Skip to content

Latest commit

 

History

History
260 lines (174 loc) · 11.8 KB

README.md

File metadata and controls

260 lines (174 loc) · 11.8 KB
Raw

django-spicy-id

A drop-in replacement for Django's AutoField that gives you "Stripe-style" self-identifying string object IDs, like user_1234.

Status: Stable. No warranty, see LICENSE.txt.

PyPI version PyPI Supported Python Versions Test status

Table of Contents

What is a "spicy" id?

It's a made-up name (because I couldn't think of a better one) for a numeric primary keys type that is shown and manipulated as a string, prefixed with a constant value.

Here are some examples. You can use this library to make your Django row IDs look like:

  • user_1234
  • account-00000000deadbeef
  • bloop:1a2k3841x

Although you should always treat these values as opaque and never decode or parse the string's contents elsewhere (see Errors), you can think of every spicy id as being composed of:

<prefix> <separator> <encoded_value>
  • prefix: A fixed string value that will be the same for all IDs of this record type, forever.
  • separator: A configurable separator which, like prefix, is fixed forever; usually _ (the default) or - (another popular choice).
  • encoded_value: The numeric portion of the id. This library supports using base 16 (hex) or base 62.

Importantly, the underlying database value is still stored and retrieved as a numeric type, just like an AutoField, SmallAutoField, or BigAutoField.

Why use spicy ids?

Briefly: Because they're so much nicer for humans to work with.

  • Readability: While row-level primary keys are typically treated as "anonymous" (as in "not something anyone should have to care about"), the fact is these values still show up in lots of situations: They're in URLs, dumped in logfiles, shown in queries, and so on. In these situations, it's just plain faster to understand "what am I looking at" when the identifier itself tells you its type.
  • Conflict and accident prevention: When your systems require you to pass around typed identifiers like acct_1234 and invoice_5432beef, certain kinds of accidents become impossible. For example, HTTP DELETE /users/invoice_21 fails fast.
  • Future-proofing: Adopting spicy IDs means your systems and APIs are developed to accept a basically-opaque string as an ID. While their underlying type is numeric, in very advanced situations you may be able to migrate to a different type or datastore "behind" the abstraction the string ID creates.

For a more detailed look at this pattern, see Stripe's "Object IDs: Designing APIs for Humans".

Installation

Requirements

This package supports and is tested against the latest patch versions of:

  • Python: 3.10, 3.11, 3.12
  • Django: 3.2, 4.2
  • MySQL: 5.7, 8.0
  • PostgreSQL: 9.5, 10, 11, 12
  • SQLite: 3.9.0+

All database backends are tested with the latest versions of their drivers. SQLite is also tested on GitHub Actions' latest macOS virtual environment.

Note: Django 4.x is recommended due to an upstream bug that is present in 3.x. See #6 for further details.

Instructions

pip install django_spicy_id

Usage

Given the following example model:

from django.db import models
from django_spicy_id import SpicyBigAutoField

class User(models.model):
    id = SpicyBigAutoField(primary_key=True, prefix='usr')

Example usage:

>>> u = models.User.objects.create()
>>> u.id
'usr_1'
>>> u2 = models.User.objects.create(id=123456789)
>>> u2.id
'usr_8M0kX'
>>> found_user = models.User.objects.filter(id='usr_8M0kX').first()
>>> found_user == u2
True

Field types

  • SpicyBigAutoField: A spicy id which is backed by a BigAutoField (i.e. 64-bit int) column.
  • SpicyAutoField: A spicy id which is backed by a AutoField (i.e. 32-bit int) column.
  • SpicySmallAutoField: A spicy id which is backed by a SmallAutoField (i.e. 16-bit int) column.

Required Parameters

The following parameters are required at declaration:

  • prefix: The prefix to use in the encoded form. Typically this is a short, descriptive string, like user or acct and similar. Note: This library does not ensure the string you provide is unique within your project. You should ensure that.

