A Python library to make storing files simple and easy.
⚠️ Although there are extensive tests within this project for Python versions 3.8 to 3.12, it is a young project and there may be bugs and security holes. Be sure and test thoroughly prior to use in a production environment.
It is primarily intended to deal with file uploads to a static files directory or an object service like AWS S3 or Linode. Files can by stored synchronously (for WSGI servers like Django, Flask or Pyramid) or asynchronously (for ASGI servers like Starlette, FastAPI or Django Channels).
Supports multiple storage services simultaneously or even the same service with multiple configurations.
Upload filters are easy to create and a few are included by default.
The library is available through the Python Package Index and can be installed with pip.
pip install filestorage
Different handlers have additional library requirements that can be optionally installed. For instance, the async local file handler requirements can be installed using:
pip install "filestorage[aio_file]"
The extras are:
aio_file
- requirements for async local file handling.s3
- requirements for storing to an AWS S3 bucket.
Interaction with the library is primarily accomplished through a global store
object. Any Python file can access this global object by importing it.
from filestorage import store
store.finalized # == False
If you are uncomfortable by the existance of a global store, fear not! You can make a new instance of the store using the StorageContainer class.
my_store = StorageContainer()
, the use that in your program.
The store can hold multiple configurations that are accessed through its indices:
store['portraits'].finalized # == False
store['more']['and more'].finalized # == False
PORTRAIT_STORE = store['portraits'] # global variable for later use!
The store and any of its other sub-configurations can be saved and referenced prior to setting up any configuration. This allows the store or any other sub-configuration to be imported anywhere and stored within global variables as needed.
Trying to use the store to save files prior to providing it a handler will result in an error.
store.save_data(filename='file.txt', data=b'spamity spam')
# FilestorageConfigError: No handler provided for store
So it's time to give it a handler.
Although any file can import the store, it will not actually be usable until it is provided some kind of configuration. You do this by giving the store a handler. This library includes a few kinds of handlers with different configuration setups. For testing it's easiest to use the DummyHandler:
from filestore.handlers import DummmyHandler
store.handler = DummyHandler()
store['portraits'] = DummyHandler()
The DummyHandler doesn't need any other configuration and just stores any saved files in memory.
In your app, you would would normally perform this configuration after all files have been imported but before the app starts - for instance in an initialization step where the app configuration is read. Different frameworks (Pyramid/Django/Starlette) will have different mechanisms for this.
Now you can try to save some data into the store:
store.save_data(filename='file.txt', data=b'spamity spam')
# 'file.txt' - the name of the file that was saved
store.exists(filename='file.txt')
# True
Handlers and filters might adjust the filename in the process of storing the file. The return value gives you feedback of the actual name of the file that was saved to the store. This might be useful for storing in a database for later reference.
You can set and reset the handler several times. When you want to lock in the configuration and prevent further changes, you need to finalize the store. When you do so, all handlers and filters will validate their configuration and if any handlers are missing a configuration error will be thrown. Any attempts to set a handler after this step will throw an error. This ensures that the store is properly set up once and only once.
store.finalize_config()
store.handler = DummyHandler()
# FilestorageConfigError: Setting store.handler: store already finalized!
If using an ASGI server, you may need to instead use an async startup task that contains:
await store.async_finalize_config()
The finalization step also will validate any handler configuration. For instance, the local file handler ensures its configured directory exists, the async file handler ensures the proper libraries are installed and the S3 handler verifies the credentials by saving and deleting a dummy file in the bucket.
If the store or one of the sub-stores are not intended to be used, they must explicitly have their handler set to None
. For instance, suppose
you want to use two different configurations, but don't want to use the store's base config. You can do:
store['portraits'].handler = DummyHandler()
store['backgrounds'].handler = DummyHandler()
# Indicate that we don't want anybody using the base configurations (no `store.save_data()`)
store.handler = None
# Finalize the configuration to prevent further changes
store.finalize_config()
The save/exists/delete methods can also be used on any sub-folder of the store
.
folder = store / 'one_folder' / 'another_folder'
folder.save_data(b'contents', 'file.txt')
Folders are late-binding and can thus be saved into a variable prior to the handler configuration.
