Skip to main content
Help us improve Python packaging – donate today!

Simple DynamoDB object mapper and utils

Project Description


dynamo_objects is a set of tools to work with DynamoDB in python.

It is based on boto and provides following features:

  • a simple object mapper - use object notation to work with dynamo records
  • new tables are automatically created in the database, so you just write and deploy the new code
  • transparent support for table prefixes (multiple databases or multiple environments), you don’t need to handle table prefixes in code, just set the prefix during the database connection
  • simple in-memory dynamodb mock for fast unit tests
  • supports DynamoDB local for slower tests
  • in-memory cached tables to speedup computational operations on top of DynamoDB - all data is read only once and then results are flushed back in a batch
  • additional tools - copy data from table to table, a context manager to update table throughputs and set back once operation is completed

Discussion group


$ pip install dynamo_objects

DB Connection and Table Prefixes

Database connection method adds table prefix support on top of boto’s connect_to_region method. Using the table prefix it is possible to switch the application to different set of tables for different environments (or just use different prefixes for different applications).

Use the following snippet to connect to the database:

from dynamo_objects import DynamoDatabase


Region name, and aws credentials are passed to the boto’s connect_to_region method, so you can use other ways suppored by boto to specify aws credentials. For example, it is not necessary to specify access key id and secret if you use IAM role.

The table_prefix parameter is used to prefix all the table names with the string you specify.

Like if you set the table_prefix to staging_, the application will use tables like staging_user and staging_post. And if you set the prefix to dev_ then application will use dev_user, dev_post.

If you leave the table_prefix empty then it will be just user and post. This way you can easily switch your application from one set of tables to another for different environments (development, staging, production).

To connect to the DynamoDB Local, specify the region_name=’localhost’:

from dynamo_objects import DynamoDatabase

Object Mapper

To use the object mapper, define record and table objects for each DynamoDB table:

from boto.dynamodb2.fields import HashKey, RangeKey
from dynamo_objects import DynamoRecord, DynamoTable

class Store(DynamoRecord):

    def __init__(self, **data):
        # define table fields in the __init__ method
        self.store_id = '' = ''
        self.tags = []
        super(Store, self).__init__(**data)

class StoreTable(DynamoTable):

    def __init__(self):
        super(self.__class__, self).__init__(
            throughput={'read': 3, 'write': 3},

Here the StoreTable class describes the table: table name, schema (hash and optionally range keys), throughput and record class.

And the Store class describes the table row, in the __init__ method we put all the table fields.

See more examples of table/record objects in the tests/ file.

Now the record object can be created and used like this:

store = StoreTable()
store = Store() = 'My Store'

# or initialize the fields using the constructor
store2 = Store(name='My Store 2')
# change the name = 'Another Store'

Compare this to the pure boto code where you have a dictionary-like interface:

store = Item(stores, data={
   name='My Store'
# ....
store['nmae'] = 'Another Store'

If you mistype the field name like in store['nmae'] there will be no error - you will just create a new field in the database. The main purpose of the object mapper is to prevent this.

The DynamoRecord object will raise an exception if you mistype the field name. To actually go schema-less, it is possible to override the _freeze_schema method with pass in the method body.

You can also override the _check_data method to do additional transformations before saving to the database (like convert data types or normalize/unify data format).

Find a record, update it and save:

table = MyTable()
# will raise ItemNotFound exception if record does not exist
record = table.get('my_hash', 'my_range')
record.some_field = 100

# to handle the case when there is no record int the database use
# try/except
from boto.dynamodb2.exceptions import ItemNotFound
    record = table.get('my_hash', 'my_range')
except ItemNotFound:
    # handle the record not found case
    # ...

# sometimes it is more convenient to get None for non-existing record
# `find` method will return None if record does not exist
record = table.find('my_hash', 'my_range')
if record is not None:
    record.some_field = 100

# get a record or create new one if record does not exist
record = table.get('my_hash', 'my_range', create=True)
record.some_field = 200

# delete the existing record
# `delete` method will return the deleted record, so the record data can be
# used for some additional actions like logging
record = table.delete('hash', 'range')

The create=True option for the table.get() method is useful when you want to read the data from the database or get the Null object if data is not found. For example:

class User(DynamoRecord):

  def __init__(self, *data): = 'guest'
     self.password = ''
     super(User, self).__init__(**data)

# Find the user in the database, if not found - the `user` object
# will represent guest user
user = table.get(user_id, create=True)
# print user name or 'guest' (default)

Query and scan methods have the same interface as boto’s query_2 and scan, but will convert the resulting record set into DynamoRecord objects.

