Timestore

Library for storing time series data in a PostgreSQL database.

Requirements:

• Docker (for developing)
• PostgreSQL
• TimescaleDB (if not using Docker)

Optional (if not using Docker):

• PGWeb
• Grafana

Installation

Clone the git repo, cd into the root folder and run pip install -e ..

Usage example

To set up a development environment using docker, run:

# Start a postgres database with timescaledb extension
docker run --rm -d --name timescaledb -e POSTGRES_PASSWORD=ts -p 5432:5432 timescale/timescaledb

# Start a PGWeb instance
docker run -d -p 8081:8081 --link timescaledb:db -e DATABASE_URL="postgres://postgres:ts@db:5432/postgres?sslmode=disable" sosedoff/pgweb

# Start a Grafana instance
docker run -d -p 3000:3000 --link timescaledb:db grafana/grafana

To create a new time series table, just create a dataclass-style using the timeseries decorator and run create.

import numpy as np
from typing import Optional

from timestore import timeseries, SQLTypes, _TABLE_NAME
from timestore.sql import insert, create


@timeseries
class SensorData():
    value: SQLTypes.DOUBLE_PRECISION
    location: Optional[SQLTypes.TEXT]
    metadata: SQLTypes.JSONB

create(SensorData)

To insert a datapoint, create an instance of the class and run insert:

data_point = SensorData(value=np.random.rand(), location="Berkeley", metadata={"user": "guen"})
insert(data_point)

To batch insert a list of data points, create a list of instances and use batch_insert:

data_points = [SensorData(value=val, location="Berkeley", metadata={"user": "guen"}) for val in np.random.rand(1000)]
batch_insert(data_points)

To query, run a SQL query, e.g.

SELECT time AS time, value FROM sensor_data WHERE time > current_timestamp - interval '5' minute;

Or use the query function to get a list of dataclass instances back:

data = [SensorData(*result) for result in query(SensorData, limit=10)]

For an example, see the scripts folder.