If desired, a Folder can be used as a handler too:
store.handler = DummyHandler()
store['portraits'].handler = store / 'portraits'
Filters allow mutating a file to be stored to a Handler. They are called in the order defined and a few filters are provided by this library.
For instance, it's often best not to simply store filenames provided by random Internet uploads. Although this library does scrub the filename, it's not as fool-proof as simply ignoring the provided filename and using a random string with a consistent length. The RandomizeFilename filter does just that.
from filestorage.filters import RandomizeFilename
store.handler = DummyHandler(filters=[RandomizeFilename()])
store.save_data(filename='ignored name.txt', data=b'contents')
'5d29f7e1-50c0-4dc6-a466-b2ae042e8ac0.txt'
For more on filters, see the Filter class definition.
This library can behave as a Pyramid plugin. When doing so, it will read from the Pyramid configuration and set up the handler(s) defined there. The store
will also be available on the request
object as request.store
.
To set it up, include filestorage.pyramid_config
in your configuration using the Pyramid Configurator's include method.
from pyramid.config import Configurator
def main(global_config, **settings):
config = Configurator()
config.include('filestorage.pyramid_config'). # <---
...
Add any handler configuration to your app's config file. The handler and filters can refer to any handler or filter within filestorage
, or can refer to any other package by full module path and model name.
[app:main]
# (other config settings)
# Base store with a custom handler and a custom filter
store.handler = myapp.filestorage.MyCustomHandler
store.filters[0] = myapp.filestorage.MyCustomFilter
# Portrait store with a couple of filestorage filters
store['portrait'].handler = LocalFileHandler
store['portrait'].handler.base_path = /var/www/static/uploaded_images
store['portrait'].handler.base_url = http://my.portraits/static
store['portrait'].handler.filters[0] = RandomizeFilename
store['portrait'].handler.filters[1] = ValidateExtension
store['portrait'].handler.filters[1].extensions = ['jpg', 'png']
# Another store that exists, but has been disabled.
store['not_used'].handler = None
Any parameters for filters and handlers are decoded and passed in as kwargs. Anything that looks like a list, set or dict literal is eval
ed and bare whole numbers are converted to int
. You can force anything to be a string by enclosing it in single or double quotes: "50"
.
The store is then usable in any view:
def save_file(request):
uploaded_file = request.POST['file']
filename = request.store.save_field(uploaded_file)
return Response(f'Saved to {filename}!')
Additional optional settings:
store.request_property
- (Defaultstore
) - Name of the property to use on the request object. For example, set this tomy_store
then access the store throughrequest.my_store
.store.use_global
- (DefaultTrue
) - Use the globalstore
object. If set toFalse
then therequest.store
object will independent of the globalstore
object.
This is the class for the global store
object.
Methods:
handler
- Gets the configured handler or raises an exception of no handler has been provided. Set this property to set the handler.sync_handler
- Gets the configured handler as a sync-only handler. Raises an exception if nohandler
has been set.async_handler
- Gets the configured handler as an async-only handler. Raises an exception if nohandler
has been set or if the configured handler can't be used asynchronously.finalize_config()
- Walk through all configured objects and check to ensure they have a valid configuration. Lock theStorageContainer
to prevent any further configuration changes. Will raise aFilestorageConfigError
if there's a configuration problem.async_finalize_config()
- awaitable version of the above call. Necessary for ASGI servers.finalized
-True
if the config has been finalized,False
otherwise.do_not_use
-True
if thehandler
has been set toNone
,False
otherwise.name
- String name of how this configuration is accessed.store['a'].name == "['a']"
.[*]
- Get a sub-configuration object. Raises aFilestorageConfigError
if the configuration is finalized and this configuration'shandler
hasn't been set./ 'name'
- Obtain aFolder
object with the same save/exist/delete methods as this object which write to the named sub-folder.
Once the handler is set, the store object can be used as a StorageHandler
object.
(StorageHandlerBase
or AsyncStorageHandlerBase
)
All handlers inherit from StorageHandlerBase
.
