Mongo

MongoPool class

client_default: MongoClient = MongoClient(‘mongodb://localhost:27017/’) db_default = client_default[‘eos’] db_config_default = db_config_servers_by_name[“default”]

---

source

MongoPool

 MongoPool (_cnt:int=0, meta:tspace.data.core.ObservationMeta,
            db_config:tspace.config.db.DBConfig, doc_query:dict=<factory>,
            codec_option:bson.codec_options.CodecOptions,
            query:Optional[tspace.data.core.PoolQuery]=None,
            coll_name:Optional[str]=None,
            collection:Optional[pymongo.collection.Collection]=None,
            client:Optional[pymongo.mongo_client.MongoClient]=None,
            logger:Optional[logging.Logger]=None,
            dict_logger:Optional[dict]=None)

*A MongoDB store for storing data in a time series collection.

features:

timeseries interfacing pandas.DataFrame
multikey index for fast query
typed hints

The key leads to a config with db_name and collection name with a switch for record or episode.

Attributes:

- meta: ObservationMeta, meta information of the observation
- db_config: DBConfig, database configuration
- doc_query: dict, query dict for mongodb
- codec_option: CodecOptions, codec options for mongodb
- query: PoolQuery, query for mongodb
- coll_name: str, collection name
- collection: Collection, collection for mongodb
- client: MongoClient, client for mongodb
- logger: logging.Logger, logger for mongodb
- dict_logger: dict, dict for logging*

---

source

MongoPool.__post_init__

 MongoPool.__post_init__ ()

User weakref finalizer to make sure close is called when the object is destroyed

---

source

MongoPool.load

 MongoPool.load ()

*Initialize the pool interface

This function should: - connect to db - init*

---

source

MongoPool.find_item

 MongoPool.find_item (doc_id:int)

Find a record by id.

---

source

MongoPool.close

 MongoPool.close ()

close the pool, for destructor

---

source

MongoPool.parse_query

 MongoPool.parse_query (query:tspace.data.core.PoolQuery)

One-Trick Pony: check query is valid or not return dict as mongo document query filter if query is None, return empty dict

---

source

MongoPool.store_record

 MongoPool.store_record (episode:pandas.core.frame.DataFrame)

Deposit the records of an episode into the db.

---

source

MongoPool.store_episode

 MongoPool.store_episode (episode:pandas.core.frame.DataFrame)

Deposit a DataFrame of an episode into the db.

---

source

MongoPool.store

 MongoPool.store (episode:pandas.core.frame.DataFrame)

*Deposit the records of an episode into the db.

Based on the type of the db collection, store the episode as a set of records or as a complete episode.*

---

source

MongoPool.find

 MongoPool.find (query:tspace.data.core.PoolQuery)

Find records by a query object.

---

source

MongoPool.delete

 MongoPool.delete (item_id)

Delete a record by item id.

---

source

MongoPool._count

 MongoPool._count (query:Optional[tspace.data.core.PoolQuery]=None)

*Count the number of records in the db.

for episode/record document*

---

source

MongoPool.sample

 MongoPool.sample (size:int=4,
                   query:Optional[tspace.data.core.PoolQuery]=None)

*Sample a batch of records from the db.

db.hyperparameters_collection.aggregate([ {“\(match": {"start_time": {"\)exists”: False}}}, {“$sample”: {“size”: 1}}])

if PoolQuery doesn’t contain ‘timestamp_start’ and ‘timestamp_end’, the full episode is retrieved.*

	Type	Default	Details
size	int	4	batch size, default 4
query	Optional[PoolQuery]	None
Returns	Optional[pd.DataFrame]		samples as one multi-indexed Pandas DataFrame