mbutrovich commented on code in PR #4349: URL: https://github.com/apache/datafusion-comet/pull/4349#discussion_r3269082439
########## .claude/skills/wire-datafusion-function/SKILL.md: ########## @@ -72,24 +72,50 @@ Note `inputTypes`, `dataType`, `eval` / `nullSafeEval`, ANSI mode branches, `req ### 2. Find the upstream function -Read the pinned `datafusion-spark` version from `native/Cargo.toml` rather than picking the first cached copy — `head -1` can land on an older release that lacks the function: +**Prefer a local DataFusion clone if one is available** — check `CLAUDE.md`, project memory, and the user's `~/dev` tree for an existing checkout (e.g. `~/dev/.../rust/datafusion`). The local clone is typically the latest source and avoids cargo-registry version mismatches. If no local clone is found, fall back to the cached registry crate. Review Comment: This seems overly specific to your setup unless we're saying in the contributor guide that people should be working in a `~/dev` folder. I suspect users should be telling Claude on their own where their working directories are. Also, is this general advice even true? Comet lags behind, and I could easily see an agent trying to wire something up that exists in upstream but isn't on Comet's DF version yet. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
