Repository navigation
[RFC]: add missing single precision nan modules for stats/strided #13955
Description
Description
This RFC proposes to add missing single precision nan modules for stats/strided, which include :
- snanmeanpw
- snanmeankbn
- snanmeankbn2
- snanmeanli
- snanmeanlipw
- snanmeanstdev
- snanmeanstdevpn
- snanmeanvar
- snanmeanvarpn
- sdsnanmean
- sdsnanmeanpn
- sdsnanmeanwd
- sdsnanmeanpw
- snanvariance
- snanvariancech
- snanvariancepn
- snanvariancetk
- snanvariancewd
- snanvarianceyc
- snanstdev
- snanstdevch
- snanstdevpn
- snanstdevtk
- snanstdevwd
- snanstdevyc
- snansem
- snansemch
- snansempn
- snansemtk
- snansemwd
- snansemyc
- snanrangeabs
- snanmidrangeabs
- snancumax
- snancumaxabs
- snancumin
- snancuminabs
NOTE: Priority is top to bottom.
Related Issues
None
Questions
- How many packages do we really need ? ( I have already eliminate those packages which not useful or not feasible )
- What should be the priority of all the modules mentioned here?
Other
No.
Checklist
- I have read and understood the Code of Conduct.
- Searched for existing issues and pull requests.
- The issue name begins with
RFC:.
Activity
0PrashantYadav0 commentedon Aug 5, 2026
For single precision, there are lots of modules which are expected to be present. Currently I am checking all the missing packages and soon create new RFC for them.
[-][RFC]: add missing single precision nan modules for `stats/strided`[/-][+][RFC]: add missing single precision `nan` modules for `stats/strided`[/+]AryanSharma48 commentedon Aug 11, 2026
I have implemented$n = 0$ edge-case handling, and JS bindings for this series.
snanmeankbnas an initial prototype to establish and verify the single-precision C implementation pattern, strided memory logic,Draft PR: #14175
Once maintainers review the structural approach and CI checks pass on this initial PR, I will proceed with the remaining priority modules on the list.
0PrashantYadav0 commentedon Aug 11, 2026
Hi @AryanSharma48 thank you for working on these modules. Looking forward for your contribution. Feel free to tag me if you need any help or review.
AryanSharma48 commentedon Aug 11, 2026
Hi @0PrashantYadav0! The GitHub Actions CI workflows are currently awaiting maintainer approval to run. Whenever you have a moment, could you please trigger/approve the workflow for this PR #14175? Thanks!
0PrashantYadav0 commentedon Aug 11, 2026
There is no need to run checks manually, they will start as soon as you open the PR.
AryanSharma48 commentedon Aug 11, 2026
Got it @0PrashantYadav0 , thanks! I've marked the PR #14175 as "Ready for review".
Since this is my first contribution from a new fork, GitHub automatically holds the runners and displays a "12 workflows awaiting approval from a maintainer" banner at the top of the PR. Whenever you or another maintainer have a moment to click "Approve and run" on the Actions tab, the CI checks will start running.
kgryte commentedon Aug 14, 2026
@0PrashantYadav0 There was a reason many of these packages were held up. Namely, we needed to refactor the various upstream packages, such as
snannsum, etc, to returnstructs, rather than numbers. Just took a look at some of the implementations, which are trying to usesnancount, etc. That is the wrong approach and not what is desired.0PrashantYadav0 commentedon Aug 14, 2026
Thanks @kgryte , understood. The downstream NaN stats modules should wait until upstream
snannsum*primitives return { sum, count } as a struct, rather than usingsnancountor hand-rolled NaN loops. I’ll pause downstream work and focus on upstream once that pattern is defined. If there’s a reference PR/RFC for the struct API, please share it.0PrashantYadav0 commentedon Aug 14, 2026
@kgryte As we adreay
dnannsum*modules. Do we need to refactor the existing modules instats/strided/*accordingly?0PrashantYadav0 commentedon Aug 14, 2026
New modules that needed to be added before them :
Refactors ( not new packages ):
->
struct-return { sum, count }