def _encode_key(cls, hash_key, range_key=None):
dynamizer = Dynamizer()
encoded = {cls._get_hash_key().name: dynamizer.encode(hash_key)}
if range_key:
encoded.update(
{cls._get_range_key().name: dynamizer.encode(range_key)})
return encoded
def _encode_key(cls, hash_key, range_key=None):
dynamizer = Dynamizer()
encoded = {cls._get_hash_key().name: dynamizer.encode(hash_key)}
if range_key:
encoded.update(
{cls._get_range_key().name: dynamizer.encode(range_key)})
return encoded
def __init__(self, table, data=None, loaded=False):
"""
Constructs an (unsaved) ``Item`` instance.
To persist the data in DynamoDB, you'll need to call the ``Item.save``
(or ``Item.partial_save``) on the instance.
Requires a ``table`` parameter, which should be a ``Table`` instance.
This is required, as DynamoDB's API is focus around all operations
being table-level. It's also for persisting schema around many objects.
Optionally accepts a ``data`` parameter, which should be a dictionary
of the fields & values of the item. Alternatively, an ``Item`` instance
may be provided from which to extract the data.
Optionally accepts a ``loaded`` parameter, which should be a boolean.
``True`` if it was preexisting data loaded from DynamoDB, ``False`` if
it's new data from the user. Default is ``False``.
Example::
>>> users = Table('users')
>>> user = Item(users, data={
... 'username': '******',
... 'first_name': 'John',
... 'date_joined': 1248o61592,
... })
# Change existing data.
>>> user['first_name'] = 'Johann'
# Add more data.
>>> user['last_name'] = 'Doe'
# Delete data.
>>> del user['date_joined']
# Iterate over all the data.
>>> for field, val in user.items():
... print "%s: %s" % (field, val)
username: johndoe
first_name: John
date_joined: 1248o61592
"""
self.table = table
self._loaded = loaded
self._orig_data = {}
self._data = data
self._dynamizer = Dynamizer()
if isinstance(self._data, Item):
self._data = self._data._data
if self._data is None:
self._data = {}
if self._loaded:
self._orig_data = deepcopy(self._data)
def __init__(self, table, data=None):
"""
Constructs an (unsaved) ``Item`` instance.
To persist the data in DynamoDB, you'll need to call the ``Item.save``
(or ``Item.partial_save``) on the instance.
Requires a ``table`` parameter, which should be a ``Table`` instance.
This is required, as DynamoDB's API is focus around all operations
being table-level. It's also for persisting schema around many objects.
Optionally accepts a ``data`` parameter, which should be a dictionary
of the fields & values of the item.
Example::
>>> users = Table('users')
>>> user = Item(users, data={
... 'username': '******',
... 'first_name': 'John',
... 'date_joined': 1248o61592,
... })
# Change existing data.
>>> user['first_name'] = 'Johann'
# Add more data.
>>> user['last_name'] = 'Doe'
# Delete data.
>>> del user['date_joined']
# Iterate over all the data.
>>> for field, val in user.items():
... print "%s: %s" % (field, val)
username: johndoe
first_name: John
date_joined: 1248o61592
"""
self.table = table
self._data = {}
self._orig_data = {}
self._is_dirty = False
self._dynamizer = Dynamizer()
if data:
self._data = data
self._is_dirty = True
for key in data.keys():
self._orig_data[key] = NEWVALUE
def _decode_item(self, encoded_item):
encoded_item_cp = encoded_item.copy()
# only 1-depth supported.
# TODO: provide n-depth of dict
for key, val in encoded_item_cp.iteritems():
encoded_item_cp.update({key: Dynamizer().decode(val)})
return encoded_item_cp
def __init__(self, table, data=None, loaded=False):
"""
Constructs an (unsaved) ``Item`` instance.
To persist the data in DynamoDB, you'll need to call the ``Item.save``
(or ``Item.partial_save``) on the instance.
Requires a ``table`` parameter, which should be a ``Table`` instance.
This is required, as DynamoDB's API is focus around all operations
being table-level. It's also for persisting schema around many objects.
Optionally accepts a ``data`` parameter, which should be a dictionary
of the fields & values of the item. Alternatively, an ``Item`` instance
may be provided from which to extract the data.
Optionally accepts a ``loaded`` parameter, which should be a boolean.
``True`` if it was preexisting data loaded from DynamoDB, ``False`` if
it's new data from the user. Default is ``False``.
Example::
>>> users = Table('users')
>>> user = Item(users, data={
... 'username': '******',
... 'first_name': 'John',
... 'date_joined': 1248o61592,
... })
# Change existing data.
>>> user['first_name'] = 'Johann'
# Add more data.
>>> user['last_name'] = 'Doe'
# Delete data.
>>> del user['date_joined']
# Iterate over all the data.
>>> for field, val in user.items():
... print "%s: %s" % (field, val)
username: johndoe
first_name: John
date_joined: 1248o61592
"""
self.table = table
self._loaded = loaded
self._orig_data = {}
self._data = data
self._dynamizer = Dynamizer()
if isinstance(self._data, Item):
self._data = self._data._data
if self._data is None:
self._data = {}
if self._loaded:
self._orig_data = deepcopy(self._data)
def update_item(cls, hash_key, range_key=None, attributes_to_set=None, attributes_to_add=None):
"""Update item attributes. Currently SET and ADD actions are supported."""
primary_key = cls._encode_key(hash_key, range_key)
value_names = {}
encoded_values = {}
dynamizer = Dynamizer()
set_expression = ''
if attributes_to_set:
for i, key in enumerate(attributes_to_set.keys()):
value_name = ':s{0}'.format(i)
value_names[key] = value_name
encoded_values[value_name] = dynamizer.encode(attributes_to_set[key])
set_expression = 'SET {0}'.format(
', '.join(
'{key}={value_name}'.format(key=key, value_name=value_names[key]) for key in attributes_to_set
)
)
add_expression = ''
if attributes_to_add:
for i, key in enumerate(attributes_to_add.keys()):
value_name = ':a{0}'.format(i)
value_names[key] = value_name
encoded_values[value_name] = dynamizer.encode(attributes_to_add[key])
add_expression = 'ADD {0}'.format(
', '.join(
'{key} {value_name}'.format(key=key, value_name=value_names[key]) for key in attributes_to_add
)
)
update_expression = ' '.join([set_expression, add_expression])
cls._get_connection().update_item(
cls.get_table_name(),
primary_key,
update_expression=update_expression, expression_attribute_values=encoded_values)
def __init__(self, table, data=None):
"""
Constructs an (unsaved) ``Item`` instance.
To persist the data in DynamoDB, you'll need to call the ``Item.save``
(or ``Item.partial_save``) on the instance.
Requires a ``table`` parameter, which should be a ``Table`` instance.
This is required, as DynamoDB's API is focus around all operations
being table-level. It's also for persisting schema around many objects.
Optionally accepts a ``data`` parameter, which should be a dictionary
of the fields & values of the item.
Example::
>>> users = Table('users')
>>> user = Item(users, data={
... 'username': '******',
... 'first_name': 'John',
... 'date_joined': 1248o61592,
... })
# Change existing data.
>>> user['first_name'] = 'Johann'
# Add more data.
>>> user['last_name'] = 'Doe'
# Delete data.
>>> del user['date_joined']
# Iterate over all the data.
>>> for field, val in user.items():
... print "%s: %s" % (field, val)
username: johndoe
first_name: John
date_joined: 1248o61592
"""
self.table = table
self._data = {}
self._orig_data = {}
self._is_dirty = False
self._dynamizer = Dynamizer()
if data:
self._data = data
self._is_dirty = True
for key in list(data.keys()):
self._orig_data[key] = NEWVALUE
class Table(object):
"""
Interacts & models the behavior of a DynamoDB table.
The ``Table`` object represents a set (or rough categorization) of
records within DynamoDB. The important part is that all records within the
table, while largely-schema-free, share the same schema & are essentially
namespaced for use in your application. For example, you might have a
``users`` table or a ``forums`` table.
"""
max_batch_get = 100
def __init__(self, table_name, schema=None, throughput=None, indexes=None,
global_indexes=None, connection=None):
"""
Sets up a new in-memory ``Table``.
This is useful if the table already exists within DynamoDB & you simply
want to use it for additional interactions. The only required parameter
is the ``table_name``. However, under the hood, the object will call
``describe_table`` to determine the schema/indexes/throughput. You
can avoid this extra call by passing in ``schema`` & ``indexes``.
**IMPORTANT** - If you're creating a new ``Table`` for the first time,
you should use the ``Table.create`` method instead, as it will
persist the table structure to DynamoDB.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Optionally accepts a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``global_indexes`` parameter, which should be a
list of ``GlobalBaseIndexField`` subclasses representing the desired
indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
# The simple, it-already-exists case.
>>> conn = Table('users')
# The full, minimum-extra-calls case.
>>> from boto import dynamodb2
>>> users = Table('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... HashKey('username')
... RangeKey('date_joined')
... ]),
... ], global_indexes=[
... GlobalAllIndex('UsersByZipcode', parts=[
... HashKey('zipcode'),
... RangeKey('username'),
... ],
... throughput={
... 'read':10,
... 'write":10,
... }),
... ], connection=dynamodb2.connect_to_region('us-west-2',
... aws_access_key_id='key',
... aws_secret_access_key='key',
... ))
"""
self.table_name = table_name
self.connection = connection
self.throughput = {
'read': 5,
'write': 5,
}
self.schema = schema
self.indexes = indexes
self.global_indexes = global_indexes
if self.connection is None:
self.connection = DynamoDBConnection()
if throughput is not None:
self.throughput = throughput
self._dynamizer = Dynamizer()
@classmethod
def create(cls, table_name, schema, throughput=None, indexes=None,
global_indexes=None, connection=None):
"""
Creates a new table in DynamoDB & returns an in-memory ``Table`` object.
This will setup a brand new table within DynamoDB. The ``table_name``
must be unique for your AWS account. The ``schema`` is also required
to define the key structure of the table.
**IMPORTANT** - You should consider the usage pattern of your table
up-front, as the schema & indexes can **NOT** be modified once the
table is created, requiring the creation of a new table & migrating
the data should you wish to revise it.
**IMPORTANT** - If the table already exists in DynamoDB, additional
calls to this method will result in an error. If you just need
a ``Table`` object to interact with the existing table, you should
just initialize a new ``Table`` object, which requires only the
``table_name``.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Requires a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``global_indexes`` parameter, which should be a
list of ``GlobalBaseIndexField`` subclasses representing the desired
indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
>>> users = Table.create('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... RangeKey('date_joined')
... ]), global_indexes=[
... GlobalAllIndex('UsersByZipcode', parts=[
... HashKey('zipcode'),
... RangeKey('username'),
... ],
... throughput={
... 'read':10,
... 'write':10,
... }),
... ])
"""
table = cls(table_name=table_name, connection=connection)
table.schema = schema
if throughput is not None:
table.throughput = throughput
if indexes is not None:
table.indexes = indexes
if global_indexes is not None:
table.global_indexes = global_indexes
# Prep the schema.
raw_schema = []
attr_defs = []
seen_attrs = set()
for field in table.schema:
raw_schema.append(field.schema())
# Build the attributes off what we know.
seen_attrs.add(field.name)
attr_defs.append(field.definition())
raw_throughput = {
'ReadCapacityUnits': int(table.throughput['read']),
'WriteCapacityUnits': int(table.throughput['write']),
}
kwargs = {}
kwarg_map = {
'indexes': 'local_secondary_indexes',
'global_indexes': 'global_secondary_indexes',
}
for index_attr in ('indexes', 'global_indexes'):
table_indexes = getattr(table, index_attr)
if table_indexes:
raw_indexes = []
for index_field in table_indexes:
raw_indexes.append(index_field.schema())
# Make sure all attributes specified in the indexes are
# added to the definition
for field in index_field.parts:
if field.name not in seen_attrs:
seen_attrs.add(field.name)
attr_defs.append(field.definition())
kwargs[kwarg_map[index_attr]] = raw_indexes
table.connection.create_table(
table_name=table.table_name,
attribute_definitions=attr_defs,
key_schema=raw_schema,
provisioned_throughput=raw_throughput,
**kwargs
)
return table
def _introspect_schema(self, raw_schema, raw_attributes=None):
"""
Given a raw schema structure back from a DynamoDB response, parse
out & build the high-level Python objects that represent them.
"""
schema = []
sane_attributes = {}
if raw_attributes:
for field in raw_attributes:
sane_attributes[field['AttributeName']] = field['AttributeType']
for field in raw_schema:
data_type = sane_attributes.get(field['AttributeName'], STRING)
if field['KeyType'] == 'HASH':
schema.append(
HashKey(field['AttributeName'], data_type=data_type)
)
elif field['KeyType'] == 'RANGE':
schema.append(
RangeKey(field['AttributeName'], data_type=data_type)
)
else:
raise exceptions.UnknownSchemaFieldError(
"%s was seen, but is unknown. Please report this at "
"https://github.com/boto/boto/issues." % field['KeyType']
)
return schema
def _introspect_indexes(self, raw_indexes):
"""
Given a raw index structure back from a DynamoDB response, parse
out & build the high-level Python objects that represent them.
"""
indexes = []
for field in raw_indexes:
index_klass = AllIndex
kwargs = {
'parts': []
}
if field['Projection']['ProjectionType'] == 'ALL':
index_klass = AllIndex
elif field['Projection']['ProjectionType'] == 'KEYS_ONLY':
index_klass = KeysOnlyIndex
elif field['Projection']['ProjectionType'] == 'INCLUDE':
index_klass = IncludeIndex
kwargs['includes'] = field['Projection']['NonKeyAttributes']
else:
raise exceptions.UnknownIndexFieldError(
"%s was seen, but is unknown. Please report this at "
"https://github.com/boto/boto/issues." % \
field['Projection']['ProjectionType']
)
name = field['IndexName']
kwargs['parts'] = self._introspect_schema(field['KeySchema'], None)
indexes.append(index_klass(name, **kwargs))
return indexes
def describe(self):
"""
Describes the current structure of the table in DynamoDB.
This information will be used to update the ``schema``, ``indexes``
and ``throughput`` information on the ``Table``. Some calls, such as
those involving creating keys or querying, will require this
information to be populated.
It also returns the full raw datastructure from DynamoDB, in the
event you'd like to parse out additional information (such as the
``ItemCount`` or usage information).
