Performance improvement for os.scandir on Windows
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 77.2k
- Forks
- 35.9k
- PR merge metrics
- PR metrics pending
Description
Feature or enhancement
Proposal:
As has been mentioned in numerous issues (See #119169 for example) , there are quite a few performance issues regarding os.scandir and os.walk.
Diving into the implementation - see os_scandir_impl in posixmodule.c - I noticed that we are currently using WinAPI to list a directory. Looking into the relevant WinAPI functions (FindNextFileW, FindFirstFileW) it seems redundant to implement our Python wrapper around these wrappers. I propose to use the native NT functions - for example NtQueryDirectoryFile - directly, as we are already implementing a wrapper ourselves.
After quite a bit of reading MS-Docs and looking at Kernel32.dll & NtDll.dll, I have reached a stable implementation of this proposal. Unfortunately, I could not significantly reduce the amount of syscalls being performed, however I did reduce the amount of memory allocation and copying by a ratio of 1:6.
Additionally, this new implementation should aid in future implementations around Windows file system operations (See #99454 for any easy example).
This improves the performance of both os.scandir and os.walk which is implemented over it.
Has this already been discussed elsewhere?
This is a minor feature, which does not need previous discussion elsewhere
Links to previous discussion of this feature:
No response
Linked PRs
- gh-122886
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with os_scandir_impl in Modules/posixmodule.c and review the linked PR gh-122886. Compare the Windows directory-listing implementation with the relevant WinAPI and NT functions described in the issue, then benchmark os.scandir and os.walk. Done means the Windows implementation improves allocation and copying performance without changing behavior.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- operating-systems
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 25/100