A very simple shared memory dict implementation.
Requires: Python >= 3.8
>>> # In the first Python interactive shell
>> from shared_memory_dict import SharedMemoryDict
>> smd = SharedMemoryDict(name='tokens', size=1024)
>> smd['some-key'] = 'some-value-with-any-type'
>> smd['some-key']
'some-value-with-any-type'
>>> # In either the same shell or a new Python shell on the same machine
>> existing_smd = SharedMemoryDict(name='tokens', size=1024)
>>> existing_smd['some-key']
'some-value-with-any-type'
>>> existing_smd['new-key'] = 'some-value-with-any-type'
>>> # Back in the first Python interactive shell, smd reflects this change
>> smd['new-key']
'some-value-with-any-type'
>>> # Clean up from within the second Python shell
>>> existing_smd.shm.close() # or "del existing_smd"
>>> # Clean up from within the first Python shell
>>> smd.shm.close()
>>> smd.shm.unlink() # Free and release the shared memory block at the very end
>>> del smd # use of smd after call unlink() is unsupported
The arg
name
defines the location of the memory block, so if you want to share the memory between process use the same name. The size (in bytes) occupied by the contents of the dictionary depends on the serialization used in storage. By default pickle is used.
Using pip
:
pip install shared-memory-dict
To use multiprocessing.Lock on write operations of shared memory dict set environment variable SHARED_MEMORY_USE_LOCK=1
.
We use pickle as default to read and write the data into the shared memory block.
You can create a custom serializer by implementing the dumps
and loads
methods.
Custom serializers should raise SerializationError
if the serialization fails and DeserializationError
if the deserialization fails. Both are defined in the shared_memory_dict.serializers
module.
An example of a JSON serializer extracted from serializers module:
NULL_BYTE: Final = b"\x00"
class JSONSerializer:
def dumps(self, obj: dict) -> bytes:
try:
return json.dumps(obj).encode() + NULL_BYTE
except (ValueError, TypeError):
raise SerializationError(obj)
def loads(self, data: bytes) -> dict:
data = data.split(NULL_BYTE, 1)[0]
try:
return json.loads(data)
except json.JSONDecodeError:
raise DeserializationError(data)
Note: A null byte is used to separate the dictionary contents from the bytes that are in memory.
To use the custom serializer you must set it when creating a new shared memory dict instance:
>>> smd = SharedMemoryDict(name='tokens', size=1024, serializer=JSONSerializer())
The pickle module is not secure. Only unpickle data you trust.
See more here.
There's a Django Cache Implementation with Shared Memory Dict:
# settings/base.py
CACHES = {
'default': {
'BACKEND': 'shared_memory_dict.caches.django.SharedMemoryCache',
'LOCATION': 'memory',
'OPTIONS': {'MEMORY_BLOCK_SIZE': 1024}
}
}
Install with: pip install "shared-memory-dict[django]"
With Django cache implementation the keys only expire when they're read. Be careful with memory usage
There's also a AioCache Backend Implementation with Shared Memory Dict:
From aiocache import caches
caches.set_config({
'default': {
'cache': 'shared_memory_dict.caches.aiocache.SharedMemoryCache',
'size': 1024,
},
})
This implementation is very based on aiocache SimpleMemoryCache
Install with: pip install "shared-memory-dict[aiocache]"