Example::
>>> users.describe()
{
# Lots of keys here...
}
>>> len(users.schema)
2
"""
result = self.connection.describe_table(self.table_name)
# Blindly update throughput, since what's on DynamoDB's end is likely
# more correct.
raw_throughput = result['Table']['ProvisionedThroughput']
self.throughput['read'] = int(raw_throughput['ReadCapacityUnits'])
self.throughput['write'] = int(raw_throughput['WriteCapacityUnits'])
if not self.schema:
# Since we have the data, build the schema.
raw_schema = result['Table'].get('KeySchema', [])
raw_attributes = result['Table'].get('AttributeDefinitions', [])
self.schema = self._introspect_schema(raw_schema, raw_attributes)
if not self.indexes:
# Build the index information as well.
raw_indexes = result['Table'].get('LocalSecondaryIndexes', [])
self.indexes = self._introspect_indexes(raw_indexes)
# This is leaky.
return result
def update(self, throughput, global_indexes=None):
"""
Updates table attributes in DynamoDB.
Currently, the only thing you can modify about a table after it has
been created is the throughput.
Requires a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Returns ``True`` on success.
Example::
# For a read-heavier application...
>>> users.update(throughput={
... 'read': 20,
... 'write': 10,
... })
True
# To also update the global index(es) throughput.
>>> users.update(throughput={
... 'read': 20,
... 'write': 10,
... },
... global_secondary_indexes={
... 'TheIndexNameHere': {
... 'read': 15,
... 'write': 5,
... }
... })
True
"""
self.throughput = throughput
data = {
'ReadCapacityUnits': int(self.throughput['read']),
'WriteCapacityUnits': int(self.throughput['write']),
}
gsi_data = None
if global_indexes:
gsi_data = []
for gsi_name, gsi_throughput in global_indexes.items():
gsi_data.append({
"Update": {
"IndexName": gsi_name,
"ProvisionedThroughput": {
"ReadCapacityUnits": int(gsi_throughput['read']),
"WriteCapacityUnits": int(gsi_throughput['write']),
},
},
})
self.connection.update_table(
self.table_name,
provisioned_throughput=data,
global_secondary_index_updates=gsi_data
)
return True
def delete(self):
"""
Deletes a table in DynamoDB.
**IMPORTANT** - Be careful when using this method, there is no undo.
Returns ``True`` on success.
Example::
>>> users.delete()
True
"""
self.connection.delete_table(self.table_name)
return True
def _encode_keys(self, keys):
"""
Given a flat Python dictionary of keys/values, converts it into the
nested dictionary DynamoDB expects.
Converts::
{
'username': '******',
'tags': [1, 2, 5],
}
...to...::
{
'username': {'S': 'john'},
'tags': {'NS': ['1', '2', '5']},
}
"""
raw_key = {}
for key, value in keys.items():
raw_key[key] = self._dynamizer.encode(value)
return raw_key
def get_item(self, consistent=False, attributes=None, **kwargs):
"""
Fetches an item (record) from a table in DynamoDB.
To specify the key of the item you'd like to get, you can specify the
key attributes as kwargs.
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will perform
a consistent (but more expensive) read from DynamoDB.
(Default: ``False``)
Optionally accepts an ``attributes`` parameter, which should be a
list of fieldname to fetch. (Default: ``None``, which means all fields
should be fetched)
Returns an ``Item`` instance containing all the data for that record.
Example::
# A simple hash key.
>>> john = users.get_item(username='******')
>>> john['first_name']
'John'
# A complex hash+range key.
>>> john = users.get_item(username='******', last_name='Doe')
>>> john['first_name']
'John'
# A consistent read (assuming the data might have just changed).
>>> john = users.get_item(username='******', consistent=True)
>>> john['first_name']
'Johann'
# With a key that is an invalid variable name in Python.
# Also, assumes a different schema than previous examples.
>>> john = users.get_item(**{
... 'date-joined': 127549192,
... })
>>> john['first_name']
'John'
"""
raw_key = self._encode_keys(kwargs)
item_data = self.connection.get_item(
self.table_name,
raw_key,
attributes_to_get=attributes,
consistent_read=consistent
)
if 'Item' not in item_data:
raise exceptions.ItemNotFound("Item %s couldn't be found." % kwargs)
item = Item(self)
item.load(item_data)
return item
def has_item(self, **kwargs):
"""
Return whether an item (record) exists within a table in DynamoDB.
To specify the key of the item you'd like to get, you can specify the
key attributes as kwargs.
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will perform
a consistent (but more expensive) read from DynamoDB.
(Default: ``False``)
Optionally accepts an ``attributes`` parameter, which should be a
list of fieldnames to fetch. (Default: ``None``, which means all fields
should be fetched)
Returns ``True`` if an ``Item`` is present, ``False`` if not.
Example::
# Simple, just hash-key schema.
>>> users.has_item(username='******')
True
# Complex schema, item not present.
>>> users.has_item(
... username='******',
... date_joined='2014-01-07'
... )
False
"""
try:
self.get_item(**kwargs)
except (JSONResponseError, exceptions.ItemNotFound):
return False
return True
def lookup(self, *args, **kwargs):
"""
Look up an entry in DynamoDB. This is mostly backwards compatible
with boto.dynamodb. Unlike get_item, it takes hash_key and range_key first,
although you may still specify keyword arguments instead.
Also unlike the get_item command, if the returned item has no keys
(i.e., it does not exist in DynamoDB), a None result is returned, instead
of an empty key object.
Example::
>>> user = users.lookup(username)
>>> user = users.lookup(username, consistent=True)
>>> app = apps.lookup('my_customer_id', 'my_app_id')
"""
if not self.schema:
self.describe()
for x, arg in enumerate(args):
kwargs[self.schema[x].name] = arg
ret = self.get_item(**kwargs)
if not ret.keys():
return None
return ret
def new_item(self, *args):
"""
Returns a new, blank item
This is mostly for consistency with boto.dynamodb
"""
if not self.schema:
self.describe()
data = {}
for x, arg in enumerate(args):
data[self.schema[x].name] = arg
return Item(self, data=data)
def put_item(self, data, overwrite=False):
"""
Saves an entire item to DynamoDB.
By default, if any part of the ``Item``'s original data doesn't match
what's currently in DynamoDB, this request will fail. This prevents
other processes from updating the data in between when you read the
item & when your request to update the item's data is processed, which
would typically result in some data loss.
Requires a ``data`` parameter, which should be a dictionary of the data
you'd like to store in DynamoDB.
Optionally accepts an ``overwrite`` parameter, which should be a
boolean. If you provide ``True``, this will tell DynamoDB to blindly
overwrite whatever data is present, if any.
Returns ``True`` on success.
Example::
>>> users.put_item(data={
... 'username': '******',
... 'first_name': 'Jane',
... 'last_name': 'Doe',
... 'date_joined': 126478915,
... })
True
"""
item = Item(self, data=data)
return item.save(overwrite=overwrite)
def _put_item(self, item_data, expects=None):
"""
The internal variant of ``put_item`` (full data). This is used by the
``Item`` objects, since that operation is represented at the
table-level by the API, but conceptually maps better to telling an
individual ``Item`` to save itself.
"""
kwargs = {}
if expects is not None:
kwargs['expected'] = expects
self.connection.put_item(self.table_name, item_data, **kwargs)
return True
def _update_item(self, key, item_data, expects=None):
"""
The internal variant of ``put_item`` (partial data). This is used by the
``Item`` objects, since that operation is represented at the
table-level by the API, but conceptually maps better to telling an
individual ``Item`` to save itself.
"""
raw_key = self._encode_keys(key)
kwargs = {}
if expects is not None:
kwargs['expected'] = expects
self.connection.update_item(self.table_name, raw_key, item_data, **kwargs)
return True
def delete_item(self, **kwargs):
"""
Deletes an item in DynamoDB.
**IMPORTANT** - Be careful when using this method, there is no undo.
To specify the key of the item you'd like to get, you can specify the
key attributes as kwargs.
Returns ``True`` on success.
Example::
# A simple hash key.
>>> users.delete_item(username='******')
True
# A complex hash+range key.
>>> users.delete_item(username='******', last_name='Doe')
True
# With a key that is an invalid variable name in Python.
# Also, assumes a different schema than previous examples.
>>> users.delete_item(**{
... 'date-joined': 127549192,
... })
True
"""
raw_key = self._encode_keys(kwargs)
self.connection.delete_item(self.table_name, raw_key)
return True
def get_key_fields(self):
"""
Returns the fields necessary to make a key for a table.
If the ``Table`` does not already have a populated ``schema``,
this will request it via a ``Table.describe`` call.
Returns a list of fieldnames (strings).
Example::
# A simple hash key.
>>> users.get_key_fields()
['username']
# A complex hash+range key.
>>> users.get_key_fields()
['username', 'last_name']
"""
if not self.schema:
# We don't know the structure of the table. Get a description to
# populate the schema.
self.describe()
return [field.name for field in self.schema]
def batch_write(self):
"""
Allows the batching of writes to DynamoDB.
Since each write/delete call to DynamoDB has a cost associated with it,
when loading lots of data, it makes sense to batch them, creating as
few calls as possible.
This returns a context manager that will transparently handle creating
these batches. The object you get back lightly-resembles a ``Table``
object, sharing just the ``put_item`` & ``delete_item`` methods
(which are all that DynamoDB can batch in terms of writing data).
DynamoDB's maximum batch size is 25 items per request. If you attempt
to put/delete more than that, the context manager will batch as many
as it can up to that number, then flush them to DynamoDB & continue
batching as more calls come in.
Example::
# Assuming a table with one record...
>>> with users.batch_write() as batch:
... batch.put_item(data={
... 'username': '******',
... 'first_name': 'John',
... 'last_name': 'Doe',
... 'owner': 1,
... })
... # Nothing across the wire yet.
... batch.delete_item(username='******')
... # Still no requests sent.
... batch.put_item(data={
... 'username': '******',
... 'first_name': 'Jane',
... 'last_name': 'Doe',
... 'date_joined': 127436192,
... })
... # Nothing yet, but once we leave the context, the
... # put/deletes will be sent.
"""
# PHENOMENAL COSMIC DOCS!!! itty-bitty code.
return BatchTable(self)
def _build_filters(self, filter_kwargs, using=QUERY_OPERATORS):
"""
An internal method for taking query/scan-style ``**kwargs`` & turning
them into the raw structure DynamoDB expects for filtering.
"""
filters = {}
for field_and_op, value in filter_kwargs.items():
field_bits = field_and_op.split('__')
fieldname = '__'.join(field_bits[:-1])
try:
op = using[field_bits[-1]]
except KeyError:
raise exceptions.UnknownFilterTypeError(
"Operator '%s' from '%s' is not recognized." % (
field_bits[-1],
field_and_op
)
)
lookup = {
'AttributeValueList': [],
'ComparisonOperator': op,
}
# Special-case the ``NULL/NOT_NULL`` case.
if field_bits[-1] == 'null':
del lookup['AttributeValueList']
if value is False:
lookup['ComparisonOperator'] = 'NOT_NULL'
else:
lookup['ComparisonOperator'] = 'NULL'
# Special-case the ``BETWEEN`` case.
elif field_bits[-1] == 'between':
if len(value) == 2 and isinstance(value, (list, tuple)):
lookup['AttributeValueList'].append(
self._dynamizer.encode(value[0])
)
lookup['AttributeValueList'].append(
self._dynamizer.encode(value[1])
)
# Special-case the ``IN`` case
elif field_bits[-1] == 'in':
for val in value:
lookup['AttributeValueList'].append(self._dynamizer.encode(val))
else:
# Fix up the value for encoding, because it was built to only work
# with ``set``s.
if isinstance(value, (list, tuple)):
value = set(value)
lookup['AttributeValueList'].append(
self._dynamizer.encode(value)
)
# Finally, insert it into the filters.
filters[fieldname] = lookup
return filters
def query(self, limit=None, index=None, reverse=False, consistent=False,
attributes=None, max_page_size=None, **filter_kwargs):
"""
**WARNING:** This method is provided **strictly** for
backward-compatibility. It returns results in an incorrect order.
If you are writing new code, please use ``Table.query_2``.
"""
reverse = not reverse
return self.query_2(limit=limit, index=index, reverse=reverse,
consistent=consistent, attributes=attributes,
max_page_size=max_page_size, **filter_kwargs)
def query_2(self, limit=None, index=None, reverse=False,
consistent=False, attributes=None, max_page_size=None,
**filter_kwargs):
"""
Queries for a set of matching items in a DynamoDB table.
Queries can be performed against a hash key, a hash+range key or
against any data stored in your local secondary indexes.
**Note** - You can not query against arbitrary fields within the data
stored in DynamoDB.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts a ``limit`` parameter, which should be an integer
count of the total number of items to return. (Default: ``None`` -
all results)
Optionally accepts an ``index`` parameter, which should be a string of
name of the local secondary index you want to query against.
(Default: ``None``)
Optionally accepts a ``reverse`` parameter, which will present the
results in reverse order. (Default: ``False`` - normal order)
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will force a consistent read of
the data (more expensive). (Default: ``False`` - use eventually
consistent reads)
Optionally accepts a ``attributes`` parameter, which should be a
tuple. If you provide any attributes only these will be fetched
from DynamoDB. This uses the ``AttributesToGet`` and set's
``Select`` to ``SPECIFIC_ATTRIBUTES`` API.
Optionally accepts a ``max_page_size`` parameter, which should be an
integer count of the maximum number of items to retrieve
**per-request**. This is useful in making faster requests & prevent
the scan from drowning out other queries. (Default: ``None`` -
fetch as many as DynamoDB will return)
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
# Look for last names equal to "Doe".
>>> results = users.query(last_name__eq='Doe')
>>> for res in results:
... print res['first_name']
'John'
'Jane'
# Look for last names beginning with "D", in reverse order, limit 3.
>>> results = users.query(
... last_name__beginswith='D',
... reverse=True,
... limit=3
... )
>>> for res in results:
... print res['first_name']
'Alice'
'Jane'
'John'
# Use an LSI & a consistent read.
>>> results = users.query(
... date_joined__gte=1236451000,
... owner__eq=1,
... index='DateJoinedIndex',
... consistent=True
... )
>>> for res in results:
... print res['first_name']
'Alice'
'Bob'
'John'
'Fred'
"""
if self.schema:
if len(self.schema) == 1:
if len(filter_kwargs) <= 1:
if not self.global_indexes or not len(self.global_indexes):
# If the schema only has one field, there's <= 1 filter
# param & no Global Secondary Indexes, this is user
# error. Bail early.
raise exceptions.QueryError(
"You must specify more than one key to filter on."
