django-postgres-partition 0.1.4

Partition support for django, based on django-postgres-extra

Django Postgres Partition

Fork of https://github.com/SectorLabs/django-postgres-extra with partition support only.

Status: Unstable, only change is renamed module.

pip install django-postgres-partition

Features

• PostgreSQL 11.x declarative table partitioning

  • Supports range and list partitioning

Prerequisites

• PostgreSQL 13+
• Django 5.2+
• Python 3.11+
• psycopg 3.2+

Usage

psql_partition.models.PostgresPartitionedModel adds support for PostgreSQL Declarative Table Partitioning.

The following partitioning methods are available:

• PARTITION BY RANGE
• PARTITION BY LIST
• PARTITION BY HASH

Creating partitioned tables

Partitioned tables are declared like regular Django models with a special base class and two extra options to set the partitioning method and key. Once declared, they behave like regular Django models.

Declaring the model

Inherit your model from psql_partition.models.PostgresPartitionedModel and declare a child class named PartitioningMeta. On the meta class, specify the partitioning method and key.

• Use psql_partition.types.PostgresPartitioningMethod.RANGE to PARTITION BY RANGE
• Use psql_partition.types.PostgresPartitioningMethod.LIST to PARTITION BY LIST
• Use psql_partition.types.PostgresPartitioningMethod.HASH to PARTITION BY HASH

from django.db import models

from psql_partition.types import PostgresPartitioningMethod
from psql_partition.models import PostgresPartitionedModel

class MyModel(PostgresPartitionedModel):
    class PartitioningMeta:
        method = PostgresPartitioningMethod.RANGE
        key = ["timestamp"]

    name = models.TextField()
    timestamp = models.DateTimeField()

Generating a migration

Run the following command to automatically generate a migration:

python manage.py pgmakemigrations

This will generate a migration that creates the partitioned table with a default partition.

Warning:

Always use python manage.py pgmakemigrations for partitioned models.

The model must be created by the psql_partition.backend.migrations.operations.PostgresCreatePartitionedModel operation.

Do not use the standard python manage.py makemigrations command for partitioned models. Django will issue a standard django.db.migrations.operations.CreateModel operation. Doing this will not create a partitioned table and all subsequent operations will fail.

Automatically managing partitions

The python manage.py pgpartition command can help you automatically create new partitions ahead of time and delete old ones for time-based partitioning.

You can run this command manually as needed, schedule to run it periodically or run it every time you release a new version of your app.

Warning:

We DO NOT recommend that you set up this command to automatically delete partitions without manual review.

Specify --skip-delete to not delete partitions automatically. Run the command manually periodically without the --yes flag to review partitions to be deleted.

Command-line options

Long flag

Short flag

Default

Description

--yes

-y

False

Specifies yes to all questions. You will NOT be asked for confirmation before partition deletion.

--using

-u

'default'

Optional name of the database connection to use.

--skip-create

False

Whether to skip creating partitions.

--skip-delete

False

Whether to skip deleting partitions.

Configuration

In order to use the command, you have to declare an instance of psql_partition.partitioning.PostgresPartitioningManager and set PSQLEXTRA_PARTITIONING_MANAGER to a string with the import path to your instance of psql_partition.partitioning.PostgresPartitioningManager.

For example:

# myapp/partitioning.py
from psql_partition.partitioning import PostgresPartitioningManager

manager = PostgresPartitioningManager(...)

# myapp/settings.py
PSQLEXTRA_PARTITIONING_MANAGER = 'myapp.partitioning.manager'

Time-based partitioning

from dateutil.relativedelta import relativedelta

from psql_partition.partitioning import (
    PostgresPartitioningManager,
    PostgresCurrentTimePartitioningStrategy,
    PostgresTimePartitionSize,
    partition_by_current_time,
)
from psql_partition.partitioning.config import PostgresPartitioningConfig

manager = PostgresPartitioningManager([
    # 3 partitions ahead, each partition is one month
    # delete partitions older than 6 months
    # partitions will be named `[table_name]_[year]_[3-letter month name]`.
    PostgresPartitioningConfig(
        model=MyPartitionedModel,
        strategy=PostgresCurrentTimePartitioningStrategy(
            size=PostgresTimePartitionSize(months=1),
            count=3,
            max_age=relativedelta(months=6),
        ),
    ),
    # 6 partitions ahead, each partition is two weeks
    # delete partitions older than 8 months
    # partitions will be named `[table_name]_[year]_week_[week number]`.
    PostgresPartitioningConfig(
        model=MyPartitionedModel,
        strategy=PostgresCurrentTimePartitioningStrategy(
            size=PostgresTimePartitionSize(weeks=2),
            count=6,
            max_age=relativedelta(months=8),
        ),
    ),
    # 12 partitions ahead, each partition is 5 days
    # old partitions are never deleted, `max_age` is not set
    # partitions will be named `[table_name]_[year]_[month]_[month day number]`.
    PostgresPartitioningConfig(
        model=MyPartitionedModel,
        strategy=PostgresCurrentTimePartitioningStrategy(
            size=PostgresTimePartitionSize(days=5),
            count=12,
        ),
    ),
])

Changing a time partitioning strategy

When switching partitioning strategies, you might encounter the problem that partitions for part of a particular range already exist.