The async version of the Handler can be used for either synchronous or asynchronous operations. The StorageHandlerBase
by itself can only be used for synchronous operations and any async_*
method calls will throw an error. To make a new custom handler, start with the handler template.
⚠️ Ensure your forms include the attribute enctype=”multipart/form-data” or your uploaded files will be empty. Short example and more detail.
Parameters:
base_url
- Optional string - The URL prefix for any saved file. For example:'https://eppx.com/static/'
. If not set, theget_url
method will only return the path string, not the full URL.path
- Optional list or string - The path within the store (and URL) for any saved file. For example:['folder', 'subfolder']
filters
- Optional list of Filters to apply, in order, when saving any file through this handler. For example:[RandomizeFilename()]
⚠️ All time related methods return the specified time in the timezone of the handler.
Methods:
get_url(filename: str)
- Return the full URL of the given filename. If thebase_url
parameter isn't set, will only return the path string instead of the full URL.sanitize_filename(filename: str)
- Return the string stripped of dangerous characters.- Synchronous methods:
exists(filename: str)
-True
if the given file exists in the store, false otherwise.size(filename: str)
-int
The size in bytes of the given file.get_accessed_time(filename: str)
-datetime
The time of last access for the given file.get_created_time(filename: str)
-datetime
The time of creation for the given file. This returns different times based on operating system. Click here for more detail.get_modified_time(filename: str)
-datetime
The time of last modification for the given file.delete(filename: str)
- Deletes the given file from the store.save_file(filename: str, data: BinaryIO)
- Save the binary IO object to the given file.save_data(filename: str, data: bytes)
- Save the binary data to the given file.save_field(field: cgi.FieldStorage)
- Save the given field storage object.
- Asynchronous methods: (all will throw a
FilestorageConfigError
if the handler doesn't support async operations.)async_exists(filename: str)
- Awaitable versionasync_size(filename: str)
- Awaitable versionasync_get_accessed_time(filename: str)
- Awaitable versionasync_get_created_time(filename: str)
- Awaitable versionasync_get_modified_time(filename: str)
- Awaitable versionasync_delete(filename: str)
- Awaitable versionasync_save_file(filename: str, data: BinaryIO)
- Awaitable versionasync_save_data(filename: str, data: binary)
- Awaitable versionasync_save_field(field: cgi.FieldStorage)
- Awaitable version
Abstract Methods to be overridden when sub-classing:
_validate()
- Check to ensure the provided configuration is correct. Can be an async method or return aFuture
object.- Synchronous methods: (All get passed a FileItem object)
_exists(item: FileItem)
- ReturnsTrue
/False
to indicate if the item exists in the storage container._size(filename: str)
- Returns the size in bytes of the given file._get_accessed_time(filename: str)
Returns the time of last access for the given file._get_created_time(filename: str)
Returns the time of creation for the given file._get_modified_time(filename: str)
Returns the time of last modification for the given file._delete(item: FileItem)
- Remove the item from the storage container._save(item: FileItem)
- Save the item to the storage container and return the name of the file saved.
- Asynchronous methods:
async _async_exists(item: FileItem)
- async version, returnsTrue
orFalse
.async_size(filename: str)
- async version.async_get_accessed_time(filename: str)
async version, returns the time of last access for the given file.async_get_created_time(filename: str)
async version, returns the time of creation for the given file.async_get_modified_time(filename: str)
async version, returns the time of last modification for the given file.async _async_delete(item: FileItem)
- async version.async _async_save(item: FileItem)
- async version, returns the stored filename.
The FilterBase
is used as a base class for any Filters. These are not intended to be used directly, but to be passed as an optional list to a Handler through the filters
parameter. To make a new custom filter, start with the filter template.
Properties:
async_ok
-True
/False
to indicate if this Filter is OK to be used asynchronously.
Methods:
call(item: FileItem)
- Returns the filtered FileItem.async_call(item: FileItem)
- Awaitable version ofcall
. If the filter can't be used asynchronously, will raise aFilestorageConfigError
.validate()
- Checks to ensure the Filter is configured correctly. Might return an Awaitable.