)
if attributes is not None:
select = 'SPECIFIC_ATTRIBUTES'
else:
select = None
results = ResultSet(
max_page_size=max_page_size
)
kwargs = filter_kwargs.copy()
kwargs.update({
'limit': limit,
'index': index,
'reverse': reverse,
'consistent': consistent,
'select': select,
'attributes_to_get': attributes,
})
results.to_call(self._query, **kwargs)
return results
def query_count(self, index=None, consistent=False, **filter_kwargs):
"""
Queries the exact count of matching items in a DynamoDB table.
Queries can be performed against a hash key, a hash+range key or
against any data stored in your local secondary indexes.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts an ``index`` parameter, which should be a string of
name of the local secondary index you want to query against.
(Default: ``None``)
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will force a consistent read of
the data (more expensive). (Default: ``False`` - use eventually
consistent reads)
Returns an integer which represents the exact amount of matched
items.
Example::
# Look for last names equal to "Doe".
>>> users.query_count(last_name__eq='Doe')
5
# Use an LSI & a consistent read.
>>> users.query_count(
... date_joined__gte=1236451000,
... owner__eq=1,
... index='DateJoinedIndex',
... consistent=True
... )
2
"""
key_conditions = self._build_filters(
filter_kwargs,
using=QUERY_OPERATORS
)
raw_results = self.connection.query(
self.table_name,
index_name=index,
consistent_read=consistent,
select='COUNT',
key_conditions=key_conditions,
)
return int(raw_results.get('Count', 0))
def _query(self, limit=None, index=None, reverse=False, consistent=False,
exclusive_start_key=None, select=None, attributes_to_get=None,
**filter_kwargs):
"""
The internal method that performs the actual queries. Used extensively
by ``ResultSet`` to perform each (paginated) request.
"""
kwargs = {
'limit': limit,
'index_name': index,
'consistent_read': consistent,
'select': select,
'attributes_to_get': attributes_to_get,
}
if reverse:
kwargs['scan_index_forward'] = False
if exclusive_start_key:
kwargs['exclusive_start_key'] = {}
for key, value in exclusive_start_key.items():
kwargs['exclusive_start_key'][key] = \
self._dynamizer.encode(value)
# Convert the filters into something we can actually use.
kwargs['key_conditions'] = self._build_filters(
filter_kwargs,
using=QUERY_OPERATORS
)
raw_results = self.connection.query(
self.table_name,
**kwargs
)
results = []
last_key = None
for raw_item in raw_results.get('Items', []):
item = Item(self)
item.load({
'Item': raw_item,
})
results.append(item)
if raw_results.get('LastEvaluatedKey', None):
last_key = {}
for key, value in raw_results['LastEvaluatedKey'].items():
last_key[key] = self._dynamizer.decode(value)
return {
'results': results,
'last_key': last_key,
}
def scan(self, limit=None, segment=None, total_segments=None,
max_page_size=None, attributes=None, **filter_kwargs):
"""
Scans across all items within a DynamoDB table.
Scans can be performed against a hash key or a hash+range key. You can
additionally filter the results after the table has been read but
before the response is returned.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts a ``limit`` parameter, which should be an integer
count of the total number of items to return. (Default: ``None`` -
all results)
Optionally accepts a ``segment`` parameter, which should be an integer
of the segment to retrieve on. Please see the documentation about
Parallel Scans (Default: ``None`` - no segments)
Optionally accepts a ``total_segments`` parameter, which should be an
integer count of number of segments to divide the table into.
Please see the documentation about Parallel Scans (Default: ``None`` -
no segments)
Optionally accepts a ``max_page_size`` parameter, which should be an
integer count of the maximum number of items to retrieve
**per-request**. This is useful in making faster requests & prevent
the scan from drowning out other queries. (Default: ``None`` -
fetch as many as DynamoDB will return)
Optionally accepts an ``attributes`` parameter, which should be a
tuple. If you provide any attributes only these will be fetched
from DynamoDB. This uses the ``AttributesToGet`` and set's
``Select`` to ``SPECIFIC_ATTRIBUTES`` API.
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
# All results.
>>> everything = users.scan()
# Look for last names beginning with "D".
>>> results = users.scan(last_name__beginswith='D')
>>> for res in results:
... print res['first_name']
'Alice'
'John'
'Jane'
# Use an ``IN`` filter & limit.
>>> results = users.scan(
... age__in=[25, 26, 27, 28, 29],
... limit=1
... )
>>> for res in results:
... print res['first_name']
'Alice'
"""
results = ResultSet(
max_page_size=max_page_size
)
kwargs = filter_kwargs.copy()
kwargs.update({
'limit': limit,
'segment': segment,
'total_segments': total_segments,
'attributes': attributes,
})
results.to_call(self._scan, **kwargs)
return results
def _scan(self, limit=None, exclusive_start_key=None, segment=None,
total_segments=None, attributes=None, **filter_kwargs):
"""
The internal method that performs the actual scan. Used extensively
by ``ResultSet`` to perform each (paginated) request.
"""
kwargs = {
'limit': limit,
'segment': segment,
'total_segments': total_segments,
'attributes_to_get': attributes,
}
if exclusive_start_key:
kwargs['exclusive_start_key'] = {}
for key, value in exclusive_start_key.items():
kwargs['exclusive_start_key'][key] = \
self._dynamizer.encode(value)
# Convert the filters into something we can actually use.
kwargs['scan_filter'] = self._build_filters(
filter_kwargs,
using=FILTER_OPERATORS
)
raw_results = self.connection.scan(
self.table_name,
**kwargs
)
results = []
last_key = None
for raw_item in raw_results.get('Items', []):
item = Item(self)
item.load({
'Item': raw_item,
})
results.append(item)
if raw_results.get('LastEvaluatedKey', None):
last_key = {}
for key, value in raw_results['LastEvaluatedKey'].items():
last_key[key] = self._dynamizer.decode(value)
return {
'results': results,
'last_key': last_key,
}
def batch_get(self, keys, consistent=False):
"""
Fetches many specific items in batch from a table.
Requires a ``keys`` parameter, which should be a list of dictionaries.
Each dictionary should consist of the keys values to specify.
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, a strongly consistent read will be
used. (Default: False)
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
>>> results = users.batch_get(keys=[
... {
... 'username': '******',
... },
... {
... 'username': '******',
... },
... {
... 'username': '******',
... },
... ])
>>> for res in results:
... print res['first_name']
'John'
'Jane'
'Fred'
"""
# We pass the keys to the constructor instead, so it can maintain it's
# own internal state as to what keys have been processed.
results = BatchGetResultSet(keys=keys, max_batch_get=self.max_batch_get)
results.to_call(self._batch_get, consistent=False)
return results
def _batch_get(self, keys, consistent=False):
"""
The internal method that performs the actual batch get. Used extensively
by ``BatchGetResultSet`` to perform each (paginated) request.
"""
items = {
self.table_name: {
'Keys': [],
},
}
if consistent:
items[self.table_name]['ConsistentRead'] = True
for key_data in keys:
raw_key = {}
for key, value in key_data.items():
raw_key[key] = self._dynamizer.encode(value)
items[self.table_name]['Keys'].append(raw_key)
raw_results = self.connection.batch_get_item(request_items=items)
results = []
unprocessed_keys = []
for raw_item in raw_results['Responses'].get(self.table_name, []):
item = Item(self)
item.load({
'Item': raw_item,
})
results.append(item)
raw_unproccessed = raw_results.get('UnprocessedKeys', {})
for raw_key in raw_unproccessed.get('Keys', []):
py_key = {}
for key, value in raw_key.items():
py_key[key] = self._dynamizer.decode(value)
unprocessed_keys.append(py_key)
return {
'results': results,
# NEVER return a ``last_key``. Just in-case any part of
# ``ResultSet`` peeks through, since much of the
# original underlying implementation is based on this key.
'last_key': None,
'unprocessed_keys': unprocessed_keys,
}
def count(self):
"""
Returns a (very) eventually consistent count of the number of items
in a table.
Lag time is about 6 hours, so don't expect a high degree of accuracy.
Example::
>>> users.count()
6
"""
info = self.describe()
return info['Table'].get('ItemCount', 0)
class Item(object):
"""
An object representing the item data within a DynamoDB table.
An item is largely schema-free, meaning it can contain any data. The only
limitation is that it must have data for the fields in the ``Table``'s
schema.
This object presents a dictionary-like interface for accessing/storing
data. It also tries to intelligently track how data has changed throughout
the life of the instance, to be as efficient as possible about updates.
"""
def __init__(self, table, data=None):
"""
Constructs an (unsaved) ``Item`` instance.
To persist the data in DynamoDB, you'll need to call the ``Item.save``
(or ``Item.partial_save``) on the instance.
Requires a ``table`` parameter, which should be a ``Table`` instance.
This is required, as DynamoDB's API is focus around all operations
being table-level. It's also for persisting schema around many objects.
Optionally accepts a ``data`` parameter, which should be a dictionary
of the fields & values of the item.
Example::
>>> users = Table('users')
>>> user = Item(users, data={
... 'username': '******',
... 'first_name': 'John',
... 'date_joined': 1248o61592,
... })
# Change existing data.
>>> user['first_name'] = 'Johann'
# Add more data.
>>> user['last_name'] = 'Doe'
# Delete data.
>>> del user['date_joined']
# Iterate over all the data.
>>> for field, val in user.items():
... print "%s: %s" % (field, val)
username: johndoe
first_name: John
date_joined: 1248o61592
"""
self.table = table
self._data = {}
self._orig_data = {}
self._is_dirty = False
self._dynamizer = Dynamizer()
if data:
self._data = data
self._is_dirty = True
for key in list(data.keys()):
self._orig_data[key] = NEWVALUE
def __getitem__(self, key):
return self._data.get(key, None)
def __setitem__(self, key, value):
# Stow the original value if present, so we can track what's changed.
if key in self._data:
self._orig_data[key] = self._data[key]
else:
# Use a marker to indicate we've never seen a value for this key.
self._orig_data[key] = NEWVALUE
self._data[key] = value
self._is_dirty = True
def __delitem__(self, key):
if not key in self._data:
return
# Stow the original value, so we can track what's changed.
value = self._data[key]
del self._data[key]
self._orig_data[key] = value
self._is_dirty = True
def keys(self):
return list(self._data.keys())
def values(self):
return list(self._data.values())
def items(self):
return list(self._data.items())
def get(self, key, default=None):
return self._data.get(key, default)
def __iter__(self):
for key in self._data:
yield self._data[key]
def __contains__(self, key):
return key in self._data
def needs_save(self):
"""
Returns whether or not the data has changed on the ``Item``.
Example:
>>> user.needs_save()
False
>>> user['first_name'] = 'Johann'
>>> user.needs_save()
True
"""
return self._is_dirty
def mark_clean(self):
"""
Marks an ``Item`` instance as no longer needing to be saved.
Example:
>>> user.needs_save()
False
>>> user['first_name'] = 'Johann'
>>> user.needs_save()
True
>>> user.mark_clean()
>>> user.needs_save()
False
"""
self._orig_data = {}
self._is_dirty = False
def mark_dirty(self):
"""
Marks an ``Item`` instance as needing to be saved.
Example:
>>> user.needs_save()
False
>>> user.mark_dirty()
>>> user.needs_save()
True
"""
self._is_dirty = True
def load(self, data):
"""
This is only useful when being handed raw data from DynamoDB directly.
If you have a Python datastructure already, use the ``__init__`` or
manually set the data instead.
Largely internal, unless you know what you're doing or are trying to
mix the low-level & high-level APIs.
"""
self._data = {}
for field_name, field_value in list(data.get('Item', {}).items()):
self[field_name] = self._dynamizer.decode(field_value)
self.mark_clean()
def get_keys(self):
"""
Returns a Python-style dict of the keys/values.
Largely internal.
"""
key_fields = self.table.get_key_fields()
key_data = {}
for key in key_fields:
key_data[key] = self[key]
return key_data
def get_raw_keys(self):
"""
Returns a DynamoDB-style dict of the keys/values.
Largely internal.
"""
raw_key_data = {}
for key, value in list(self.get_keys().items()):
raw_key_data[key] = self._dynamizer.encode(value)
return raw_key_data
def build_expects(self, fields=None):
"""
Builds up a list of expecations to hand off to DynamoDB on save.
Largely internal.
"""
expects = {}
if fields is None:
fields = list(self._data.keys()) + list(self._orig_data.keys())
# Only uniques.
fields = set(fields)
for key in fields:
expects[key] = {
'Exists': True,
}
value = None
# Check for invalid keys.
if not key in self._orig_data and not key in self._data:
raise ValueError("Unknown key %s provided." % key)
# States:
# * New field (_data & _orig_data w/ marker)
# * Unchanged field (only _data)
# * Modified field (_data & _orig_data)
# * Deleted field (only _orig_data)
if not key in self._orig_data:
# Existing field unchanged.
value = self._data[key]
else:
if key in self._data:
if self._orig_data[key] is NEWVALUE:
# New field.
expects[key]['Exists'] = False
else:
# Existing field modified.
value = self._orig_data[key]
else:
# Existing field deleted.
value = self._orig_data[key]
if value is not None:
expects[key]['Value'] = self._dynamizer.encode(value)
return expects
def prepare_full(self):
"""
Runs through all fields & encodes them to be handed off to DynamoDB
as part of an ``save`` (``put_item``) call.
Largely internal.
"""
# This doesn't save on it's own. Rather, we prepare the datastructure
# and hand-off to the table to handle creation/update.
final_data = {}
for key, value in list(self._data.items()):
final_data[key] = self._dynamizer.encode(value)
return final_data
def prepare_partial(self):
"""
Runs through **ONLY** the changed/deleted fields & encodes them to be
handed off to DynamoDB as part of an ``partial_save`` (``update_item``)
call.
Largely internal.
"""
# This doesn't save on it's own. Rather, we prepare the datastructure
# and hand-off to the table to handle creation/update.
final_data = {}
# Loop over ``_orig_data`` so that we only build up data that's changed.
for key, value in list(self._orig_data.items()):
if key in self._data:
# It changed.
final_data[key] = {
'Action': 'PUT',
'Value': self._dynamizer.encode(self._data[key])
}
else:
# It was deleted.
final_data[key] = {
'Action': 'DELETE',
}
return final_data
def partial_save(self):
"""
Saves only the changed data to DynamoDB.
