ORM For Dynamodb
Project description
Dynamojo
:exclamation: Important Notice :exclamation:
Version 1.0.0 is an important update that introduces breaking changes:
- Pydantic dependency is now at 2.x
DynamojoBase._config
attribute is now an abstractclassmethod namedDynamojoBase.__config()
DynamojoBase._config
must be declared in subclasses ofDynamojoBase
and must return aDynamojoConfig
object
- The following methods are now async:
DynamojoBase.delete
DynamojoBase.fetch
DynamojoBase.query
DynamojoBase.save
DynamojoBase.update
Because one table is better than more
Dynamojo takes the concept of Dynamodb Single Table design and creates a modeling framework for it. This library is opinionated in the following ways:
- Indexes should be generic. They could mean different things for different types of items. An index attribute shouldn't imply that it is always a date, color, etc.
- When using generic indexes the attributes should shadow a human readable attribute. For instance if you have a partition key named "pk" that for items that represent users stores their userid, then there should also be an attribute named userid.
- When creating models for item types that will be stored in the database the developer should only have to worry about their access patterns in terms of the human readable attributes, not be in the weeds of the index design of the table. Mapping items to indexes should happen in code, not in the table definition itself
- Table and Global Secondary indexes should always define a sortkey. There is no reason not to. It's better to have it in cases where you don't need it than to need it and not have it.
Dynamojo is built on top of Pydantic with some bells and whistles:
- put, update, delete, and query db objects
- Dynamically map attributes to the index of your choice. EG: attribute "userId" automatically populates the partition key named "pk"
- Dynamically join attributes into another using a delimiter. For instance create a field that is
<userid>~<date>~<action>
to use as a sort key for fast queries - Mutate attributes when set
- Create models that subclass other models. A common pattern is to define a base class for your project that has a baseline of methods that you will need other than db operations. Different item types would then be created as models that are subclasses from the base class. See test.py
- Flag attributes as immutable so they can't be modified once set
- Use all of the features of
put_item()
,query()
,delete()
, andupdate()
that you normally could withboto3.client("dynamodb")
Limitations:
- Dynamojo doesn't do scans because scans are dumb. I will die on the hill of defending that statement.
- If you have so much data that replicating indexed data into human readable columns is too expensive then this library may not be for you. But if you have that much data you should have a staff of engineers that can write your own library.
See test.py for examples
This library is very opinionated about how the table's indexes should be structured. Below is Terraform that shows the correct way to set up the table. Index keys are never referenced directly when using the table. Rely on IndexMap for that. Since LSI's can only be created at table creation time, and all indexes cost nothing if not used, we go ahead and create all of the indexes that AWS will allow us to when the table is created.
resource "aws_dynamodb_table" "test_table" {
name = "test-dynamojo"
hash_key = "pk"
range_key = "sk"
billing_mode = "PAY_PER_REQUEST"
# LSI attributes
dynamic "attribute" {
for_each = range(5)
content {
name = "lsi${attribute.value}_sk"
type = "S"
}
}
# GSI pk attributes
dynamic "attribute" {
for_each = range(20)
content {
name = "gsi${attribute.value}_pk"
type = "S"
}
}
# GSI sk attributes
dynamic "attribute" {
for_each = range(20)
content {
name = "gsi${attribute.value}_sk"
type = "S"
}
}
attribute {
name = "pk"
type = "S"
}
attribute {
name = "sk"
type = "S"
}
# GSI's
dynamic "global_secondary_index" {
for_each = range(20)
content {
name = "gsi${global_secondary_index.value}"
hash_key = "gsi${global_secondary_index.value}_pk"
range_key = "gsi${global_secondary_index.value}_sk"
projection_type = "ALL"
}
}
# LSI's
dynamic "local_secondary_index" {
for_each = range(5)
content {
name = "lsi${local_secondary_index.value}"
range_key = "lsi${local_secondary_index.value}_sk"
projection_type = "ALL"
}
}
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file dynamojo-1.0.3.tar.gz
.
File metadata
- Download URL: dynamojo-1.0.3.tar.gz
- Upload date:
- Size: 12.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: poetry/1.8.2 CPython/3.12.2 Darwin/23.1.0
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 578dde2b17c562a8fbea4e147f0a5b8ad4d799ceaba3adde2c7b1564dabcb60c |
|
MD5 | da7c6a47bb2b44a52e4105493d1bcca9 |
|
BLAKE2b-256 | b2a922a5656649635a2f9681d52c3889301d30a5490c18ea2649dd08971e7adb |
File details
Details for the file dynamojo-1.0.3-py3-none-any.whl
.
File metadata
- Download URL: dynamojo-1.0.3-py3-none-any.whl
- Upload date:
- Size: 12.2 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: poetry/1.8.2 CPython/3.12.2 Darwin/23.1.0
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 8316b4a26a035a86fec06c2240812c0ef26da9ff8bbaa3cfd5c1cb351f8ed7c7 |
|
MD5 | 12024749b8baaa3027e757ba5ec462b3 |
|
BLAKE2b-256 | 68d18b377fc2dc0e955853e625a3f623dc3c879ea61f67607427a042588e580a |