s3-as-a-datastore 2.7.0

S3-as-a-datastore is a library that lives on top of botocore and boto3, as a way to use S3 as a key-value datastore instead of a real datastore

S3-as-a-datastore is a library that lives on top of botocore and boto3, as a way to use S3 as a key-value datastore instead of a real datastore

DISCLAIMER: This is NOT a real datastore, only the illusion of one. If you have remotely high I/O, this is NOT the library for you.

Motivation

S3 is really inexpensive compared to Memcache, or RDS.

For example, this is the RDS cost

while this is S3 cost

If a service doesn’t have a lot of traffic, keeping up a RDS deployment is wasteful because it stands idle but incurring cost. S3 doesn’t have that problem. For services that has low read/writes operations, or only has CRD without the U (if you don’t know what that means, read CRUD), saving things in S3 gets similar results. As long as data isn’t getting upgrade, only written and read, S3 can be used. However, Writing to S3 requires a lot of documentation reading if you’re not used to it. This library is an interface to communication with S3 like a very pseudo-ORM way.

Installation

pip3 install s3aads

Idea

The main idea is a database is mapped to a bucket, and a table is the top level “folder” of s3. The rest of nested “folders” are columns. Because the way buckets work in S3, they must be unique for all S3 buckets. This also mean the combination of keys must be unique

NOTE: There are quotations around “folder” because files in a S3 bucket are flat, and there aren’t really folders.

Example

Database: joeyism-test
Table: daily-data

id | year | month | day | data
------------------------------
 1 | 2020 |    01 |  01 | ["a", "b"]
 2 | 2020 |    01 |  01 | ["c", "d"]
 3 | 2020 |    01 |  01 | ["abk20dj3i"]

is mapped to

joeyism-test/daily-data/1/2020/01/01  ->  ["a", "b"]
joeyism-test/daily-data/2/2020/01/01  ->  ["c", "d"]
joeyism-test/daily-data/3/2020/01/01  ->  ["abk20dj3i"]

but it can be called with

from s3aads import Table
table = Table(name="daily-data", database="joeyism-test", columns=["id", "year", "month", "day"])
table.select(id=1, year=2020, month="01", day="01") # b'["a", "b"]'
table.select(id=2, year=2020, month="01", day="01") # b'["c", "d"]'
table.select(id=3, year=2020, month="01", day="01") # b'["abk20dj3i"]'

Usage

Example

from s3aads import Database, Table
db = Database("joeyism-test")
db.create()

table = Table(name="daily-data", database=db, columns=["id", "year", "month", "day"])
table.insert(id=1, year=2020, month="01", day="01", data=b'["a", "b"]')
table.insert(id=2, year=2020, month="01", day="01", data=b'["c", "d"]')
table.insert(id=2, year=2020, month="01", day="01", data=b'["abk20dj3i"]')

table.select(id=1, year=2020, month="01", day="01") # b'["a", "b"]'
table.select(id=2, year=2020, month="01", day="01") # b'["c", "d"]'
table.select(id=3, year=2020, month="01", day="01") # b'["abk20dj3i"]'

table.delete(id=1, year=2020, month="01", day="01")
table.delete(id=2, year=2020, month="01", day="01")
table.delete(id=3, year=2020, month="01", day="01")

API

Database

Database(name)

• name: name of the table

Properties

tables: list of tables for that Database (S3 Bucket)

Methods

create(): Create the database (S3 Bucket) if it doesn’t exist

get_table(table_name) -> Table: Pass in a table name and returns the Table object

drop_table(table_name): Fully drops table

Class methods

list_databases(): List all available databases (S3 Buckets)

Table

Table(name, database, columns=[])

• name: name of the table

• database: Database object. If a string is passed instead, it’ll attempt to fetch the Database object

• columns (default: []): Table columns

Properties

keys: list of all keys in that table. Essentially, list the name of all files in the folder

Full Param Methods

The following methods require all the params to be passed in order for it to work.

delete(**kwargs): If you pass the params, it’ll delete that row of data

insert(data:bytes, **kwargs): If you pass the params and value for data, it’ll insert that row of bytes data

insert_string(data:string, **kwargs): If you pass the params and value for data, it’ll insert that row of string data

select(**kwargs) -> bytes: If you pass the params, it’ll select that row of data and return the value as bytes

select_string(**kwargs) -> string: If you pass the params, it’ll select that row of data and return the value as a string

Partial Param Methods

The following methods can work with partial params passed in.

query(**kwargs) -> List[Dict[str, str]]: If you pass the params, it’ll return a list of params that is availabe in the table

Key Methods

delete_by_key(key): If you pass the full key/path of the file, it’ll delete that row/file

insert_by_key(key, data: bytes): If you pass the full key/path of the file and the data (in bytes), it’ll insert that row/file with the data

select_by_key(key) -> bytes: If you pass the full key/path of the file, it’ll select that row/file and return the data

query_by_key(key="", sort_by=None) -> List[str]: If you pass the full or partial key/path of the file, it’ll return a list of keys that matches the pattern

• sort_by: Possible values are Key, LastModified, ETag, Size, StorageClass

Methods

distinct(columns: List[str]) -> List[Tuple]: If you pass a list of columns, it’ll return a list of distinct tuple combinations based on those columns

random_key() -> str: Returns a random key to data

random() -> Dict: Returns a set of params and data of a random data

count() -> int: Returns the number of objects in the table

<first_column_name>s() -> List: Taking the name of the first column, returns a list of unique values.

<n_column_name>s() -> List: Taking the name of the Nth column, returns a list of unique values.

• For example, a table with columns ["id", "name"] will have the method table.ids() which will return a list of unique ids