Extremely useful for high-volume/high-write data sets, this allows
you to update only a handful of fields rather than having to push
entire items. This prevents many accidental overwrite situations as
well as saves on the amount of data to transfer over the wire.
Returns ``True`` on success, ``False`` if no save was performed or
the write failed.
Example::
>>> user['last_name'] = 'Doh!'
# Only the last name field will be sent to DynamoDB.
>>> user.partial_save()
"""
if not self.needs_save():
return False
key = self.get_keys()
# Build a new dict of only the data we're changing.
final_data = self.prepare_partial()
# Build expectations of only the fields we're planning to update.
expects = self.build_expects(fields=list(self._orig_data.keys()))
returned = self.table._update_item(key, final_data, expects=expects)
# Mark the object as clean.
self.mark_clean()
return returned
def save(self, overwrite=False):
"""
Saves all data to DynamoDB.
By default, this attempts to ensure that none of the underlying
data has changed. If any fields have changed in between when the
``Item`` was constructed & when it is saved, this call will fail so
as not to cause any data loss.
If you're sure possibly overwriting data is acceptable, you can pass
an ``overwrite=True``. If that's not acceptable, you may be able to use
``Item.partial_save`` to only write the changed field data.
Optionally accepts an ``overwrite`` parameter, which should be a
boolean. If you provide ``True``, the item will be forcibly overwritten
within DynamoDB, even if another process changed the data in the
meantime. (Default: ``False``)
Returns ``True`` on success, ``False`` if no save was performed.
Example::
>>> user['last_name'] = 'Doh!'
# All data on the Item is sent to DynamoDB.
>>> user.save()
# If it fails, you can overwrite.
>>> user.save(overwrite=True)
"""
if not self.needs_save():
return False
final_data = self.prepare_full()
expects = None
if overwrite is False:
# Build expectations about *all* of the data.
expects = self.build_expects()
returned = self.table._put_item(final_data, expects=expects)
# Mark the object as clean.
self.mark_clean()
return returned
def delete(self):
"""
Deletes the item's data to DynamoDB.
Returns ``True`` on success.
Example::
# Buh-bye now.
>>> user.delete()
"""
key_data = self.get_keys()
return self.table.delete_item(**key_data)
def __init__(self,
table_name,
schema=None,
throughput=None,
indexes=None,
connection=None):
"""
Sets up a new in-memory ``Table``.
This is useful if the table already exists within DynamoDB & you simply
want to use it for additional interactions. The only required parameter
is the ``table_name``. However, under the hood, the object will call
``describe_table`` to determine the schema/indexes/throughput. You
can avoid this extra call by passing in ``schema`` & ``indexes``.
**IMPORTANT** - If you're creating a new ``Table`` for the first time,
you should use the ``Table.create`` method instead, as it will
persist the table structure to DynamoDB.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Optionally accepts a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
# The simple, it-already-exists case.
>>> conn = Table('users')
# The full, minimum-extra-calls case.
>>> from boto import dynamodb2
>>> users = Table('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... RangeKey('date_joined')
... ]),
... ],
... connection=dynamodb2.connect_to_region('us-west-2',
... aws_access_key_id='key',
... aws_secret_access_key='key',
... ))
"""
self.table_name = table_name
self.connection = connection
self.throughput = {
'read': 5,
'write': 5,
}
self.schema = schema
self.indexes = indexes
if self.connection is None:
self.connection = DynamoDBConnection()
if throughput is not None:
self.throughput = throughput
self._dynamizer = Dynamizer()
class Table(object):
"""
Interacts & models the behavior of a DynamoDB table.
The ``Table`` object represents a set (or rough categorization) of
records within DynamoDB. The important part is that all records within the
table, while largely-schema-free, share the same schema & are essentially
namespaced for use in your application. For example, you might have a
``users`` table or a ``forums`` table.
"""
max_batch_get = 100
def __init__(self,
table_name,
schema=None,
throughput=None,
indexes=None,
connection=None):
"""
Sets up a new in-memory ``Table``.
This is useful if the table already exists within DynamoDB & you simply
want to use it for additional interactions. The only required parameter
is the ``table_name``. However, under the hood, the object will call
``describe_table`` to determine the schema/indexes/throughput. You
can avoid this extra call by passing in ``schema`` & ``indexes``.
**IMPORTANT** - If you're creating a new ``Table`` for the first time,
you should use the ``Table.create`` method instead, as it will
persist the table structure to DynamoDB.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Optionally accepts a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
# The simple, it-already-exists case.
>>> conn = Table('users')
# The full, minimum-extra-calls case.
>>> from boto import dynamodb2
>>> users = Table('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... RangeKey('date_joined')
... ]),
... ],
... connection=dynamodb2.connect_to_region('us-west-2',
... aws_access_key_id='key',
... aws_secret_access_key='key',
... ))
"""
self.table_name = table_name
self.connection = connection
self.throughput = {
'read': 5,
'write': 5,
}
self.schema = schema
self.indexes = indexes
if self.connection is None:
self.connection = DynamoDBConnection()
if throughput is not None:
self.throughput = throughput
self._dynamizer = Dynamizer()
@classmethod
def create(cls,
table_name,
schema,
throughput=None,
indexes=None,
connection=None):
"""
Creates a new table in DynamoDB & returns an in-memory ``Table`` object.
This will setup a brand new table within DynamoDB. The ``table_name``
must be unique for your AWS account. The ``schema`` is also required
to define the key structure of the table.
**IMPORTANT** - You should consider the usage pattern of your table
up-front, as the schema & indexes can **NOT** be modified once the
table is created, requiring the creation of a new table & migrating
the data should you wish to revise it.
**IMPORTANT** - If the table already exists in DynamoDB, additional
calls to this method will result in an error. If you just need
a ``Table`` object to interact with the existing table, you should
just initialize a new ``Table`` object, which requires only the
``table_name``.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Requires a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
>>> users = Table.create('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... RangeKey('date_joined')
... ]),
... ])
"""
table = cls(table_name=table_name, connection=connection)
table.schema = schema
if throughput is not None:
table.throughput = throughput
if indexes is not None:
table.indexes = indexes
# Prep the schema.
raw_schema = []
attr_defs = []
for field in table.schema:
raw_schema.append(field.schema())
# Build the attributes off what we know.
attr_defs.append(field.definition())
raw_throughput = {
'ReadCapacityUnits': int(table.throughput['read']),
'WriteCapacityUnits': int(table.throughput['write']),
}
kwargs = {}
if table.indexes:
# Prep the LSIs.
raw_lsi = []
for index_field in table.indexes:
raw_lsi.append(index_field.schema())
# Again, build the attributes off what we know.
# HOWEVER, only add attributes *NOT* already seen.
attr_define = index_field.definition()
for part in attr_define:
attr_names = [attr['AttributeName'] for attr in attr_defs]
if not part['AttributeName'] in attr_names:
attr_defs.append(part)
kwargs['local_secondary_indexes'] = raw_lsi
table.connection.create_table(table_name=table.table_name,
attribute_definitions=attr_defs,
key_schema=raw_schema,
provisioned_throughput=raw_throughput,
**kwargs)
return table
def _introspect_schema(self, raw_schema):
"""
Given a raw schema structure back from a DynamoDB response, parse
out & build the high-level Python objects that represent them.
"""
schema = []
for field in raw_schema:
if field['KeyType'] == 'HASH':
schema.append(HashKey(field['AttributeName']))
elif field['KeyType'] == 'RANGE':
schema.append(RangeKey(field['AttributeName']))
else:
raise exceptions.UnknownSchemaFieldError(
"%s was seen, but is unknown. Please report this at "
"https://github.com/boto/boto/issues." % field['KeyType'])
return schema
def _introspect_indexes(self, raw_indexes):
"""
Given a raw index structure back from a DynamoDB response, parse
out & build the high-level Python objects that represent them.
"""
indexes = []
for field in raw_indexes:
index_klass = AllIndex
kwargs = {'parts': []}
if field['Projection']['ProjectionType'] == 'ALL':
index_klass = AllIndex
elif field['Projection']['ProjectionType'] == 'KEYS_ONLY':
index_klass = KeysOnlyIndex
elif field['Projection']['ProjectionType'] == 'INCLUDE':
index_klass = IncludeIndex
kwargs['includes'] = field['Projection']['NonKeyAttributes']
else:
raise exceptions.UnknownIndexFieldError(
"%s was seen, but is unknown. Please report this at "
"https://github.com/boto/boto/issues." % \
field['Projection']['ProjectionType']
)
name = field['IndexName']
kwargs['parts'] = self._introspect_schema(field['KeySchema'])
indexes.append(index_klass(name, **kwargs))
return indexes
def describe(self):
"""
Describes the current structure of the table in DynamoDB.
This information will be used to update the ``schema``, ``indexes``
and ``throughput`` information on the ``Table``. Some calls, such as
those involving creating keys or querying, will require this
information to be populated.
It also returns the full raw datastructure from DynamoDB, in the
event you'd like to parse out additional information (such as the
``ItemCount`` or usage information).
Example::
>>> users.describe()
{
# Lots of keys here...
}
>>> len(users.schema)
2
"""
result = self.connection.describe_table(self.table_name)
# Blindly update throughput, since what's on DynamoDB's end is likely
# more correct.
raw_throughput = result['Table']['ProvisionedThroughput']
self.throughput['read'] = int(raw_throughput['ReadCapacityUnits'])
self.throughput['write'] = int(raw_throughput['WriteCapacityUnits'])
if not self.schema:
# Since we have the data, build the schema.
raw_schema = result['Table'].get('KeySchema', [])
self.schema = self._introspect_schema(raw_schema)
if not self.indexes:
# Build the index information as well.
raw_indexes = result['Table'].get('LocalSecondaryIndexes', [])
self.indexes = self._introspect_indexes(raw_indexes)
# This is leaky.
return result
def update(self, throughput):
"""
Updates table attributes in DynamoDB.
Currently, the only thing you can modify about a table after it has
been created is the throughput.
Requires a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Returns ``True`` on success.
Example::
# For a read-heavier application...
>>> users.update(throughput={
... 'read': 20,
... 'write': 10,
... })
True
"""
self.throughput = throughput
self.connection.update_table(
self.table_name, {
'ReadCapacityUnits': int(self.throughput['read']),
'WriteCapacityUnits': int(self.throughput['write']),
})
return True
def delete(self):
"""
Deletes a table in DynamoDB.
**IMPORTANT** - Be careful when using this method, there is no undo.
Returns ``True`` on success.
Example::
>>> users.delete()
True
"""
self.connection.delete_table(self.table_name)
return True
def _encode_keys(self, keys):
"""
Given a flat Python dictionary of keys/values, converts it into the
nested dictionary DynamoDB expects.
Converts::
{
'username': '******',
'tags': [1, 2, 5],
}
...to...::
{
'username': {'S': 'john'},
'tags': {'NS': ['1', '2', '5']},
}
"""
raw_key = {}
for key, value in keys.items():
raw_key[key] = self._dynamizer.encode(value)
return raw_key
def get_item(self, consistent=False, **kwargs):
"""
Fetches an item (record) from a table in DynamoDB.
To specify the key of the item you'd like to get, you can specify the
key attributes as kwargs.
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will perform
a consistent (but more expensive) read from DynamoDB.
(Default: ``False``)
Returns an ``Item`` instance containing all the data for that record.
Example::
# A simple hash key.
>>> john = users.get_item(username='******')
>>> john['first_name']
'John'
# A complex hash+range key.
>>> john = users.get_item(username='******', last_name='Doe')
>>> john['first_name']
'John'
# A consistent read (assuming the data might have just changed).
>>> john = users.get_item(username='******', consistent=True)
>>> john['first_name']
'Johann'
# With a key that is an invalid variable name in Python.
# Also, assumes a different schema than previous examples.
>>> john = users.get_item(**{
... 'date-joined': 127549192,
... })
>>> john['first_name']
'John'
"""
raw_key = self._encode_keys(kwargs)
item_data = self.connection.get_item(self.table_name,
raw_key,
consistent_read=consistent)
item = Item(self)
item.load(item_data)
return item
def lookup(self, *args, **kwargs):
"""
Look up an entry in DynamoDB. This is mostly backwards compatible
with boto.dynamodb. Unlike get_item, it takes hash_key and range_key first,
although you may still specify keyword arguments instead.
Also unlike the get_item command, if the returned item has no keys
(i.e., it does not exist in DynamoDB), a None result is returned, instead
of an empty key object.
Example::
>>> user = users.lookup(username)
>>> user = users.lookup(username, consistent=True)
>>> app = apps.lookup('my_customer_id', 'my_app_id')
"""
if not self.schema:
self.describe()
for x, arg in enumerate(args):
kwargs[self.schema[x].name] = arg
ret = self.get_item(**kwargs)
if not ret.keys():
return None
return ret
def new_item(self, *args):
"""
Returns a new, blank item
This is mostly for consistency with boto.dynamodb
"""
if not self.schema:
self.describe()
data = {}
for x, arg in enumerate(args):
data[self.schema[x].name] = arg
return Item(self, data=data)
def put_item(self, data, overwrite=False):
"""
Saves an entire item to DynamoDB.
By default, if any part of the ``Item``'s original data doesn't match
what's currently in DynamoDB, this request will fail. This prevents
other processes from updating the data in between when you read the
item & when your request to update the item's data is processed, which
would typically result in some data loss.
Requires a ``data`` parameter, which should be a dictionary of the data
you'd like to store in DynamoDB.
Optionally accepts an ``overwrite`` parameter, which should be a
boolean. If you provide ``True``, this will tell DynamoDB to blindly
overwrite whatever data is present, if any.
Returns ``True`` on success.
Example::
>>> users.put_item(data={
... 'username': '******',
... 'first_name': 'Jane',
... 'last_name': 'Doe',
... 'date_joined': 126478915,
... })
True
"""
item = Item(self, data=data)
return item.save(overwrite=overwrite)
def _put_item(self, item_data, expects=None):
"""
The internal variant of ``put_item`` (full data). This is used by the
``Item`` objects, since that operation is represented at the
table-level by the API, but conceptually maps better to telling an
individual ``Item`` to save itself.
"""
kwargs = {}
if expects is not None:
kwargs['expected'] = expects
self.connection.put_item(self.table_name, item_data, **kwargs)
return True
def _update_item(self, key, item_data, expects=None):
"""
The internal variant of ``put_item`` (partial data). This is used by the
``Item`` objects, since that operation is represented at the
table-level by the API, but conceptually maps better to telling an
individual ``Item`` to save itself.
