Feature Table Usage¶

In [0]:

import tecton
import pandas
from datetime import datetime, timedelta

ws = tecton.get_workspace("prod")

import tecton import pandas from datetime import datetime, timedelta ws = tecton.get_workspace("prod")

Load a Feature Table¶

In [0]:

ft = ws.get_feature_table("user_login_counts")
ft.summary()

ft = ws.get_feature_table("user_login_counts") ft.summary()

Out[13]:

Name	user_login_counts
Workspace	prod
Description	User login counts over time.
Created At	2022-06-07 17:57:09 UTC
Owner	derek@tecton.ai
Last Modified By	jake@tecton.ai
Family
Source Filename	fraud/features/feature_tables/user_login_counts.py
Tags	{'release': 'production'}
Type	FeatureTable
URL	https://staging.tecton.ai/app/repo/prod/features/user_login_counts
Entities	fraud_user
Features	user_login_count_7d, user_login_count_30d
Feature Services
Timestamp Key	timestamp
Online Materialization	Enabled
Offline Materialization	Enabled
Online Join Keys	user_id
Offline Join Keys	user_id
Serving TTL	7 days
Schema	user_id string timestamp timestamp user_login_count_7d long user_login_count_30d long

Ingest Feature Data¶

In [0]:

df = pandas.DataFrame([{"user_id": "user_1", 
                        "timestamp": pandas.Timestamp(datetime.now()),
                        "user_login_count_7d": 15,
                        "user_login_count_30d": 35},
                       {"user_id": "user_1", 
                        "timestamp": pandas.Timestamp(datetime.now() - timedelta(days=7)),
                        "user_login_count_7d": 20,
                        "user_login_count_30d": 40},
                       {"user_id": "user_2", 
                        "timestamp": pandas.Timestamp(datetime.now()),
                        "user_login_count_7d": 100,
                        "user_login_count_30d": 200}])
ft.ingest(df)

df = pandas.DataFrame([{"user_id": "user_1", "timestamp": pandas.Timestamp(datetime.now()), "user_login_count_7d": 15, "user_login_count_30d": 35}, {"user_id": "user_1", "timestamp": pandas.Timestamp(datetime.now() - timedelta(days=7)), "user_login_count_7d": 20, "user_login_count_30d": 40}, {"user_id": "user_2", "timestamp": pandas.Timestamp(datetime.now()), "user_login_count_7d": 100, "user_login_count_30d": 200}]) ft.ingest(df)

Check materialization status¶

Now that the ingestion has been kicked off, you can check the job status in your Notebook or the Web UI.

In [0]:

ft.materialization_status()

ft.materialization_status()

All the displayed times are in UTC time zone Out[17]:

TYPE	WINDOW_START_TIME	WINDOW_END_TIME	STATUS	ATTEMPT	JOB_CREATED_AT	JOB_LOGS
INGEST	N/A	N/A	SUCCESS	1	2022-06-07 18:05:59	https://tecton-production.cloud.databricks.com/?o=1249442739035107#job/59589832183771/run/67436640

Read Features Online¶

In [0]:

ft.get_online_features(join_keys={"user_id": "user_1"}).to_dict()

ft.get_online_features(join_keys={"user_id": "user_1"}).to_dict()

Out[18]: {'user_login_count_30d': 35, 'user_login_count_7d': 15}

Read Features Offline¶

In [0]:

ft.get_historical_features(entities=pandas.DataFrame([{"user_id": "user_1"}])).to_pandas()

ft.get_historical_features(entities=pandas.DataFrame([{"user_id": "user_1"}])).to_pandas()

Out[19]:

	user_id	user_login_count_7d	user_login_count_30d	timestamp
0	user_1	15	35	2022-06-07 17:57:51.410874
1	user_1	20	40	2022-05-31 17:57:51.410897