clickhouse-guidelisted

# ClickHouse Tools: `schema`, `query`, `explain`, and `execute` when writes are allowed. ClickHouse is a column store built for aggregation over very large tables: shape your queries accordingly. ## Aggregate, do not scan - Selecting raw rows from a billion-row table is the wrong tool. Use `count()`, `sum()`, `uniq()`, `quantile()` and `GROUP BY` with a time bucket (`toStartOfHour`, `toDate`). - Filter on the table's primary key or partition key first; `schema` shows the engine, sort key and partition key of a table. - Rows returned are capped (200 by default); the cap is on rows, not on the data scanned, so a bad query can still take the full statement timeout. ## Readonly By default the session is `readonly=1`: only SELECT-like statements. `INSERT`, `ALTER`, `DROP` are refused by the server. `execute` appears only when the user turns read-only off. ## Useful specifics - `explain` shows the query plan; `EXPLAIN ESTIMATE` in a `query` shows how many parts and rows would be read, which is the number to look at before a heavy query. - Dates are UTC unless the column has a time zone. - `FINAL` is expensive; use it only when the user needs the deduplicated view of a ReplacingMergeTree.