# parameters are the same as for boto's query_2
# returns array of records
# don't use when you expect a lot of data, because it will
# fetch all the data from the database and convert to DynamoRecord
# before returning
records = table.query(hash__eq='value', range__gte=50)
records = table.scan(some_field__gte=10)
# get count
count = table.query_count(hash__eq='value', range__gte=50)

Table object also supports the atomic counter update:

# increment the `counter_name` field by 2 for the
# item with hash key = `myhashkey`
table.update_counter('myhashkey', counter_name=2)

# decrement the `counter_name` field by 2 for the
# item with hash key = `myhashkey` and rangekey = 'myrange'
table.update_counter('myhashkey', 'myrange', counter_name=-2)

And it is possible to use boto’s objects directly:

table = MyTable()
# the boto Table object
boto_table = table.table
# ...

record = table.get('my_hash', 'my_range')
# the boto Item object
boto_item = record._item
# ...

Memory tables

Memory tables can be used to cache DynamoDB access in-memory. Every record is only read once and no data is written until you call the save_data or save_data_batch method.

# StoreTable is a regular table definition, DynamoTable subclass
from myschema import StoreTable
from dynamo_objects.memorydb import MemoryTable

class StoreMemoryTable(MemoryTable):

    def __init__(self):
        super(StoreMemoryTable, self).__init__(StoreTable())

Here we define a StoreMemoryTable class for in-memory table which wraps the StoreTable (a regular table definition). Now we can do this:

table = StoreMemoryTable()
# read records with store_id = 1 and store_id = 2
record = table.get(1)
record2 = table.get(2)
# data is not actually saved yet, no write db operations
# ...
# read same records again - will fetch from memory, no db read operations
record = table.get(1)
record2 = table.get(2)
# ...
# data is not actually saved yet, no write db operations
# Now we flush all the data back to DynamoDB
# the `save_data_batch` will use the `batch write` DynamoDB operation

This can be very useful if you do some computational operations and need to read / write a lot of small objects to the database. Depending on the data structure the used read / write throughput and the whole processing time can be noticeably reduced.

Testing and DynamoDB Mock

It is possible to run unit tests using the real DynamoDB connection using the table prefixes feature: you can choose some special table prefix like xx_unit_tests_. This way you’ll have a set of tables for your unit tests.

But this approach is not practical - tests will be slow and will consume the read/write operations (and this will cost money).

Amazon provides a DynamoDB emulator in java but it is problematic to use it during development, because it is slow and consumes a lot of memory.

The solution is a simple in-memory DynamoDB mock module. It is a fast, but very approximate dynamo emulation without permanent data storage.

To enable the mock, just import the dynamock module:

from dynamo_objects import database
# once imported, the `dynamock` module will mock real DynamoDB
# operations and forward them to the simple implementation which
# keeps all the data in memory
from dynamo_objects import dynamock

There is an example of the mock usage in the tests/ module.

This base test module can be used for any project to test parts of code which work with DynamoDB. You can find examples of unit tests under the tests folder. The database schema is described in the tests/

To run all tests use nosetests (install with pip install nose):


By default it will use the in-memory DynamoDB mock. To run tests against the DynamoDB Local use following commands:

# in the first terminal window launch the local dynamodb
# script will download it if necessary

# in another terminal window run the tests
DYNAMODB_MOCK= nosetests

I use fast in-memory mock to run tests locally, during the development.

On the CI server tests a launched two times - first against the in-memory mock and then one more time against the DynamoDB local.

Here is an example of the shell script to do this:

# Run fast tests with in-memory mock

# Run slow tests with DynamoDB local
pushd path/to/folder/with/dynamodb-local
  java -Djava.library.path=./DynamoDBLocal_lib -jar ./DynamoDBLocal.jar -inMemory -sharedDb &
echo "Started local dynamodb: $PID"
DYNAMODB_MOCK= nosetests
kill -9 $PID

Additional Tools

The database module contains few additional useful tools.

The copy_item and copy_table_data methods allow to copy data from table to table (for example, you may want to copy some data from staging to production):

db = database.Database()
# note: table_prefix is empty, so we can explicitly set table names
    region_name='...', ...
num_copied = db.copy_table_data('table_name', 'staging_table_name')

# copy and transform data
def transform_fn(record): = 'staging_' +
db.copy_table_data('table_name', 'staging_table_name', transform=transform_fn)

There are also some other useful methods to create the table, wait until the new table becomes active, delete the table, etc.

The TableThroughput class is a context manager to update (usually set higher) throughput limits and put them back after some operation. It is useful when you need to do something what requires a high read/write throughput.

Using the TableThroughput it is possible to set high limits just before the operation and set them back just after it:

high_throughputs = {
    'table_one': { 'table': { 'read': 100, 'write': 50, }, },
    'table_two': {
        'table': { 'read': 60, 'write': 120, },
        'SomeGlobalIndex': { 'read': 1, 'write': 120 }

with database.TablesThroughput(high_throughputs):
    # now throughputs are high
# now throughputs are low again (same as before the operation)

Release history Release notifications

This version
History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
dynamo_objects-1.0.17.tar.gz (20.8 kB) Copy SHA256 hash SHA256 Source None Apr 26, 2016

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page