Tutorials

At the end you will have chained multiple datagov commands into two real, end-to-end recipes — distinct from Getting Started's atomic, single-command paths. Only commands that exist today are used here.

Check a dataset for PII before sharing it

You have a CSV you're about to hand to someone outside your team. First, see its shape without looking at raw values:

datagov inspect examples/pii-fixture.csv --output json

The envelope's dataset.row_count and dataset.column_count confirm this is a small, 10-row, 10-column file — cheap to scan in full. Now run the actual PII scan on the same file:

datagov pii scan examples/pii-fixture.csv

Every finding is entity-typed and confidence-scored, with masked evidence only — record_uuid, ssn, credit_card_number, ipv4_address, and more all surface as findings without ever printing a raw value. If you want the check to fail loudly in a script or CI job rather than just reporting, add a threshold:

datagov pii scan examples/pii-fixture.csv --fail-on 0.8
echo "exit code: $?"

Against this fixture, --fail-on 0.8 exits 12 (mac_address and credit_card_number findings score at or above 0.8) — a scriptable gate: "don't share this file until these findings are resolved."

Compare SQL across two dialects for a migration

You're migrating a query from SQL Server (T-SQL) to Postgres and want to understand what changes before you commit to it. Start by parsing the source under its real dialect, to confirm datagov understands the construct you're migrating:

cat examples/sql/transpile/tsql_to_postgres/source.sql

SELECT t.state, COUNT(*) AS total FROM customers t INNER JOIN orders o ON t.id = o.customer_id GROUP BY t.state ORDER BY t.state;

datagov sql parse examples/sql/transpile/tsql_to_postgres/source.sql --dialect tsql --output json

The AST under extensions.sql_parse.ast confirms the statement type, tables (customers t, orders o), the join, and the grouping — before you transform anything. Now transpile it:

datagov sql transpile examples/sql/transpile/tsql_to_postgres/source.sql --from tsql --to postgres

SELECT t.state, COUNT(*) AS total FROM customers AS t INNER JOIN orders AS o ON t.id = o.customer_id GROUP BY t.state ORDER BY t.state

The rewrite adds explicit AS before table aliases — implicit aliasing (customers t) is valid T-SQL but Postgres's canonical form spells out AS. Confirm the target parses cleanly under its own dialect too:

datagov sql parse examples/sql/transpile/tsql_to_postgres/expected.sql --dialect postgres

No errors, exit 0 — the transpiled query is valid Postgres. For a construct with no equivalent in the target dialect (Snowflake's QUALIFY into ANSI, for example), transpile prints an explicit warning instead of silently producing SQL that won't run — see Commands for that example.

Next: Use Cases for the broader scenarios these commands address.