Mongo

MongoBuffer class

MongoBuffer

 MongoBuffer (pool:Optional[tspace.storage.pool.mongo.MongoPool]=None,
              batch_size:int, driver:tspace.config.drivers.Driver,
              truck:tspace.config.vehicles.Truck,
              meta:tspace.data.core.ObservationMeta,
              db_config:tspace.config.db.DBConfig,
              torque_table_row_names:list[str],
              query:Optional[tspace.data.core.PoolQuery]=None, logger:Opti
              onal[<module'logging'from'/opt/hostedtoolcache/Python/3.11.9
              /x64/lib/python3.11/logging/__init__.py'>]=None,
              dict_logger:Optional[dict]=None)

*A Buffer connected with a MongoDB database pool

Args:

batch_size: number of documents for batch sampling
driver: driver of the vehicle
truck: subject vehicle
meta: metadata of the observation
db_config: config for mongodb (str) or DBConfig (dict)
the key leads to a config with db_name and collection name with a switch for record or episode:
    - string for db server name
    - or string of the format "usr:password@host:port"
        for mongo_cluster:
            Host="10.10.0.4",  # url for the database server
            port="23000",  # port for the database server
            user_name="admin",  # username for the database server
            password="ty02ydhVqDj3QFjT",  # password for the database server
            ==> mongo_key = "admin:ty02ydhVqDj3QFjT@10.10.0.4:23000"
torque_table_row_names: list of torque table row names, e.g. ['r0, r1, ..., r9']
query: query for the pool
pool: pool of the database, a [`MongoPool`](https://Binjian.github.io/tspace/05.storage.pool.mongo.html#mongopool) instance
logger: logger for the buffer
dict_logger: dict logger for the buffer*

source

MongoBuffer.__post_init__

 MongoBuffer.__post_init__ ()

set logger and load pool

source

MongoBuffer.load

 MongoBuffer.load ()

load pool from database

source

MongoBuffer.find

 MongoBuffer.find (idx)

find a record by index

source

MongoBuffer.close

 MongoBuffer.close ()

close the pool and the database

source

MongoBuffer.sample

 MongoBuffer.sample ()

*Sampling a batch of records or episodes from the pool.

Decoding the batch observations from mongodb nested dicts to pandas dataframe flatten is used to adapt to model interface

Return:

A quadruple of numpy arrays (states, actions, rewards, next_states)*

source

MongoBuffer.decode_batch_records

 MongoBuffer.decode_batch_records (df:pandas.core.frame.DataFrame)

*Sample a batch of records from the pool.

Args:

df: dataframe of sampled from the record collection of the pool

Return:

A quadruple of numpy arrays (states, actions, rewards, next_states)*