Optional Parameters

In addition to all parameters you can provide a normal AutoField, each of the field types above supports the following additional optional paramters:

  • encoding: What numeric encoding scheme to use. One of django_spicy_id.ENCODING_BASE_62 (default), django_spicy_id.ENCODING_BASE_58, or django_spicy_id.ENCODING_HEX.
  • sep: The separator character. Defaults to _. Can be any string.
  • pad: Whether the encoded portion of the id should be zero-padded so that all values are the same string length. Either False (default) or True.
    • Example without padding: user_8M0kX
    • Example with padding: user_0000008M0kX
  • randomize: If True, the default value of a new record will be generated randomly using secrets.randbelow(). If False (the default), works just like a normal AutoField i.e. the default value comes from the database upon INSERT.
    • When randomize is set, an error will be thrown if default is also set, since randomize is essentially a special and built-in default function.
    • Since randomize installs a special default function, a new but unsaved model instance will have a non-None value for object.id / object.pk. You must used object._state.adding to determine whether an instance is a new but unsaved object.
    • If you use this feature, be aware of its hazards:
      • The generated ID may conflict with an existing row, with probability determined by the birthday problem (i.e. the column size and the size of the existing dataset).
      • A conflict can also arise if two processes generate the same value for secrets.randbelow() (i.e. if system entropy is identical or misconfigured for some reason).

Registering URLs

When installing routes that must match a specific spicy id, you can use the get_url_converter() helper method to install a Django custom path converter.

Using this method will ensure that only valid spicy ID strings for that field will be presented to your view.

Example:

# models.py
class User(models.model):
    id = SpicyBigAutoField(primary_key=True, prefix='usr')
# urls.py
from . import models
from django.urls import path, register_converter
from django_spicy_id import get_url_converter

# Register the pattern for `User.id` as "spicy_user_id". You should do this
# once for each unique spicy ID field.
register_converter(get_url_converter(models.User, 'id'), 'spicy_user_id')

urlpatterns = [
    path('users/<spicy_user_id:id>', views.user_detail),
    ...
]
# views.py

def user_detail(request, id):
  user = models.User.objects.get(id=id)
  ...

Django REST Framework

Django REST Framework (DRF) works mostly without issue with django-spicy-id. However, an additional step is needed so that DRF treats spicy ID fields as strings, not integers, in serializers.

You can use the included utility function to monkey patch DRF. It is safe to call this method multiple times.

from django_spicy_id import monkey_patch_drf

monkey_patch_drf()

Field Attributes

The following attributes are available on the field once constructed

.validate_string(strval)

Checks whether strval is a legal value for the field, throwing django_spicy_id.errors.MalformedSpicyIdError if not.

.re

A compiled regex which can be used to validate a string.

.re_pattern

A string regex pattern which can be used to validate a string. Unlike the pattern used in re, this pattern does not include the leading ^ and trailing $ boundary characters, making it easier to use in things like Django url patterns.

You probably don't need to use this directly, instead see get_url_converter().

Utility methods

These utility methods are provided on the top-level django_spicy_id module.

get_url_converter(model_class, field_name)

Returns a Django custom path converter for field_name on model_class.

See Registering URLs for example usage.

Errors

django.db.utils.ProgrammingError

Thrown when attempting to access or query this field using an illegal value. Some examples of this situation:

  • Providing a spicy id with the wrong prefix or separator (e.g id="acct_1234" where id="invoice_1234" is expected).
  • Providing a string with illegal characters in it (i.e. where the encoded part isn't decodable)
  • Providing an unpadded value when padding is enabled.
  • Providing a padded value when padded is disabled.

You can consider these situations analogous to providing a wrongly-typed object to any other field type, for example SomeModel.objects.filter(id=object()).

You can avoid this situation by validating inputs first. See Field Attributes.

🚨 Warning: Regardless of field configuration, the string value of a spicy id must always be treated as an exact value. Just like you would never modify the contents of a UUID4, a spicy id string must never be translated, re-interpreted, or changed by a client.

django_spicy_id.MalformedSpicyIdError

A subclass of ValueError, raised by .validate_string(strval) when the provided string is invalid for the field's configuration.

Tips and tricks

Don't change field configuration

Changing prefix, sep, pad, or encoding after you have started using the field should be considered a breaking change for any external callers.

Although the stored row IDs are never changed, any spicy IDs generated previously, with a different encoding configuration, may now be invalid or (potentially catastrophically) resolve to a different object.

For just one example, user_10 would naturally refer to a different numeric row id if parsed as hex versus base62 or base58. You should avoid changing the field configuration.

Changelog

See CHANGELOG.md for a summary of changes.