"""
raw_key = self._encode_keys(key)
kwargs = {}
if expects is not None:
kwargs['expected'] = expects
self.connection.update_item(self.table_name, raw_key, item_data,
**kwargs)
return True
def delete_item(self, **kwargs):
"""
Deletes an item in DynamoDB.
**IMPORTANT** - Be careful when using this method, there is no undo.
To specify the key of the item you'd like to get, you can specify the
key attributes as kwargs.
Returns ``True`` on success.
Example::
# A simple hash key.
>>> users.delete_item(username='******')
True
# A complex hash+range key.
>>> users.delete_item(username='******', last_name='Doe')
True
# With a key that is an invalid variable name in Python.
# Also, assumes a different schema than previous examples.
>>> users.delete_item(**{
... 'date-joined': 127549192,
... })
True
"""
raw_key = self._encode_keys(kwargs)
self.connection.delete_item(self.table_name, raw_key)
return True
def get_key_fields(self):
"""
Returns the fields necessary to make a key for a table.
If the ``Table`` does not already have a populated ``schema``,
this will request it via a ``Table.describe`` call.
Returns a list of fieldnames (strings).
Example::
# A simple hash key.
>>> users.get_key_fields()
['username']
# A complex hash+range key.
>>> users.get_key_fields()
['username', 'last_name']
"""
if not self.schema:
# We don't know the structure of the table. Get a description to
# populate the schema.
self.describe()
return [field.name for field in self.schema]
def batch_write(self):
"""
Allows the batching of writes to DynamoDB.
Since each write/delete call to DynamoDB has a cost associated with it,
when loading lots of data, it makes sense to batch them, creating as
few calls as possible.
This returns a context manager that will transparently handle creating
these batches. The object you get back lightly-resembles a ``Table``
object, sharing just the ``put_item`` & ``delete_item`` methods
(which are all that DynamoDB can batch in terms of writing data).
DynamoDB's maximum batch size is 25 items per request. If you attempt
to put/delete more than that, the context manager will batch as many
as it can up to that number, then flush them to DynamoDB & continue
batching as more calls come in.
Example::
# Assuming a table with one record...
>>> with users.batch_write() as batch:
... batch.put_item(data={
... 'username': '******',
... 'first_name': 'John',
... 'last_name': 'Doe',
... 'owner': 1,
... })
... # Nothing across the wire yet.
... batch.delete_item(username='******')
... # Still no requests sent.
... batch.put_item(data={
... 'username': '******',
... 'first_name': 'Jane',
... 'last_name': 'Doe',
... 'date_joined': 127436192,
... })
... # Nothing yet, but once we leave the context, the
... # put/deletes will be sent.
"""
# PHENOMENAL COSMIC DOCS!!! itty-bitty code.
return BatchTable(self)
def _build_filters(self, filter_kwargs, using=QUERY_OPERATORS):
"""
An internal method for taking query/scan-style ``**kwargs`` & turning
them into the raw structure DynamoDB expects for filtering.
"""
filters = {}
for field_and_op, value in filter_kwargs.items():
field_bits = field_and_op.split('__')
fieldname = '__'.join(field_bits[:-1])
try:
op = using[field_bits[-1]]
except KeyError:
raise exceptions.UnknownFilterTypeError(
"Operator '%s' from '%s' is not recognized." %
(field_bits[-1], field_and_op))
lookup = {
'AttributeValueList': [],
'ComparisonOperator': op,
}
# Special-case the ``NULL/NOT_NULL`` case.
if field_bits[-1] == 'null':
del lookup['AttributeValueList']
if value is False:
lookup['ComparisonOperator'] = 'NOT_NULL'
else:
lookup['ComparisonOperator'] = 'NULL'
# Special-case the ``BETWEEN`` case.
elif field_bits[-1] == 'between':
if len(value) == 2 and isinstance(value, (list, tuple)):
lookup['AttributeValueList'].append(
self._dynamizer.encode(value[0]))
lookup['AttributeValueList'].append(
self._dynamizer.encode(value[1]))
else:
# Fix up the value for encoding, because it was built to only work
# with ``set``s.
if isinstance(value, (list, tuple)):
value = set(value)
lookup['AttributeValueList'].append(
self._dynamizer.encode(value))
# Finally, insert it into the filters.
filters[fieldname] = lookup
return filters
def query(self,
limit=None,
index=None,
reverse=False,
consistent=False,
attributes=None,
**filter_kwargs):
"""
Queries for a set of matching items in a DynamoDB table.
Queries can be performed against a hash key, a hash+range key or
against any data stored in your local secondary indexes.
**Note** - You can not query against arbitrary fields within the data
stored in DynamoDB.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts a ``limit`` parameter, which should be an integer
count of the total number of items to return. (Default: ``None`` -
all results)
Optionally accepts an ``index`` parameter, which should be a string of
name of the local secondary index you want to query against.
(Default: ``None``)
Optionally accepts a ``reverse`` parameter, which will present the
results in reverse order. (Default: ``None`` - normal order)
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will force a consistent read of
the data (more expensive). (Default: ``False`` - use eventually
consistent reads)
Optionally accepts a ``attributes`` parameter, which should be a
tuple. If you provide any attributes only these will be fetched
from DynamoDB. This uses the ``AttributesToGet`` and set's
``Select`` to ``SPECIFIC_ATTRIBUTES`` API.
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
# Look for last names equal to "Doe".
>>> results = users.query(last_name__eq='Doe')
>>> for res in results:
... print res['first_name']
'John'
'Jane'
# Look for last names beginning with "D", in reverse order, limit 3.
>>> results = users.query(
... last_name__beginswith='D',
... reverse=True,
... limit=3
... )
>>> for res in results:
... print res['first_name']
'Alice'
'Jane'
'John'
# Use an LSI & a consistent read.
>>> results = users.query(
... date_joined__gte=1236451000,
... owner__eq=1,
... index='DateJoinedIndex',
... consistent=True
... )
>>> for res in results:
... print res['first_name']
'Alice'
'Bob'
'John'
'Fred'
"""
if self.schema:
if len(self.schema) == 1 and len(filter_kwargs) <= 1:
raise exceptions.QueryError(
"You must specify more than one key to filter on.")
if attributes is not None:
select = 'SPECIFIC_ATTRIBUTES'
else:
select = None
results = ResultSet()
kwargs = filter_kwargs.copy()
kwargs.update({
'limit': limit,
'index': index,
'reverse': reverse,
'consistent': consistent,
'select': select,
'attributes_to_get': attributes
})
results.to_call(self._query, **kwargs)
return results
def query_count(self, index=None, consistent=False, **filter_kwargs):
"""
Queries the exact count of matching items in a DynamoDB table.
Queries can be performed against a hash key, a hash+range key or
against any data stored in your local secondary indexes.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts an ``index`` parameter, which should be a string of
name of the local secondary index you want to query against.
(Default: ``None``)
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will force a consistent read of
the data (more expensive). (Default: ``False`` - use eventually
consistent reads)
Returns an integer which represents the exact amount of matched
items.
Example::
# Look for last names equal to "Doe".
>>> users.query_count(last_name__eq='Doe')
5
# Use an LSI & a consistent read.
>>> users.query_count(
... date_joined__gte=1236451000,
... owner__eq=1,
... index='DateJoinedIndex',
... consistent=True
... )
2
"""
key_conditions = self._build_filters(filter_kwargs,
using=QUERY_OPERATORS)
raw_results = self.connection.query(
self.table_name,
index_name=index,
consistent_read=consistent,
select='COUNT',
key_conditions=key_conditions,
)
return int(raw_results.get('Count', 0))
def _query(self,
limit=None,
index=None,
reverse=False,
consistent=False,
exclusive_start_key=None,
select=None,
attributes_to_get=None,
**filter_kwargs):
"""
The internal method that performs the actual queries. Used extensively
by ``ResultSet`` to perform each (paginated) request.
"""
kwargs = {
'limit': limit,
'index_name': index,
'scan_index_forward': reverse,
'consistent_read': consistent,
'select': select,
'attributes_to_get': attributes_to_get
}
if exclusive_start_key:
kwargs['exclusive_start_key'] = {}
for key, value in exclusive_start_key.items():
kwargs['exclusive_start_key'][key] = \
self._dynamizer.encode(value)
# Convert the filters into something we can actually use.
kwargs['key_conditions'] = self._build_filters(filter_kwargs,
using=QUERY_OPERATORS)
raw_results = self.connection.query(self.table_name, **kwargs)
results = []
last_key = None
for raw_item in raw_results.get('Items', []):
item = Item(self)
item.load({
'Item': raw_item,
})
results.append(item)
if raw_results.get('LastEvaluatedKey', None):
last_key = {}
for key, value in raw_results['LastEvaluatedKey'].items():
last_key[key] = self._dynamizer.decode(value)
return {
'results': results,
'last_key': last_key,
}
def scan(self,
limit=None,
segment=None,
total_segments=None,
**filter_kwargs):
"""
Scans across all items within a DynamoDB table.
Scans can be performed against a hash key or a hash+range key. You can
additionally filter the results after the table has been read but
before the response is returned.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts a ``limit`` parameter, which should be an integer
count of the total number of items to return. (Default: ``None`` -
all results)
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
# All results.
>>> everything = users.scan()
# Look for last names beginning with "D".
>>> results = users.scan(last_name__beginswith='D')
>>> for res in results:
... print res['first_name']
'Alice'
'John'
'Jane'
# Use an ``IN`` filter & limit.
>>> results = users.scan(
... age__in=[25, 26, 27, 28, 29],
... limit=1
... )
>>> for res in results:
... print res['first_name']
'Alice'
"""
results = ResultSet()
kwargs = filter_kwargs.copy()
kwargs.update({
'limit': limit,
'segment': segment,
'total_segments': total_segments,
})
results.to_call(self._scan, **kwargs)
return results
def _scan(self,
limit=None,
exclusive_start_key=None,
segment=None,
total_segments=None,
**filter_kwargs):
"""
The internal method that performs the actual scan. Used extensively
by ``ResultSet`` to perform each (paginated) request.
"""
kwargs = {
'limit': limit,
'segment': segment,
'total_segments': total_segments,
}
if exclusive_start_key:
kwargs['exclusive_start_key'] = {}
for key, value in exclusive_start_key.items():
kwargs['exclusive_start_key'][key] = \
self._dynamizer.encode(value)
# Convert the filters into something we can actually use.
kwargs['scan_filter'] = self._build_filters(filter_kwargs,
using=FILTER_OPERATORS)
raw_results = self.connection.scan(self.table_name, **kwargs)
results = []
last_key = None
for raw_item in raw_results.get('Items', []):
item = Item(self)
item.load({
'Item': raw_item,
})
results.append(item)
if raw_results.get('LastEvaluatedKey', None):
last_key = {}
for key, value in raw_results['LastEvaluatedKey'].items():
last_key[key] = self._dynamizer.decode(value)
return {
'results': results,
'last_key': last_key,
}
def batch_get(self, keys, consistent=False):
"""
Fetches many specific items in batch from a table.
Requires a ``keys`` parameter, which should be a list of dictionaries.
Each dictionary should consist of the keys values to specify.
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, a strongly consistent read will be
used. (Default: False)
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
>>> results = users.batch_get(keys=[
... {
... 'username': '******',
... },
... {
... 'username': '******',
... },
... {
... 'username': '******',
... },
... ])
>>> for res in results:
... print res['first_name']
'John'
'Jane'
'Fred'
"""
# We pass the keys to the constructor instead, so it can maintain it's
# own internal state as to what keys have been processed.
results = BatchGetResultSet(keys=keys,
max_batch_get=self.max_batch_get)
results.to_call(self._batch_get, consistent=False)
return results
def _batch_get(self, keys, consistent=False):
"""
The internal method that performs the actual batch get. Used extensively
by ``BatchGetResultSet`` to perform each (paginated) request.
"""
items = {
self.table_name: {
'Keys': [],
},
}
if consistent:
items[self.table_name]['ConsistentRead'] = True
for key_data in keys:
raw_key = {}
for key, value in key_data.items():
raw_key[key] = self._dynamizer.encode(value)
items[self.table_name]['Keys'].append(raw_key)
raw_results = self.connection.batch_get_item(request_items=items)
results = []
unprocessed_keys = []
for raw_item in raw_results['Responses'].get(self.table_name, []):
item = Item(self)
item.load({
'Item': raw_item,
})
results.append(item)
raw_unproccessed = raw_results.get('UnprocessedKeys', {})
for raw_key in raw_unproccessed.get('Keys', []):
py_key = {}
for key, value in raw_key.items():
py_key[key] = self._dynamizer.decode(value)
unprocessed_keys.append(py_key)
return {
'results': results,
# NEVER return a ``last_key``. Just in-case any part of
# ``ResultSet`` peeks through, since much of the
# original underlying implementation is based on this key.
'last_key': None,
'unprocessed_keys': unprocessed_keys,
}
def count(self):
"""
Returns a (very) eventually consistent count of the number of items
in a table.
Lag time is about 6 hours, so don't expect a high degree of accuracy.
Example::
>>> users.count()
6
"""
info = self.describe()
return info['Table'].get('ItemCount', 0)
class Table(object):
"""
Interacts & models the behavior of a DynamoDB table.
The ``Table`` object represents a set (or rough categorization) of
records within DynamoDB. The important part is that all records within the
table, while largely-schema-free, share the same schema & are essentially
namespaced for use in your application. For example, you might have a
``users`` table or a ``forums`` table.
"""
max_batch_get = 100
def __init__(self, table_name, schema=None, throughput=None, indexes=None, connection=None):
"""
Sets up a new in-memory ``Table``.
This is useful if the table already exists within DynamoDB & you simply
want to use it for additional interactions. The only required parameter
is the ``table_name``. However, under the hood, the object will call
``describe_table`` to determine the schema/indexes/throughput. You
can avoid this extra call by passing in ``schema`` & ``indexes``.