Abstract Methods to be overridden when sub-classing:
_apply(item: FileItem)
- Returns the filtered FileItem. It can optionally be an async method that will be awaited on. Can raise an exception if the item fails the filter._validate()
- Check to ensure the provided configuration is correct. Can be an async method or return aFuture
object.
This is the basic item that's passed through Filters to the Handlers. It is based on a NamedTuple and is therefore immutable.
Parameters, which are also Properties:
filename
- Stringpath
- Optional tuple of the path under which thefilename
can be accessed or stored. Example:('folder', 'subfolder')
.data
- Optional BinaryIO of the contents to store. Example:BytesIO(b'contents')
.media_type
- Optional string describing the media (or MIME) type contained within the data. If not provided, the type will be guessed as needed based on the filename's extension.
Properties:
url_path
- Relative path and filename for use in a URL. Example:'folder/subfolder/file.txt'
fs_path
- Relative path and filename for use in the local filesystem. Example when running on Windows:'folder\subfolder\file.txt'
.has_data
-True
if data is populated.content_type
- String indicating themedia_type
to be used in HTTP headers as needed.
Methods:
copy(**kwargs)
- Create a copy of this object, overriding any specific parameter. Example:item.copy(filename='new_name.txt')
. This is very useful in a Filter that changes the filename or other properties.
Context Manager:
The FileItem can be used as a context manager, where it will modify the read/seek methods of the underlying object to fit the requested stream as necessary - either sync or async.
with FileItem as f
- Returns an object (f
) that behaves as an open file.f.read()
andf.seek()
methods are supported.async with FileItem as f
- Returns an object (f
) that behaves as an async open file.await f.read()
andawait f.seek()
methods are supported.
All are importable via the exceptions
sub-package. For example:
from filestorage.exceptions import FilestorageError
- FilestorageError - Base class for any exceptions raised by this library.
- FileNotAllowed - The provided file is not allowed, either through a Filter or from a Handler.
- FileExtensionNotAllowed - The provided file with the given extension is not allowed, either through a Filter or from a Handler.
- FilestorageConfigError - There was some problem with the configuration.
All handlers are subclasses of the StorageHandler class. These can be imported via the handlers
sub-package. For example:
from filestorage.handlers import LocalFileHandler
store.handler = LocalFileHandler(base_path='/home/www/uploads`)
Store files on the local file system using synchronous methods.
Async not OK.
Parameters:
base_path
- Where to store the files on the local filesystemauto_make_dir
- Automatically create the directory as needed.
Store files on the local file system using asynchronous methods.
Async OK.
⚠️ Requires theaiofiles
library, which will be installed withpip install filestorage['aio_file']
Parameters:
base_path
- Where to store the files on the local filesystemauto_make_dir
(defualt:False
)- Automatically create the directory as needed.allow_sync_methods
(default:True
) - WhenFalse
, all synchronous calls throw aRuntimeError
. Might be helpful in preventing accidentally using the syncsave
/exists
/delete
methods, which would block the async loop too.
Store files to an S3 bucket. This handler works for synchronous and asynchronous calls.
Async OK.
⚠️ Requires theaioboto3
library, which will be installed withpip install filestorage['s3']
⚠️ Requires appropriate AWS permissions to the S3 bucket.
Parameters:
bucket_name
- required - AWS bucket to store the files in.acl
(default:'public-read'
) - Access-Control to apply to newly saved files. Other interesting options are'private'
,'authenticate-read'
, and'bucket-owner-read'
. See the S3 documentation for more information.connect_timeout
(default:5
) - Seconds to wait for a connection event before timeout.num_retries
(default:5
- How many times the library will attempt to connect before failing.read_timeout
(default:10
) Seconds to wait for a read event before timeout.keepalive_timeout
(default:12
) - Send a packet every few seconds to keep active connections open.host_url
- When using non-AWS S3 service (like Linode), use this url to connect. (Example:'https://us-east-1.linodeobjects.com'
)region_name
- Overrides any region_name defined in the AWS configuration file or theAWS_DEFAULT_REGION
environment variable. Required if using AWS S3 and the value is not already set elsewhere.addressing_style
- Overrides any S3.addressing_style set in the AWS configuration file.allow_sync_methods
(default:True
) - WhenFalse
, all synchronous calls throw aRuntimeError
. Might be helpful in preventing accidentally using the syncsave
/exists
/delete
methods, which would block the async loop too.
