Connect Google BigQuery to Excel, Sheets and AI

Nested data you can actually see

BigQuery's strength is also what makes it awkward to browse: records nest, arrays repeat, and half the tools that connect to it show you a column called payload and nothing more.

• Nested records are flattened in the catalog as parent.child, so the Query Builder shows the fields inside a STRUCT rather than the STRUCT itself.
• GEOGRAPHY values come through as their text form instead of an opaque object.
• INTERVAL columns are handled explicitly — the Google client library throws on them, and rather than let that take a query down, the raw value is read instead.
• Repeated and record columns are left out of value-frequency profiling, because asking for the distinct values of an array is not a question BigQuery answers.

A dataset is what other engines call a database

The vocabulary mismatch is worth stating plainly, because it decides what you type. The connection binds a project and a default dataset, and it is the dataset name that appears in the middle of a cross-source reference.

• Within a single BigQuery connection, write table names unqualified — the job already has your default dataset set. Never write the project.
• Across sources, the middle part is the dataset.
• One BigQuery connection reads one project. Joining two projects is a federated query across two connections, not a cross-database join inside one.

-- bq_analytics is the connection, events is the BigQuery dataset
SELECT   s.utm_source, COUNT(*) AS sessions
FROM     bq_analytics.events.sessions  s
JOIN     pg_crm.public.customers       c ON c.id = s.customer_id
WHERE    c.plan = 'enterprise'
GROUP BY s.utm_source