redshift-grep-columns

Redshift Cross-Table Column Grep

Find every column whose name or comment matches a keyword across all tables in one or all schemas. Single MCP call per schema via search_columns(table_name=None).

When to use / NOT

• Use to find FK / shared-key columns across many tables (typical pre-JOIN reconnaissance).
• Use to audit naming consistency ("are we calling this customer_id or cust_no?").
• NOT for a single known table — use MCP search_columns(kw, schema, table) directly.
• NOT for single-column lookup — use MCP get_column_comment.
• NOT for table-name search — use /redshift-grep-tables.

Inputs

Form	Behavior
<keyword>	grep across all schemas in the cluster
<keyword> --schema <name>	scope to one schema
<keyword> --schema <s1>,<s2>	scope to listed schemas
<kw1> <kw2>	space-separated keywords (OR logic on MCP side)
--max-list N	results per page (default 50)

Flow

Step 1 — schema-wide search via MCP

The MCP server's search_columns accepts an optional table_name; omit it for schema-wide column search in one call:

search_columns(keywords, schema_name=<schema>, table_name=None)

This returns one row per matching column with table_name included, ranked by hit_count across name + comment. Paginate via has_more if total > 50.

Procedure:

• If --schema given: one call per listed schema.
• Else: list_schemas(include_comments=true) → enumerate user schemas; one search_columns per schema (schema-wide each).
• Aggregate.

Latency: ~0.7s per schema-wide call (12K-column schemas). For clusters with many schemas, prompt user to scope via --schema before running.

Step 2 — render

Group by <schema>.<table>, comment-first. One blank line between groups:

Matches for "customer_id" (12 hits in 3 schemas, 8 tables):

dbt_marts.fct_orders
  customer_id   bigint   Customer reference (FK to dim_users.id)

dbt_marts.fct_returns
  customer_id   bigint   Customer who initiated the return

dbt_staging.stg_users
  id            bigint   Customer's primary key
  cust_no       bigint   Legacy customer no, alias for id