In order to combat this, you can use the psql_partition.partitioning.PostgresTimePartitioningStrategy and specify the start_datetime parameter. As a result, no partitions will be created before the given date/time.

Custom strategy

You can create a custom partitioning strategy by implementing the psql_partition.partitioning.PostgresPartitioningStrategy interface.

You can look at psql_partition.partitioning.PostgresCurrentTimePartitioningStrategy as an example.

Manually managing partitions

If you are using list or hash partitioning, you most likely have a fixed amount of partitions that can be created up front using migrations or using the schema editor.

Using migration operations

Adding a range partition

Use the psql_partition.backend.migrations.operations.PostgresAddRangePartition operation to add a new range partition. Only use this operation when your partitioned model uses psql_partition.types.PostgresPartitioningMethod.RANGE.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresAddRangePartition

class Migration(migrations.Migration):
    operations = [
        PostgresAddRangePartition(
           model_name="mypartitionedmodel",
           name="pt1",
           from_values="2019-01-01",
           to_values="2019-02-01",
        ),
    ]

Adding a list partition

Use the psql_partition.backend.migrations.operations.PostgresAddListPartition operation to add a new list partition. Only use this operation when your partitioned model uses psql_partition.types.PostgresPartitioningMethod.LIST.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresAddListPartition

class Migration(migrations.Migration):
    operations = [
        PostgresAddListPartition(
           model_name="mypartitionedmodel",
           name="pt1",
           values=["car", "boat"],
        ),
    ]

Adding a hash partition

Use the psql_partition.backend.migrations.operations.PostgresAddHashPartition operation to add a new list partition. Only use this operation when your partitioned model uses psql_partition.types.PostgresPartitioningMethod.HASH.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresAddHashPartition

class Migration(migrations.Migration):
    operations = [
        PostgresAddHashPartition(
           model_name="mypartitionedmodel",
           name="pt1",
           modulus=3,
           remainder=1,
        ),
    ]

Adding a default partition

Use the psql_partition.backend.migrations.operations.PostgresAddDefaultPartition operation to add a new list partition.

Note that you can only have one default partition per partitioned table/model. An error will be thrown if you try to create a second default partition.

If you used python manage.py pgmakemigrations to generate a migration for your newly created partitioned model, you do not need this operation. This operation is added automatically when you create a new partitioned model.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresAddDefaultPartition

class Migration(migrations.Migration):
    operations = [
        PostgresAddDefaultPartition(
           model_name="mypartitionedmodel",
           name="default",
        ),
    ]

Deleting a default partition

Use the psql_partition.backend.migrations.operations.PostgresDeleteDefaultPartition operation to delete an existing default partition.

Warning:

Deleting the default partition and leaving your model without a default partition can be dangerous. Rows that do not fit in any other partition will fail to be inserted.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresDeleteDefaultPartition

class Migration(migrations.Migration):
    operations = [
        PostgresDeleteDefaultPartition(
           model_name="mypartitionedmodel",
           name="pt1",
        ),
    ]

Deleting a range partition

Use the psql_partition.backend.migrations.operations.PostgresDeleteRangePartition operation to delete an existing range partition. Only use this operation when your partitioned model uses psql_partition.types.PostgresPartitioningMethod.RANGE.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresDeleteRangePartition

class Migration(migrations.Migration):
    operations = [
        PostgresDeleteRangePartition(
           model_name="mypartitionedmodel",
           name="pt1",
        ),
    ]

Deleting a list partition

Use the psql_partition.backend.migrations.operations.PostgresDeleteListPartition operation to delete an existing range partition. Only use this operation when your partitioned model uses psql_partition.types.PostgresPartitioningMethod.LIST.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresDeleteListPartition

class Migration(migrations.Migration):
    operations = [
        PostgresDeleteListPartition(
           model_name="mypartitionedmodel",
           name="pt1",
        ),
    ]

Deleting a hash partition

Use the psql_partition.backend.migrations.operations.PostgresDeleteHashPartition operation to delete an existing range partition. Only use this operation when your partitioned model uses psql_partition.types.PostgresPartitioningMethod.HASH.

from django.db import migrations, models

from psql_partition.backend.migrations.operations import PostgresDeleteHashPartition

class Migration(migrations.Migration):
    operations = [
        PostgresDeleteHashPartition(
           model_name="mypartitionedmodel",
           name="pt1",
        ),
    ]

Using the schema editor

Use the psql_partition.backend.PostgresSchemaEditor to manage partitions directly in a more imperative fashion. The schema editor is used by the migration operations described above.