**IMPORTANT** - If you're creating a new ``Table`` for the first time,
you should use the ``Table.create`` method instead, as it will
persist the table structure to DynamoDB.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Optionally accepts a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
# The simple, it-already-exists case.
>>> conn = Table('users')
# The full, minimum-extra-calls case.
>>> from boto import dynamodb2
>>> users = Table('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... RangeKey('date_joined')
... ]),
... ],
... connection=dynamodb2.connect_to_region('us-west-2',
... aws_access_key_id='key',
... aws_secret_access_key='key',
... ))
"""
self.table_name = table_name
self.connection = connection
self.throughput = {"read": 5, "write": 5}
self.schema = schema
self.indexes = indexes
if self.connection is None:
self.connection = DynamoDBConnection()
if throughput is not None:
self.throughput = throughput
self._dynamizer = Dynamizer()
@classmethod
def create(cls, table_name, schema, throughput=None, indexes=None, connection=None):
"""
Creates a new table in DynamoDB & returns an in-memory ``Table`` object.
This will setup a brand new table within DynamoDB. The ``table_name``
must be unique for your AWS account. The ``schema`` is also required
to define the key structure of the table.
**IMPORTANT** - You should consider the usage pattern of your table
up-front, as the schema & indexes can **NOT** be modified once the
table is created, requiring the creation of a new table & migrating
the data should you wish to revise it.
**IMPORTANT** - If the table already exists in DynamoDB, additional
calls to this method will result in an error. If you just need
a ``Table`` object to interact with the existing table, you should
just initialize a new ``Table`` object, which requires only the
``table_name``.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Requires a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
>>> users = Table.create('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... RangeKey('date_joined')
... ]),
... ])
"""
table = cls(table_name=table_name, connection=connection)
table.schema = schema
if throughput is not None:
table.throughput = throughput
if indexes is not None:
table.indexes = indexes
# Prep the schema.
raw_schema = []
attr_defs = []
for field in table.schema:
raw_schema.append(field.schema())
# Build the attributes off what we know.
attr_defs.append(field.definition())
raw_throughput = {
"ReadCapacityUnits": int(table.throughput["read"]),
"WriteCapacityUnits": int(table.throughput["write"]),
}
kwargs = {}
if table.indexes:
# Prep the LSIs.
raw_lsi = []
for index_field in table.indexes:
raw_lsi.append(index_field.schema())
# Again, build the attributes off what we know.
# HOWEVER, only add attributes *NOT* already seen.
attr_define = index_field.definition()
for part in attr_define:
attr_names = [attr["AttributeName"] for attr in attr_defs]
if not part["AttributeName"] in attr_names:
attr_defs.append(part)
kwargs["local_secondary_indexes"] = raw_lsi
table.connection.create_table(
table_name=table.table_name,
attribute_definitions=attr_defs,
key_schema=raw_schema,
provisioned_throughput=raw_throughput,
**kwargs
)
return table
def _introspect_schema(self, raw_schema):
"""
Given a raw schema structure back from a DynamoDB response, parse
out & build the high-level Python objects that represent them.
"""
schema = []
for field in raw_schema:
if field["KeyType"] == "HASH":
schema.append(HashKey(field["AttributeName"]))
elif field["KeyType"] == "RANGE":
schema.append(RangeKey(field["AttributeName"]))
else:
raise exceptions.UnknownSchemaFieldError(
"%s was seen, but is unknown. Please report this at "
"https://github.com/boto/boto/issues." % field["KeyType"]
)
return schema
def _introspect_indexes(self, raw_indexes):
"""
Given a raw index structure back from a DynamoDB response, parse
out & build the high-level Python objects that represent them.
"""
indexes = []
for field in raw_indexes:
index_klass = AllIndex
kwargs = {"parts": []}
if field["Projection"]["ProjectionType"] == "ALL":
index_klass = AllIndex
elif field["Projection"]["ProjectionType"] == "KEYS_ONLY":
index_klass = KeysOnlyIndex
elif field["Projection"]["ProjectionType"] == "INCLUDE":
index_klass = IncludeIndex
kwargs["includes"] = field["Projection"]["NonKeyAttributes"]
else:
raise exceptions.UnknownIndexFieldError(
"%s was seen, but is unknown. Please report this at "
"https://github.com/boto/boto/issues." % field["Projection"]["ProjectionType"]
)
name = field["IndexName"]
kwargs["parts"] = self._introspect_schema(field["KeySchema"])
indexes.append(index_klass(name, **kwargs))
return indexes
def describe(self):
"""
Describes the current structure of the table in DynamoDB.
This information will be used to update the ``schema``, ``indexes``
and ``throughput`` information on the ``Table``. Some calls, such as
those involving creating keys or querying, will require this
information to be populated.
It also returns the full raw datastructure from DynamoDB, in the
event you'd like to parse out additional information (such as the
``ItemCount`` or usage information).
Example::
>>> users.describe()
{
# Lots of keys here...
}
>>> len(users.schema)
2
"""
result = self.connection.describe_table(self.table_name)
# Blindly update throughput, since what's on DynamoDB's end is likely
# more correct.
raw_throughput = result["Table"]["ProvisionedThroughput"]
self.throughput["read"] = int(raw_throughput["ReadCapacityUnits"])
self.throughput["write"] = int(raw_throughput["WriteCapacityUnits"])
if not self.schema:
# Since we have the data, build the schema.
raw_schema = result["Table"].get("KeySchema", [])
self.schema = self._introspect_schema(raw_schema)
if not self.indexes:
# Build the index information as well.
raw_indexes = result["Table"].get("LocalSecondaryIndexes", [])
self.indexes = self._introspect_indexes(raw_indexes)
# This is leaky.
return result
def update(self, throughput):
"""
Updates table attributes in DynamoDB.
Currently, the only thing you can modify about a table after it has
been created is the throughput.
Requires a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Returns ``True`` on success.
Example::
# For a read-heavier application...
>>> users.update(throughput={
... 'read': 20,
... 'write': 10,
... })
True
"""
self.throughput = throughput
self.connection.update_table(
self.table_name,
{"ReadCapacityUnits": int(self.throughput["read"]), "WriteCapacityUnits": int(self.throughput["write"])},
)
return True
def delete(self):
"""
Deletes a table in DynamoDB.
**IMPORTANT** - Be careful when using this method, there is no undo.
Returns ``True`` on success.
Example::
>>> users.delete()
True
"""
self.connection.delete_table(self.table_name)
return True
def _encode_keys(self, keys):
"""
Given a flat Python dictionary of keys/values, converts it into the
nested dictionary DynamoDB expects.
Converts::
{
'username': '******',
'tags': [1, 2, 5],
}
...to...::
{
'username': {'S': 'john'},
'tags': {'NS': ['1', '2', '5']},
}
"""
raw_key = {}
for key, value in keys.items():
raw_key[key] = self._dynamizer.encode(value)
return raw_key
def get_item(self, consistent=False, **kwargs):
"""
Fetches an item (record) from a table in DynamoDB.
To specify the key of the item you'd like to get, you can specify the
key attributes as kwargs.
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will perform
a consistent (but more expensive) read from DynamoDB.
(Default: ``False``)
Returns an ``Item`` instance containing all the data for that record.
Example::
# A simple hash key.
>>> john = users.get_item(username='******')
>>> john['first_name']
'John'
# A complex hash+range key.
>>> john = users.get_item(username='******', last_name='Doe')
>>> john['first_name']
'John'
# A consistent read (assuming the data might have just changed).
>>> john = users.get_item(username='******', consistent=True)
>>> john['first_name']
'Johann'
# With a key that is an invalid variable name in Python.
# Also, assumes a different schema than previous examples.
>>> john = users.get_item(**{
... 'date-joined': 127549192,
... })
>>> john['first_name']
'John'
"""
raw_key = self._encode_keys(kwargs)
item_data = self.connection.get_item(self.table_name, raw_key, consistent_read=consistent)
item = Item(self)
item.load(item_data)
return item
def put_item(self, data, overwrite=False):
"""
Saves an entire item to DynamoDB.
By default, if any part of the ``Item``'s original data doesn't match
what's currently in DynamoDB, this request will fail. This prevents
other processes from updating the data in between when you read the
item & when your request to update the item's data is processed, which
would typically result in some data loss.
Requires a ``data`` parameter, which should be a dictionary of the data
you'd like to store in DynamoDB.
Optionally accepts an ``overwrite`` parameter, which should be a
boolean. If you provide ``True``, this will tell DynamoDB to blindly
overwrite whatever data is present, if any.
Returns ``True`` on success.
Example::
>>> users.put_item(data={
... 'username': '******',
... 'first_name': 'Jane',
... 'last_name': 'Doe',
... 'date_joined': 126478915,
... })
True
"""
item = Item(self, data=data)
return item.save(overwrite=overwrite)
def _put_item(self, item_data, expects=None):
"""
The internal variant of ``put_item`` (full data). This is used by the
``Item`` objects, since that operation is represented at the
table-level by the API, but conceptually maps better to telling an
individual ``Item`` to save itself.
"""
kwargs = {}
if expects is not None:
kwargs["expected"] = expects
self.connection.put_item(self.table_name, item_data, **kwargs)
return True
def _update_item(self, key, item_data, expects=None):
"""
The internal variant of ``put_item`` (partial data). This is used by the
``Item`` objects, since that operation is represented at the
table-level by the API, but conceptually maps better to telling an
individual ``Item`` to save itself.
"""
raw_key = self._encode_keys(key)
kwargs = {}
if expects is not None:
kwargs["expected"] = expects
self.connection.update_item(self.table_name, raw_key, item_data, **kwargs)
return True
def delete_item(self, **kwargs):
"""
Deletes an item in DynamoDB.
**IMPORTANT** - Be careful when using this method, there is no undo.
To specify the key of the item you'd like to get, you can specify the
key attributes as kwargs.
Returns ``True`` on success.
Example::
# A simple hash key.
>>> users.delete_item(username='******')
True
# A complex hash+range key.
>>> users.delete_item(username='******', last_name='Doe')
True
# With a key that is an invalid variable name in Python.
# Also, assumes a different schema than previous examples.
>>> users.delete_item(**{
... 'date-joined': 127549192,
... })
True
"""
raw_key = self._encode_keys(kwargs)
self.connection.delete_item(self.table_name, raw_key)
return True
def get_key_fields(self):
"""
Returns the fields necessary to make a key for a table.
If the ``Table`` does not already have a populated ``schema``,
this will request it via a ``Table.describe`` call.
Returns a list of fieldnames (strings).
Example::
# A simple hash key.
>>> users.get_key_fields()
['username']
# A complex hash+range key.
>>> users.get_key_fields()
['username', 'last_name']
"""
if not self.schema:
# We don't know the structure of the table. Get a description to
# populate the schema.
self.describe()
return [field.name for field in self.schema]
def batch_write(self):
"""
Allows the batching of writes to DynamoDB.
Since each write/delete call to DynamoDB has a cost associated with it,
when loading lots of data, it makes sense to batch them, creating as
few calls as possible.
This returns a context manager that will transparently handle creating
these batches. The object you get back lightly-resembles a ``Table``
object, sharing just the ``put_item`` & ``delete_item`` methods
(which are all that DynamoDB can batch in terms of writing data).
DynamoDB's maximum batch size is 25 items per request. If you attempt
to put/delete more than that, the context manager will batch as many
as it can up to that number, then flush them to DynamoDB & continue
batching as more calls come in.
Example::
# Assuming a table with one record...
>>> with users.batch_write() as batch:
... batch.put_item(data={
... 'username': '******',
... 'first_name': 'John',
... 'last_name': 'Doe',
... 'owner': 1,
... })
... # Nothing across the wire yet.
... batch.delete_item(username='******')
... # Still no requests sent.
... batch.put_item(data={
... 'username': '******',
... 'first_name': 'Jane',
... 'last_name': 'Doe',
... 'date_joined': 127436192,
... })
... # Nothing yet, but once we leave the context, the
... # put/deletes will be sent.
"""
# PHENOMENAL COSMIC DOCS!!! itty-bitty code.
return BatchTable(self)
def _build_filters(self, filter_kwargs, using=QUERY_OPERATORS):
"""
An internal method for taking query/scan-style ``**kwargs`` & turning
them into the raw structure DynamoDB expects for filtering.
"""
filters = {}
for field_and_op, value in filter_kwargs.items():
field_bits = field_and_op.split("__")
fieldname = "__".join(field_bits[:-1])
try:
op = using[field_bits[-1]]
except KeyError:
raise exceptions.UnknownFilterTypeError(
"Operator '%s' from '%s' is not recognized." % (field_bits[-1], field_and_op)
)
lookup = {"AttributeValueList": [], "ComparisonOperator": op}
# Special-case the ``NULL/NOT_NULL`` case.
if field_bits[-1] == "null":
del lookup["AttributeValueList"]
if value is False:
lookup["ComparisonOperator"] = "NOT_NULL"
else:
lookup["ComparisonOperator"] = "NULL"
# Special-case the ``BETWEEN`` case.
elif field_bits[-1] == "between":
if len(value) == 2 and isinstance(value, (list, tuple)):
lookup["AttributeValueList"].append(self._dynamizer.encode(value[0]))
lookup["AttributeValueList"].append(self._dynamizer.encode(value[1]))
else:
# Fix up the value for encoding, because it was built to only work
# with ``set``s.
if isinstance(value, (list, tuple)):
value = set(value)
lookup["AttributeValueList"].append(self._dynamizer.encode(value))
# Finally, insert it into the filters.
filters[fieldname] = lookup
return filters
def query(self, limit=None, index=None, reverse=False, consistent=False, attributes=None, **filter_kwargs):
"""
Queries for a set of matching items in a DynamoDB table.
Queries can be performed against a hash key, a hash+range key or
against any data stored in your local secondary indexes.
**Note** - You can not query against arbitrary fields within the data
stored in DynamoDB.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts a ``limit`` parameter, which should be an integer
count of the total number of items to return. (Default: ``None`` -
all results)
Optionally accepts an ``index`` parameter, which should be a string of
name of the local secondary index you want to query against.
(Default: ``None``)
Optionally accepts a ``reverse`` parameter, which will present the
results in reverse order. (Default: ``None`` - normal order)
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will force a consistent read of
the data (more expensive). (Default: ``False`` - use eventually
consistent reads)
Optionally accepts a ``attributes`` parameter, which should be a
tuple. If you provide any attributes only these will be fetched
from DynamoDB. This uses the ``AttributesToGet`` and set's
``Select`` to ``SPECIFIC_ATTRIBUTES`` API.
