Atomik provides you a python-friendly way to manage writing new file and folders in an atomic way.
The main way to use the library is with to use the file
and folder
context manager. They will provide
an easy interface to make sure the destination file/folder remain consistent.
This means you can be sure that you will not leave a file half written or a folder with half the files you wanted to write into in case of failure. When overwriting you can be sure that the data will be either the previous or the next version.
This will ensure either all the data is correctly written to file or none of it
with atomik.file("./result.json") as f:
print(stream, file=f)
with atomik.folder("./result_folder") as dir:
with open(Path(dir, "result_1.json") as f:
print(result_1, file=f)
with open(Path(dir, "result_2.json") as f:
print(result_2, file=f)
By default, atomik will not overwrite files/folders (but it's configurable)
Atomik uses the new renameat2
syscall to allow to rename file in an atomic way.
The syscall doesn't allow to overwrite a non-empty folder when renaming.
This is achieved using the exchange
flag to swap the two folder and then the source
folder is cleaned. This may leave the src
folder in case of interruption/issue when deleting
but the destination folder will still be written in an atomic way.
The rename syscall only work if the source and destination are on the same filesystem
This library's main goal is to provide a simple interface to rename file in a safe and simple way, without having to dive deep into OS specific documentation and limitations behind it. If you want to use the syscall directly I recommend these two libraries which helped the development of Atomik a lot: