file_id)
    file.write(data)

await bucket.delete(file_id)

Simple benchmark (lower is better):

find_one

<!--  -->

insert_one

iterate over cursor