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
# Look for last names equal to "Doe".
>>> results = users.query(last_name__eq='Doe')
>>> for res in results:
... print res['first_name']
'John'
'Jane'
# Look for last names beginning with "D", in reverse order, limit 3.
>>> results = users.query(
... last_name__beginswith='D',
... reverse=True,
... limit=3
... )
>>> for res in results:
... print res['first_name']
'Alice'
'Jane'
'John'
# Use an LSI & a consistent read.
>>> results = users.query(
... date_joined__gte=1236451000,
... owner__eq=1,
... index='DateJoinedIndex',
... consistent=True
... )
>>> for res in results:
... print res['first_name']
'Alice'
'Bob'
'John'
'Fred'
"""
if self.schema:
if len(self.schema) == 1 and len(filter_kwargs) <= 1:
raise exceptions.QueryError("You must specify more than one key to filter on.")
if attributes is not None:
select = "SPECIFIC_ATTRIBUTES"
else:
select = None
results = ResultSet()
kwargs = filter_kwargs.copy()
kwargs.update(
{
"limit": limit,
"index": index,
"reverse": reverse,
"consistent": consistent,
"select": select,
"attributes_to_get": attributes,
}
)
results.to_call(self._query, **kwargs)
return results
def query_count(self, index=None, consistent=False, **filter_kwargs):
"""
Queries the exact count of matching items in a DynamoDB table.
Queries can be performed against a hash key, a hash+range key or
against any data stored in your local secondary indexes.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts an ``index`` parameter, which should be a string of
name of the local secondary index you want to query against.
(Default: ``None``)
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, it will force a consistent read of
the data (more expensive). (Default: ``False`` - use eventually
consistent reads)
Returns an integer which represents the exact amount of matched
items.
Example::
# Look for last names equal to "Doe".
>>> users.query_count(last_name__eq='Doe')
5
# Use an LSI & a consistent read.
>>> users.query_count(
... date_joined__gte=1236451000,
... owner__eq=1,
... index='DateJoinedIndex',
... consistent=True
... )
2
"""
key_conditions = self._build_filters(filter_kwargs, using=QUERY_OPERATORS)
raw_results = self.connection.query(
self.table_name, index_name=index, consistent_read=consistent, select="COUNT", key_conditions=key_conditions
)
return int(raw_results.get("Count", 0))
def _query(
self,
limit=None,
index=None,
reverse=False,
consistent=False,
exclusive_start_key=None,
select=None,
attributes_to_get=None,
**filter_kwargs
):
"""
The internal method that performs the actual queries. Used extensively
by ``ResultSet`` to perform each (paginated) request.
"""
kwargs = {
"limit": limit,
"index_name": index,
"scan_index_forward": reverse,
"consistent_read": consistent,
"select": select,
"attributes_to_get": attributes_to_get,
}
if exclusive_start_key:
kwargs["exclusive_start_key"] = {}
for key, value in exclusive_start_key.items():
kwargs["exclusive_start_key"][key] = self._dynamizer.encode(value)
# Convert the filters into something we can actually use.
kwargs["key_conditions"] = self._build_filters(filter_kwargs, using=QUERY_OPERATORS)
raw_results = self.connection.query(self.table_name, **kwargs)
results = []
last_key = None
for raw_item in raw_results.get("Items", []):
item = Item(self)
item.load({"Item": raw_item})
results.append(item)
if raw_results.get("LastEvaluatedKey", None):
last_key = {}
for key, value in raw_results["LastEvaluatedKey"].items():
last_key[key] = self._dynamizer.decode(value)
return {"results": results, "last_key": last_key}
def scan(self, limit=None, segment=None, total_segments=None, **filter_kwargs):
"""
Scans across all items within a DynamoDB table.
Scans can be performed against a hash key or a hash+range key. You can
additionally filter the results after the table has been read but
before the response is returned.
To specify the filters of the items you'd like to get, you can specify
the filters as kwargs. Each filter kwarg should follow the pattern
``<fieldname>__<filter_operation>=<value_to_look_for>``.
Optionally accepts a ``limit`` parameter, which should be an integer
count of the total number of items to return. (Default: ``None`` -
all results)
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
# All results.
>>> everything = users.scan()
# Look for last names beginning with "D".
>>> results = users.scan(last_name__beginswith='D')
>>> for res in results:
... print res['first_name']
'Alice'
'John'
'Jane'
# Use an ``IN`` filter & limit.
>>> results = users.scan(
... age__in=[25, 26, 27, 28, 29],
... limit=1
... )
>>> for res in results:
... print res['first_name']
'Alice'
"""
results = ResultSet()
kwargs = filter_kwargs.copy()
kwargs.update({"limit": limit, "segment": segment, "total_segments": total_segments})
results.to_call(self._scan, **kwargs)
return results
def _scan(self, limit=None, exclusive_start_key=None, segment=None, total_segments=None, **filter_kwargs):
"""
The internal method that performs the actual scan. Used extensively
by ``ResultSet`` to perform each (paginated) request.
"""
kwargs = {"limit": limit, "segment": segment, "total_segments": total_segments}
if exclusive_start_key:
kwargs["exclusive_start_key"] = {}
for key, value in exclusive_start_key.items():
kwargs["exclusive_start_key"][key] = self._dynamizer.encode(value)
# Convert the filters into something we can actually use.
kwargs["scan_filter"] = self._build_filters(filter_kwargs, using=FILTER_OPERATORS)
raw_results = self.connection.scan(self.table_name, **kwargs)
results = []
last_key = None
for raw_item in raw_results.get("Items", []):
item = Item(self)
item.load({"Item": raw_item})
results.append(item)
if raw_results.get("LastEvaluatedKey", None):
last_key = {}
for key, value in raw_results["LastEvaluatedKey"].items():
last_key[key] = self._dynamizer.decode(value)
return {"results": results, "last_key": last_key}
def batch_get(self, keys, consistent=False):
"""
Fetches many specific items in batch from a table.
Requires a ``keys`` parameter, which should be a list of dictionaries.
Each dictionary should consist of the keys values to specify.
Optionally accepts a ``consistent`` parameter, which should be a
boolean. If you provide ``True``, a strongly consistent read will be
used. (Default: False)
Returns a ``ResultSet``, which transparently handles the pagination of
results you get back.
Example::
>>> results = users.batch_get(keys=[
... {
... 'username': '******',
... },
... {
... 'username': '******',
... },
... {
... 'username': '******',
... },
... ])
>>> for res in results:
... print res['first_name']
'John'
'Jane'
'Fred'
"""
# We pass the keys to the constructor instead, so it can maintain it's
# own internal state as to what keys have been processed.
results = BatchGetResultSet(keys=keys, max_batch_get=self.max_batch_get)
results.to_call(self._batch_get, consistent=False)
return results
def _batch_get(self, keys, consistent=False):
"""
The internal method that performs the actual batch get. Used extensively
by ``BatchGetResultSet`` to perform each (paginated) request.
"""
items = {self.table_name: {"Keys": []}}
if consistent:
items[self.table_name]["ConsistentRead"] = True
for key_data in keys:
raw_key = {}
for key, value in key_data.items():
raw_key[key] = self._dynamizer.encode(value)
items[self.table_name]["Keys"].append(raw_key)
raw_results = self.connection.batch_get_item(request_items=items)
results = []
unprocessed_keys = []
for raw_item in raw_results["Responses"].get(self.table_name, []):
item = Item(self)
item.load({"Item": raw_item})
results.append(item)
raw_unproccessed = raw_results.get("UnprocessedKeys", {})
for raw_key in raw_unproccessed.get("Keys", []):
py_key = {}
for key, value in raw_key.items():
py_key[key] = self._dynamizer.decode(value)
unprocessed_keys.append(py_key)
return {
"results": results,
# NEVER return a ``last_key``. Just in-case any part of
# ``ResultSet`` peeks through, since much of the
# original underlying implementation is based on this key.
"last_key": None,
"unprocessed_keys": unprocessed_keys,
}
def count(self):
"""
Returns a (very) eventually consistent count of the number of items
in a table.
Lag time is about 6 hours, so don't expect a high degree of accuracy.
Example::
>>> users.count()
6
"""
info = self.describe()
return info["Table"].get("ItemCount", 0)
def __init__(self, table_name, schema=None, throughput=None, indexes=None, connection=None):
"""
Sets up a new in-memory ``Table``.
This is useful if the table already exists within DynamoDB & you simply
want to use it for additional interactions. The only required parameter
is the ``table_name``. However, under the hood, the object will call
``describe_table`` to determine the schema/indexes/throughput. You
can avoid this extra call by passing in ``schema`` & ``indexes``.
**IMPORTANT** - If you're creating a new ``Table`` for the first time,
you should use the ``Table.create`` method instead, as it will
persist the table structure to DynamoDB.
Requires a ``table_name`` parameter, which should be a simple string
of the name of the table.
Optionally accepts a ``schema`` parameter, which should be a list of
``BaseSchemaField`` subclasses representing the desired schema.
Optionally accepts a ``throughput`` parameter, which should be a
dictionary. If provided, it should specify a ``read`` & ``write`` key,
both of which should have an integer value associated with them.
Optionally accepts a ``indexes`` parameter, which should be a list of
``BaseIndexField`` subclasses representing the desired indexes.
Optionally accepts a ``connection`` parameter, which should be a
``DynamoDBConnection`` instance (or subclass). This is primarily useful
for specifying alternate connection parameters.
Example::
# The simple, it-already-exists case.
>>> conn = Table('users')
# The full, minimum-extra-calls case.
>>> from boto import dynamodb2
>>> users = Table('users', schema=[
... HashKey('username'),
... RangeKey('date_joined', data_type=NUMBER)
... ], throughput={
... 'read':20,
... 'write': 10,
... }, indexes=[
... KeysOnlyIndex('MostRecentlyJoined', parts=[
... RangeKey('date_joined')
... ]),
... ],
... connection=dynamodb2.connect_to_region('us-west-2',
... aws_access_key_id='key',
... aws_secret_access_key='key',
... ))
"""
self.table_name = table_name
self.connection = connection
self.throughput = {"read": 5, "write": 5}
self.schema = schema
self.indexes = indexes
if self.connection is None:
self.connection = DynamoDBConnection()
if throughput is not None:
self.throughput = throughput
self._dynamizer = Dynamizer()
class Item(object):
"""
An object representing the item data within a DynamoDB table.
An item is largely schema-free, meaning it can contain any data. The only
limitation is that it must have data for the fields in the ``Table``'s
schema.
This object presents a dictionary-like interface for accessing/storing
data. It also tries to intelligently track how data has changed throughout
the life of the instance, to be as efficient as possible about updates.
"""
def __init__(self, table, data=None):
"""
Constructs an (unsaved) ``Item`` instance.
To persist the data in DynamoDB, you'll need to call the ``Item.save``
(or ``Item.partial_save``) on the instance.
Requires a ``table`` parameter, which should be a ``Table`` instance.
This is required, as DynamoDB's API is focus around all operations
being table-level. It's also for persisting schema around many objects.
Optionally accepts a ``data`` parameter, which should be a dictionary
of the fields & values of the item.
Example::
>>> users = Table('users')
>>> user = Item(users, data={
... 'username': '******',
... 'first_name': 'John',
... 'date_joined': 1248o61592,
... })
# Change existing data.
>>> user['first_name'] = 'Johann'
# Add more data.
>>> user['last_name'] = 'Doe'
# Delete data.
>>> del user['date_joined']
# Iterate over all the data.
>>> for field, val in user.items():
... print "%s: %s" % (field, val)
username: johndoe
first_name: John
date_joined: 1248o61592
"""
self.table = table
self._data = {}
self._orig_data = {}
self._is_dirty = False
self._dynamizer = Dynamizer()
if data:
self._data = data
self._is_dirty = True
for key in data.keys():
self._orig_data[key] = NEWVALUE
def __getitem__(self, key):
return self._data.get(key, None)
def __setitem__(self, key, value):
# Stow the original value if present, so we can track what's changed.
if key in self._data:
self._orig_data[key] = self._data[key]
else:
# Use a marker to indicate we've never seen a value for this key.
self._orig_data[key] = NEWVALUE
self._data[key] = value
self._is_dirty = True
def __delitem__(self, key):
if not key in self._data:
return
# Stow the original value, so we can track what's changed.
value = self._data[key]
del self._data[key]
self._orig_data[key] = value
self._is_dirty = True
def keys(self):
return self._data.keys()
def values(self):
return self._data.values()
def items(self):
return self._data.items()
def get(self, key, default=None):
return self._data.get(key, default)
def __iter__(self):
for key in self._data:
yield self._data[key]
def __contains__(self, key):
return key in self._data
def needs_save(self):
"""
Returns whether or not the data has changed on the ``Item``.
Example:
>>> user.needs_save()
False
>>> user['first_name'] = 'Johann'
>>> user.needs_save()
True
"""
return self._is_dirty
def mark_clean(self):
"""
Marks an ``Item`` instance as no longer needing to be saved.
Example:
>>> user.needs_save()
False
>>> user['first_name'] = 'Johann'
>>> user.needs_save()
True
>>> user.mark_clean()
>>> user.needs_save()
False
"""
self._orig_data = {}
self._is_dirty = False
def mark_dirty(self):
"""
Marks an ``Item`` instance as needing to be saved.
Example:
>>> user.needs_save()
False
>>> user.mark_dirty()
>>> user.needs_save()
True
"""
self._is_dirty = True
def load(self, data):
"""
This is only useful when being handed raw data from DynamoDB directly.
If you have a Python datastructure already, use the ``__init__`` or
manually set the data instead.
Largely internal, unless you know what you're doing or are trying to
mix the low-level & high-level APIs.
"""
self._data = {}
for field_name, field_value in data.get('Item', {}).items():
self[field_name] = self._dynamizer.decode(field_value)
self.mark_clean()
def get_keys(self):
"""
Returns a Python-style dict of the keys/values.
Largely internal.
"""
key_fields = self.table.get_key_fields()
key_data = {}
for key in key_fields:
key_data[key] = self[key]
return key_data
def get_raw_keys(self):
"""
Returns a DynamoDB-style dict of the keys/values.
Largely internal.
"""
raw_key_data = {}
for key, value in self.get_keys().items():
raw_key_data[key] = self._dynamizer.encode(value)
return raw_key_data
def build_expects(self, fields=None):
"""
Builds up a list of expecations to hand off to DynamoDB on save.
Largely internal.
