docs(skills): fix stale datasets create flags and add --no-input#153
Conversation
- Replace non-existent --label/--table-name flags with --name/--description in hotdata and hotdata-analytics WORKFLOWS.md chain examples - Fix dataset upload workflow: remove --file/--url (not supported on datasets create; parquet uploads belong in managed databases) - Document --no-input global flag in hotdata/SKILL.md
There was a problem hiding this comment.
Documentation fix verified against the CLI source:
datasets createflags (--name/--description/--sql/--query-id) matchcommand.rs:441-465; the removed--label/--table-name/--file/--urlflags genuinely don't exist.--no-inputglobal flag confirmed atmain.rs:52.- Parquet → managed databases redirect is correct (
databases tables load --file).
Accurate and consistent across all three files.
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
- Fix hotdata-analytics WORKFLOWS: datasets source is SQL/query-id only, not CSV/JSON/URL (remove stale description) - Fix hotdata WORKFLOWS Chain epic: databases create/load now uses --catalog and --table flags, not --name and imprecise load syntax
There was a problem hiding this comment.
Verified all documented flags against the CLI source (src/command.rs, src/main.rs, src/util.rs). The corrected datasets create --name/--description/--sql/--query-id, the databases create/load --catalog/--table/--file examples, and the new --no-input global flag all match the actual command definitions. Accurate documentation fix.
- Datasets only support --sql and --query-id (not CSV/JSON/URL/file) - Correct SQL prefix: <catalog>.<schema>.<table> not 'connection name' - Correct managed DB CLI: databases load not databases tables load - Fix rule of thumb: parquet files -> databases, SQL snapshot -> datasets - Fix skill decision table: remove misleading CSV/JSON/URL/stdin label
| | **Refresh upstream** | `datasets refresh` (URL/query sources) | Replace via `tables load` again | | ||
| | **Best for** | SQL or saved-query snapshot | Parquet files you own; catalog-style `alias.schema.table` | | ||
| | **SQL prefix** | `datasets.<schema>.<table>` (often `datasets.main.*`) | `<catalog>.<schema>.<table>` where catalog = `--catalog` alias | | ||
| | **CLI** | `hotdata datasets create --sql “…”` | `hotdata databases create --catalog` + `databases load` | |
There was a problem hiding this comment.
super nit: curly quotes “…” inside this inline-code command read as literal characters and aren't valid shell quoting if copied. Straight quotes match the rest of the command examples in this doc. (not blocking)
| | **CLI** | `hotdata datasets create --sql “…”` | `hotdata databases create --catalog` + `databases load` | | |
| | **CLI** | `hotdata datasets create --sql "…"` | `hotdata databases create --catalog` + `databases load` | |
There was a problem hiding this comment.
Verified all documented CLI flags against the actual command definitions in src/command.rs and src/main.rs — every change is accurate:
datasets create --name/--description/--sql/--query-idmatches the source; the removed--label/--table-name/--file/--urlgenuinely don't exist ondatasets create.databases create --catalog/--tableanddatabases load --catalog/--table/--file/--url/--upload-idmatch the source.--no-inputglobal flag exists (main.rs:52) and the documented error-in-non-TTY/CI behavior matchesutil.rs.
One non-blocking super nit left inline. Approving.
create only accepts --catalog (managed DB) or --dataset-id. list and delete accept --connection-id; create does not. Pre-existing error in hotdata-search/SKILL.md.
There was a problem hiding this comment.
Docs-only PR. Verified all updated flags against src/command.rs and src/main.rs: datasets create --name/--description/--sql/--query-id, databases create --catalog/--table, databases load --catalog/--table/--file/--url/--upload-id, the new --no-input global flag, and the indexes create catalog/dataset-only scopes all match the implementation. The curly-quote super nit on WORKFLOWS.md:89 is already tracked in an existing thread.
Summary
--label/--table-namewith--name/--descriptionin chain examples across hotdata and hotdata-analytics WORKFLOWS.mddatasets createhas no--file/--urlsource; parquet uploads belong in managed databases--no-inputglobal flag tohotdata/SKILL.md(added in 0.4.x, was undocumented)Files changed
skills/hotdata/SKILL.md— add--no-inputto global CLI optionsskills/hotdata/references/WORKFLOWS.md— fix Chain epic flow and dataset upload workflowskills/hotdata-analytics/references/WORKFLOWS.md— fix Chain materialize examples and naming tip