Adding a range partition

from django.db import connection

connection.schema_editor().add_range_partition(
    model=MyPartitionedModel,
    name="pt1",
    from_values="2019-01-01",
    to_values="2019-02-01",
)

Adding a list partition

from django.db import connection

connection.schema_editor().add_list_partition(
    model=MyPartitionedModel,
    name="pt1",
    values=["car", "boat"],
)

Adding a hash partition

from django.db import connection

connection.schema_editor().add_hash_partition(
    model=MyPartitionedModel,
    name="pt1",
    modulus=3,
    remainder=1,
)

Adding a default partition

from django.db import connection

connection.schema_editor().add_default_partition(
    model=MyPartitionedModel,
    name="default",
)

Deleting a partition

from django.db import connection

connection.schema_editor().delete_partition(
    model=MyPartitionedModel,
    name="default",
)

Original readme

:white_check_mark:	Tests
:memo:	License
:package:	PyPi
:four_leaf_clover:	Code coverage
	Django Versions	2.0, 2.1, 2.2, 3.0, 3.1, 3.2, 4.0, 4.1, 4.2, 5.0, 5.1, 5.2
	Python Versions	3.6, 3.7, 3.8, 3.9, 3.10, 3.11, 3.12, 3.13
	Psycopg Versions	2, 3
:book:	Documentation	Read The Docs
:warning:	Upgrade	Upgrade from v1.x
:checkered_flag:	Installation	Installation Guide
:fire:	Features	Features & Documentation
:droplet:	Future enhancements	Potential features

django-postgres-extra aims to make all of PostgreSQL's awesome features available through the Django ORM. We do this by taking care of all the hassle. As opposed to the many small packages that are available to try to bring a single feature to Django with minimal effort. django-postgres-extra goes the extra mile, with well tested implementations, seamless migrations and much more.

With seamless we mean that any features we add will work truly seamlessly. You should not have to manually modify your migrations to work with fields and objects provided by this package.

---

:warning: This README is for v2. See the v1 branch for v1.x.

---

Major features

See the full list

• Conflict handling (atomic upsert)

  Adds support for PostgreSQL's ON CONFLICT syntax for inserts. Supports DO UPDATE and DO NOTHING. Single statement, atomic and concurrency safe upserts. Supports conditional updates as well.

• Table partitioning

  Adds support for PostgreSQL 11.x declarative table partitioning. Integrated into Django migrations. Supports all types of partitioning. Includes a command to automatically create time-based partitions.

• Views & materialized views

  Adds support for creating views & materialized views as any other model. Integrated into Django migrations.

• Locking models & tables

  Support for explicit table-level locks.

• Creating/dropping schemas

  Support for managing PostgreSQL schemas.

• Truncating tables

  Support for TRUNCATE TABLE statements (including cascading).

For Django 3.1 and older:

• Conditional unique index
• Case insensitive index

For Django 2.2 and older:

• Unique index
• HStore unique and required constraints on specific HStore keys

Working with the code

Prerequisites

• PostgreSQL 14 or newer.
• Django 5.x or newer.
• Python 3.11 or newer.

These are just for local development. CI for code analysis etc runs against these. Tests will pass on all Python, Django and PostgreSQL versions documented. Linting, formatting and type-checking the code might not work on other Python and/or Django versions.

Getting started

1. Clone the repository:

    λ git clone https://github.com/SectorLabs/django-postgres-extra.git
   

2. Create a virtual environment:

   λ cd django-postgres-extra
   λ virtualenv env
   λ source env/bin/activate
   

3. Create a postgres user for use in tests (skip if your default user is a postgres superuser):

   λ createuser --superuser psqlextra --pwprompt
   λ export DATABASE_URL=postgres://psqlextra:<password>@localhost/psqlextra
   

   Hint: if you're using virtualenvwrapper, you might find it beneficial to put the export line in $VIRTUAL_ENV/bin/postactivate so that it's always available when using this virtualenv.

4. Install the development/test dependencies:

   λ pip install -r requirements-test.txt
   

5. Run the tests:

   λ poe test
   

6. Run the benchmarks:

   λ poe benchmark
   

7. Auto-format code, sort imports and auto-fix linting errors:

   λ poe fix