"""
expects = {}
if fields is None:
fields = self._data.keys() + self._orig_data.keys()
# Only uniques.
fields = set(fields)
for key in fields:
expects[key] = {
'Exists': True,
}
value = None
# Check for invalid keys.
if not key in self._orig_data and not key in self._data:
raise ValueError("Unknown key %s provided." % key)
# States:
# * New field (_data & _orig_data w/ marker)
# * Unchanged field (only _data)
# * Modified field (_data & _orig_data)
# * Deleted field (only _orig_data)
if not key in self._orig_data:
# Existing field unchanged.
value = self._data[key]
else:
if key in self._data:
if self._orig_data[key] is NEWVALUE:
# New field.
expects[key]['Exists'] = False
else:
# Existing field modified.
value = self._orig_data[key]
else:
# Existing field deleted.
value = self._orig_data[key]
if value is not None:
expects[key]['Value'] = self._dynamizer.encode(value)
return expects
def prepare_full(self):
"""
Runs through all fields & encodes them to be handed off to DynamoDB
as part of an ``save`` (``put_item``) call.
Largely internal.
"""
# This doesn't save on it's own. Rather, we prepare the datastructure
# and hand-off to the table to handle creation/update.
final_data = {}
for key, value in self._data.items():
final_data[key] = self._dynamizer.encode(value)
return final_data
def prepare_partial(self):
"""
Runs through **ONLY** the changed/deleted fields & encodes them to be
handed off to DynamoDB as part of an ``partial_save`` (``update_item``)
call.
Largely internal.
"""
# This doesn't save on it's own. Rather, we prepare the datastructure
# and hand-off to the table to handle creation/update.
final_data = {}
# Loop over ``_orig_data`` so that we only build up data that's changed.
for key, value in self._orig_data.items():
if key in self._data:
# It changed.
final_data[key] = {
'Action': 'PUT',
'Value': self._dynamizer.encode(self._data[key])
}
else:
# It was deleted.
final_data[key] = {
'Action': 'DELETE',
}
return final_data
def partial_save(self):
"""
Saves only the changed data to DynamoDB.
Extremely useful for high-volume/high-write data sets, this allows
you to update only a handful of fields rather than having to push
entire items. This prevents many accidental overwrite situations as
well as saves on the amount of data to transfer over the wire.
Returns ``True`` on success, ``False`` if no save was performed or
the write failed.
Example::
>>> user['last_name'] = 'Doh!'
# Only the last name field will be sent to DynamoDB.
>>> user.partial_save()
"""
if not self.needs_save():
return False
key = self.get_keys()
# Build a new dict of only the data we're changing.
final_data = self.prepare_partial()
# Build expectations of only the fields we're planning to update.
expects = self.build_expects(fields=self._orig_data.keys())
returned = self.table._update_item(key, final_data, expects=expects)
# Mark the object as clean.
self.mark_clean()
return returned
def save(self, overwrite=False):
"""
Saves all data to DynamoDB.
By default, this attempts to ensure that none of the underlying
data has changed. If any fields have changed in between when the
``Item`` was constructed & when it is saved, this call will fail so
as not to cause any data loss.
If you're sure possibly overwriting data is acceptable, you can pass
an ``overwrite=True``. If that's not acceptable, you may be able to use
``Item.partial_save`` to only write the changed field data.
Optionally accepts an ``overwrite`` parameter, which should be a
boolean. If you provide ``True``, the item will be forcibly overwritten
within DynamoDB, even if another process changed the data in the
meantime. (Default: ``False``)
Returns ``True`` on success, ``False`` if no save was performed.
Example::
>>> user['last_name'] = 'Doh!'
# All data on the Item is sent to DynamoDB.
>>> user.save()
# If it fails, you can overwrite.
>>> user.save(overwrite=True)
"""
if not self.needs_save():
return False
final_data = self.prepare_full()
expects = None
if overwrite is False:
# Build expectations about *all* of the data.
expects = self.build_expects()
returned = self.table._put_item(final_data, expects=expects)
# Mark the object as clean.
self.mark_clean()
return returned
def delete(self):
"""
Deletes the item's data to DynamoDB.
Returns ``True`` on success.
Example::
# Buh-bye now.
>>> user.delete()
"""
key_data = self.get_keys()
return self.table.delete_item(**key_data)
class Item(object):
"""
An object representing the item data within a DynamoDB table.
An item is largely schema-free, meaning it can contain any data. The only
limitation is that it must have data for the fields in the ``Table``'s
schema.
This object presents a dictionary-like interface for accessing/storing
data. It also tries to intelligently track how data has changed throughout
the life of the instance, to be as efficient as possible about updates.
Empty items, or items that have no data, are considered falsey.
"""
def __init__(self, table, data=None, loaded=False):
"""
Constructs an (unsaved) ``Item`` instance.
To persist the data in DynamoDB, you'll need to call the ``Item.save``
(or ``Item.partial_save``) on the instance.
Requires a ``table`` parameter, which should be a ``Table`` instance.
This is required, as DynamoDB's API is focus around all operations
being table-level. It's also for persisting schema around many objects.
Optionally accepts a ``data`` parameter, which should be a dictionary
of the fields & values of the item.
Optionally accepts a ``loaded`` parameter, which should be a boolean.
``True`` if it was preexisting data loaded from DynamoDB, ``False`` if
it's new data from the user. Default is ``False``.
Example::
>>> users = Table('users')
>>> user = Item(users, data={
... 'username': '******',
... 'first_name': 'John',
... 'date_joined': 1248o61592,
... })
# Change existing data.
>>> user['first_name'] = 'Johann'
# Add more data.
>>> user['last_name'] = 'Doe'
# Delete data.
>>> del user['date_joined']
# Iterate over all the data.
>>> for field, val in user.items():
... print "%s: %s" % (field, val)
username: johndoe
first_name: John
date_joined: 1248o61592
"""
self.table = table
self._loaded = loaded
self._orig_data = {}
self._data = data
self._dynamizer = Dynamizer()
if self._data is None:
self._data = {}
if self._loaded:
self._orig_data = deepcopy(self._data)
def __getitem__(self, key):
return self._data.get(key, None)
def __setitem__(self, key, value):
self._data[key] = value
def __delitem__(self, key):
if not key in self._data:
return
del self._data[key]
def keys(self):
return self._data.keys()
def values(self):
return self._data.values()
def items(self):
return self._data.items()
def get(self, key, default=None):
return self._data.get(key, default)
def __iter__(self):
for key in self._data:
yield self._data[key]
def __contains__(self, key):
return key in self._data
def __nonzero__(self):
return bool(self._data)
def _determine_alterations(self):
"""
Checks the ``-orig_data`` against the ``_data`` to determine what
changes to the data are present.
Returns a dictionary containing the keys ``adds``, ``changes`` &
``deletes``, containing the updated data.
"""
alterations = {
'adds': {},
'changes': {},
'deletes': [],
}
orig_keys = set(self._orig_data.keys())
data_keys = set(self._data.keys())
# Run through keys we know are in both for changes.
for key in orig_keys.intersection(data_keys):
if self._data[key] != self._orig_data[key]:
if self._is_storable(self._data[key]):
alterations['changes'][key] = self._data[key]
else:
alterations['deletes'].append(key)
# Run through additions.
for key in data_keys.difference(orig_keys):
if self._is_storable(self._data[key]):
alterations['adds'][key] = self._data[key]
# Run through deletions.
for key in orig_keys.difference(data_keys):
alterations['deletes'].append(key)
return alterations
def needs_save(self, data=None):
"""
Returns whether or not the data has changed on the ``Item``.
Optionally accepts a ``data`` argument, which accepts the output from
``self._determine_alterations()`` if you've already called it. Typically
unnecessary to do. Default is ``None``.
Example:
>>> user.needs_save()
False
>>> user['first_name'] = 'Johann'
>>> user.needs_save()
True
"""
if data is None:
data = self._determine_alterations()
needs_save = False
for kind in ['adds', 'changes', 'deletes']:
if len(data[kind]):
needs_save = True
break
return needs_save
def mark_clean(self):
"""
Marks an ``Item`` instance as no longer needing to be saved.
Example:
>>> user.needs_save()
False
>>> user['first_name'] = 'Johann'
>>> user.needs_save()
True
>>> user.mark_clean()
>>> user.needs_save()
False
"""
self._orig_data = deepcopy(self._data)
def mark_dirty(self):
"""
DEPRECATED: Marks an ``Item`` instance as needing to be saved.
This method is no longer necessary, as the state tracking on ``Item``
has been improved to automatically detect proper state.
"""
return
def load(self, data):
"""
This is only useful when being handed raw data from DynamoDB directly.
If you have a Python datastructure already, use the ``__init__`` or
manually set the data instead.
Largely internal, unless you know what you're doing or are trying to
mix the low-level & high-level APIs.
"""
self._data = {}
for field_name, field_value in data.get('Item', {}).items():
self[field_name] = self._dynamizer.decode(field_value)
self._loaded = True
self._orig_data = deepcopy(self._data)
def get_keys(self):
"""
Returns a Python-style dict of the keys/values.
Largely internal.
"""
key_fields = self.table.get_key_fields()
key_data = {}
for key in key_fields:
key_data[key] = self[key]
return key_data
def get_raw_keys(self):
"""
Returns a DynamoDB-style dict of the keys/values.
Largely internal.
"""
raw_key_data = {}
for key, value in self.get_keys().items():
raw_key_data[key] = self._dynamizer.encode(value)
return raw_key_data
def build_expects(self, fields=None):
"""
Builds up a list of expecations to hand off to DynamoDB on save.
Largely internal.
"""
expects = {}
if fields is None:
fields = self._data.keys() + self._orig_data.keys()
# Only uniques.
fields = set(fields)
for key in fields:
expects[key] = {
'Exists': True,
}
value = None
# Check for invalid keys.
if not key in self._orig_data and not key in self._data:
raise ValueError("Unknown key %s provided." % key)
# States:
# * New field (only in _data)
# * Unchanged field (in both _data & _orig_data, same data)
# * Modified field (in both _data & _orig_data, different data)
# * Deleted field (only in _orig_data)
orig_value = self._orig_data.get(key, NEWVALUE)
current_value = self._data.get(key, NEWVALUE)
if orig_value == current_value:
# Existing field unchanged.
value = current_value
else:
if key in self._data:
if not key in self._orig_data:
# New field.
expects[key]['Exists'] = False
else:
# Existing field modified.
value = orig_value
else:
# Existing field deleted.
value = orig_value
if value is not None:
expects[key]['Value'] = self._dynamizer.encode(value)
return expects
def _is_storable(self, value):
# We need to prevent ``None``, empty string & empty set from
# heading to DDB, but allow false-y values like 0 & False make it.
if not value:
if not value in (0, 0.0, False):
return False
return True
def prepare_full(self):
"""
Runs through all fields & encodes them to be handed off to DynamoDB
as part of an ``save`` (``put_item``) call.
Largely internal.
"""
# This doesn't save on it's own. Rather, we prepare the datastructure
# and hand-off to the table to handle creation/update.
final_data = {}
for key, value in self._data.items():
if not self._is_storable(value):
continue
final_data[key] = self._dynamizer.encode(value)
return final_data
def prepare_partial(self):
"""
Runs through **ONLY** the changed/deleted fields & encodes them to be
handed off to DynamoDB as part of an ``partial_save`` (``update_item``)
call.
Largely internal.
"""
# This doesn't save on it's own. Rather, we prepare the datastructure
# and hand-off to the table to handle creation/update.
final_data = {}
fields = set()
alterations = self._determine_alterations()
for key, value in alterations['adds'].items():
final_data[key] = {
'Action': 'PUT',
'Value': self._dynamizer.encode(self._data[key])
}
fields.add(key)
for key, value in alterations['changes'].items():
final_data[key] = {
'Action': 'PUT',
'Value': self._dynamizer.encode(self._data[key])
}
fields.add(key)
for key in alterations['deletes']:
final_data[key] = {
'Action': 'DELETE',
}
fields.add(key)
return final_data, fields
def partial_save(self):
"""
Saves only the changed data to DynamoDB.
Extremely useful for high-volume/high-write data sets, this allows
you to update only a handful of fields rather than having to push
entire items. This prevents many accidental overwrite situations as
well as saves on the amount of data to transfer over the wire.
Returns ``True`` on success, ``False`` if no save was performed or
the write failed.
Example::
>>> user['last_name'] = 'Doh!'
# Only the last name field will be sent to DynamoDB.
>>> user.partial_save()
"""
key = self.get_keys()
# Build a new dict of only the data we're changing.
final_data, fields = self.prepare_partial()
if not final_data:
return False
# Remove the key(s) from the ``final_data`` if present.
# They should only be present if this is a new item, in which
# case we shouldn't be sending as part of the data to update.
for fieldname, value in key.items():
if fieldname in final_data:
del final_data[fieldname]
try:
# It's likely also in ``fields``, so remove it there too.
fields.remove(fieldname)
except KeyError:
pass
# Build expectations of only the fields we're planning to update.
expects = self.build_expects(fields=fields)
returned = self.table._update_item(key, final_data, expects=expects)
# Mark the object as clean.
self.mark_clean()
return returned
def save(self, overwrite=False):
"""
Saves all data to DynamoDB.
By default, this attempts to ensure that none of the underlying
data has changed. If any fields have changed in between when the
``Item`` was constructed & when it is saved, this call will fail so
as not to cause any data loss.
If you're sure possibly overwriting data is acceptable, you can pass
an ``overwrite=True``. If that's not acceptable, you may be able to use
``Item.partial_save`` to only write the changed field data.
Optionally accepts an ``overwrite`` parameter, which should be a
boolean. If you provide ``True``, the item will be forcibly overwritten
within DynamoDB, even if another process changed the data in the
meantime. (Default: ``False``)
Returns ``True`` on success, ``False`` if no save was performed.
Example::
>>> user['last_name'] = 'Doh!'
# All data on the Item is sent to DynamoDB.
>>> user.save()
# If it fails, you can overwrite.
>>> user.save(overwrite=True)
"""
if not self.needs_save() and not overwrite:
return False
final_data = self.prepare_full()
expects = None
if overwrite is False:
# Build expectations about *all* of the data.
expects = self.build_expects()
returned = self.table._put_item(final_data, expects=expects)
# Mark the object as clean.
self.mark_clean()
return returned
def delete(self):
"""
Deletes the item's data to DynamoDB.
Returns ``True`` on success.
Example::
# Buh-bye now.
>>> user.delete()
"""
key_data = self.get_keys()
return self.table.delete_item(**key_data)
def _encode_item(self, item):
item_cp = item.copy()
for key, val in item_cp.iteritems():
item_cp.update({key: Dynamizer().encode(val)})
return item_cp