You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
doc-gen: Migrate builtin scalar functions from code based documentation to attribute based #13671
The translation can be done manually or in semi automatic way, uncommenting a function in PR #13669 locally. This will generate local text files with attribute text which can be copy pasted to structs. However its still needed to remove code based documentation manually, add imports, etc
Please refer to PR #13646 which completed the migration for aggregation functions
changed the title [-]doc-gen: Migrate builtin functions from code based documentation to attribute based[/-][+]doc-gen: Migrate builtin scalar functions from code based documentation to attribute based[/+]on Dec 6, 2024
Activity
[-]doc-gen: Migrate builtin functions from code based documentation to attribute based[/-][+]doc-gen: Migrate builtin scalar functions from code based documentation to attribute based[/+]Chen-Yuan-Lai commentedon Dec 9, 2024
take
Chen-Yuan-Lai commentedon Dec 27, 2024
Hi @alamb @Omega359 @comphead I've split the changes into several smaller PRs based on function categories. Could you please review these PRs? Thx 🙏
BTW, I put
Cargo.tomlandCargo.lockchanges in #13928 , this PR may need to be merged first than other array functions PRsOmega359 commentedon Dec 27, 2024
Thanks for all your work on this! I'll do my best to start reviewing these PR's this weekend if no one beats me to it.
alamb commentedon Dec 28, 2024
Thank you @Chen-Yuan-Lai and @Omega359 for driving this forward
comphead commentedon Dec 28, 2024
Supposedly it was closed automatically, I think we can close it once all the PRs merged.
comphead commentedon Jan 16, 2025
Okay, I think some still left
Chen-Yuan-Lai commentedon Jan 17, 2025
@comphead I'll complete these ignored functions
alamb commentedon Jan 17, 2025
Nice work everyone and thank you @Chen-Yuan-Lai especially 🚀