fs-synapse 3.0.0

A Synapse implementation of the fsspec interface

fs-synapse

A Synapse implementation of the fsspec interface.

fs-synapse allows us to leverage the fsspec API to interface with Synapse files, folders, and projects. By learning this API, you can write code that is agnostic to where your files are physically located. This is achieved by referring to Synapse entities using URLs. Commented examples are included below.

syn://syn50545516               # Synapse project

syn://syn50557597               # Folder in the above Synapse project
syn://syn50545516/syn50557597   # Same folder, but using a full path
syn://syn50545516/TestSubDir    # Same folder, but referenced by name

syn://syn50555279               # File in the above Synapse project
syn://syn50545516/syn50555279   # Same file, but using a full path
syn://syn50545516/test.txt      # Same file, but referenced by name

syn://syn50545516/ExploratoryTests/report.json      # Nested file

Benefits

There are several benefits to using the fs-synapse API over synapseclient.

from synapsefs import SynapseFS

fs = SynapseFS()

# Or using fsspec directly
import fsspec
fs = fsspec.filesystem("syn")

Interact with Synapse using a Pythonic interface

file_url = "syn://syn50555279"

with fs.open(file_url, "a") as fp:
    fp.write("Appending some text to a Synapse file")

Access to several convenience functions

folder_url = "syn://syn50696438"

fs.makedirs(f"{folder_url}/creating/nested/folders/with/one/operation")

Refer to Synapse files and folders by name

You don't have to track as many Synapse IDs. You only need to care about the top-level projects or folders and refer to subfolders and files by name.

project_url = "syn://syn50545516"

data_url = f"{project_url}/data/raw.csv"
output_url = f"{project_url}/outputs/processed.csv"

with fs.open(data_url, "r") as data_fp, fs.open(output_url, "a") as output_fp:
    results = number_cruncher(data)
    output.write(results)

Write Synapse-agnostic code

Unfortunately, every time you use synapseclient for file and folder operations, you are hard-coding a dependency on Synapse into your project. Leveraging fs-synapse helps avoid this hard dependency and makes your code more portable to other file backends (e.g. S3). You can swap for any other file system by using their URL scheme (e.g. s3://). Here's an index of available file systems that you can swap for.

Rely on code covered by integration tests

So you don't have to write the Synapse integration tests yourself! These tests tend to be slow, so delegating that responsibilty to an externally managed package like fs-synapse keeps your test suite fast and focused on what you care about.

In your test code, you can use the memory filesystem for faster I/O instead of storing and retrieving files on Synapse.

def test_some_feature_of_your_code():
    fs = fsspec.filesystem("memory")
    cruncher = NumberCruncher(fs=fs)
    cruncher.save("report.json")
    assert fs.exists("report.json")

Migration from PyFilesystem2 to fsspec

This package previously used PyFilesystem2 (fs) as its base. It now uses fsspec. The table below maps the old API to the new one.

Initialization

Old (PyFilesystem2)	New (fsspec)
from fs import open_fs	import fsspec
fs = open_fs("syn://")	fs = fsspec.filesystem("syn")
fs = open_fs("syn://syn50545516")	fs = SynapseFS(root="syn50545516")

File operations

Old (PyFilesystem2)	New (fsspec)
fs.open(path, "r")	fs.open(path, "r")
fs.readtext(path)	fs.cat_file(path).decode()
fs.readbytes(path)	fs.cat_file(path)
fs.writetext(path, text)	fs.pipe_file(path, text.encode())
fs.writebytes(path, data)	fs.pipe_file(path, data)
fs.create(path) / fs.touch(path)	fs.touch(path)
fs.download(name, file_obj)	fs.get(path, local_path)

Directory operations

Old (PyFilesystem2)	New (fsspec)
fs.listdir(path)	fs.ls(path, detail=False) (returns full paths)
fs.makedir(path)	fs.mkdir(path)
fs.makedirs(path)	fs.makedirs(path)
fs.opendir(path)	(no equivalent; use full paths)
fs.tree(path=path)	fs.ls(path, detail=True)

Removal

Old (PyFilesystem2)	New (fsspec)
fs.remove(path)	fs.rm(path)
fs.removedir(path)	fs.rmdir(path)
fs.removetree(path)	fs.rm(path, recursive=True)

Info and metadata

Old (PyFilesystem2)	New (fsspec)
info = fs.getinfo(path, namespaces=["details", "synapse"])	info = fs.info(path)
info.name	info["name"]
info.is_dir	info["type"] == "directory"
info.get("details", "size")	info["size"]
info.get("synapse", "id")	info["synapse_id"]
info.get("synapse", "content_type")	info["synapse_content_type"]
info.get("synapse", "etag")	info["synapse_etag"]
fs.getsize(path)	fs.info(path)["size"]
fs.gettype(path)	fs.info(path)["type"]
fs.exists(path)	fs.exists(path)

Errors

Old (PyFilesystem2)	New (fsspec)
fs.errors.ResourceNotFound	FileNotFoundError
fs.errors.FileExists / DirectoryExists	FileExistsError
fs.errors.FileExpected	IsADirectoryError
fs.errors.DirectoryExpected	NotADirectoryError
fs.errors.CreateFailed	ValueError
fs.errors.ResourceInvalid	ValueError
fs.errors.RemoveRootError	PermissionError
fs.errors.DirectoryNotEmpty	OSError

PyScaffold

This project has been set up using PyScaffold 4.3. For details and usage information on PyScaffold see PyScaffold.

putup --name fs-synapse --markdown --github-actions --pre-commit --license Apache-2.0 fs-synapse