Permissions can be configured in three different ways. They can be stored in environment variables, then can be stored in a particular AWS file, or they can be passed in directly.
See the boto3 credentials documentation for details on how to configure the required keys.
⚠️ Do not hard-code secret information in your source code!
If you wish to directly provide the connection information to this handler, use the following optional parameters:
aws_access_key_id
- or use theAWS_ACCESS_KEY_ID
environment variable.aws_secret_access_key
- or use theAWS_SECRET_ACCESS_KEY
environment variable.aws_session_token
- Unnecessary when using the two previous options, but here for completeness. Note that session tokens have a maximum age of 36 hours. Could instead use theAWS_SESSION_TOKEN
environment variable.profile_name
- Which profile to use within a shared credentials file.
Handler used to test the file store - keeps any stored files in memory.
Async not OK - will fail for any async call. For an async version use the AsyncDummyHandler.
Accepts no parameters.
Properties:
files
- a dictionary of all saved files. The key is the path/filename string and the value is a FileItem object.validated
-True
if the handler was validated (which happens while finalizing the config).last_save
- The last FileItem saved.last_delete
- The last FileItem deleted.
Methods:
assert_exists(filename: str, path: Tuple[str, ...]
- Asserts that the provided filename and path have been saved.assert_size(self, filename: str, path: Tuple[str, ...], size: int)
- Assert that given file size is equal to the anticipated size.assert_get_accessed_time(self, filename: str, path: Tuple[str, ...], date: datetime)
- Assert that given file access time is equal to the anticipated time.assert_get_created_time(self, filename: str, path: Tuple[str, ...], date: datetime)
- Assert that given file creation time is equal to the anticipated time.assert_get_modified_time(self, filename: str, path: Tuple[str, ...], date: datetime)
- Assert that given file modification time is equal to the anticipated time.assert_file_contains(filename: str, path: Tuple[str, ...], data: bytes)
- Asserts that the saved file contains the given data.
Identical to the DummyHandler, but can be used asynchronously.
All are importable via the filters
sub-package. For example:
from filestorage.filter import RandomizeFilename
store.handler = DummyHandler(filters=[RandomizeFilename()])
Filter to randomize the filename. It keeps any extension within the filename, but replaces the name with a random string.
Async OK.
Parameters:
name_generator
Optional method that takes the filename (without extension) and returns a new random name. When left off, this Filter uses the uuid4 method.
Reject any filename that does not have one of the indicated extensions. Raises a FileExtensionNotAllowed
for a disallowed extension.
Async OK.
Parameters:
extensions
- List of exceptions to allow through this filter. Example:['jpg', 'png']
The DummyHandler or AsyncDummyHandler are great tools for testing your application. To keep your tests isolated, you can create a new store object and configure it for each test as needed.
from filestorage import StorageContainer
from filestorage.handlers import AsyncDummyHandler
def test_store():
store = StorageContainer()
dummy_handler = AsyncDummyHandler()
store.handler = dummy_handler
# Do whatever should trigger your app to store a file
dummy_handler.assert_exists('name.txt', ('folder', 'subfolder'))
If you need to write several tests and want to check the result, it's probably best to create a couple of simple fixtures and allow pytest to inject them as needed.
Within tests/conftest.py
:
import pytest
from filestorage import StorageContainer
from filestorage.handlers import AsyncDummyHandler
@pytest.fixture
def dummy_handler():
return AsyncDummyHandler()
@pytest.fixture
def store(dummy_handler):
store = StorageContainer()
store.handler = dummy_handler
store.finalize_config()
return store
Within tests/test_myapp.py
def test_store(store, dummy_handler):
# The configured dummy handler and store were auto-created and passed in here
# Do whatever should trigger your app to store a file
dummy_handler.assert_exists('name.txt', ('folder', 'subfolder'))