BigQuery

ObjectExplorer shows Google BigQuery in the tree as projects, datasets and tables, and reads it at the least cost there is. BigQuery bills the bytes a query scans, not the rows it returns, while its metadata and a table's stored rows are free to read. So browsing and opening a table cost nothing, and a query is priced before it runs and runs only under a limit you set.

Connecting

BigQuery uses the Google sign-in, the same one Google Cloud Storage uses: Application Default Credentials, Sign in with Google through gcloud, or a service account key, in Settings → Accounts → Google. The pane ticks Google Cloud Storage and Google BigQuery apart, so either can be turned off.

Settings → Locations → BigQuery then lists the projects your account may use, each with a tick; ticked is in the tree. A project you may use but that is not listed, such as bigquery-public-data, is added by its id. Queries run in names the project a query is billed to; left empty, it is gcloud's project, else the one the credential belongs to. The same pane says what this month's queries billed, and about how much that cost.

In the tree

bq://                            the projects
bq://my-project                  its datasets
bq://my-project/sales            its tables and views
bq://my-project/sales/orders     a table

A table, a snapshot or a clone shows a table icon; a view, a materialized view or an external table a view icon. Datasets are not watched for changes: Refresh in the right-click menu lists them again.

Opening a table

A table opens on a line of its facts — its type, how many rows, how much it stores, its location, when it last changed, and what it is partitioned and clustered by — over its first 10,000 rows. Read 10,000 More reads the next ones. Those rows are BigQuery's own listing of a table's stored rows, which is free.

The rows read are kept on this machine with the time the table last changed. Opened again, the table shows them at once, and asks BigQuery only whether it changed since; Refresh reads them again.

A view, a materialized view or an external table keeps no rows of its own: only a query reads it. It opens on its SQL, with Query It.

A query

The Query tab starts on SELECT * FROM project.dataset.table LIMIT 1000. As you type, each statement is checked by a dry run, which is free, and the line under it says what it would scan, what it would bill and about how much that costs, at BigQuery's on-demand price of $6.25 a TiB (a region's price may differ). BigQuery bills at least 10 MB for each table a query reads, and the line says so when that minimum is what you pay.

A query runs only under its Limit, the most it may bill, 1 GiB unless you change it (about $0.006). The limit goes with the query to BigQuery as well, so a query that would bill more never runs: Run stays gray, and the line says to raise the limit or scan less.

Its answer shows the first 10,000 rows, what it billed and its job's id. The answer is kept on this machine, by its SQL, the project it ran in, and when each table it read last changed. Asked again while those tables are unchanged, it comes from the disk: free, with no job at all. BigQuery's own cache, when it answers, is free too, and the line says which it was.

Where it is kept

Everything BigQuery sent is in ~/.objectexplorer/cache/bigquery: the rows of the tables you opened, the answers of the queries you ran, and a line per query job with what it billed, which is what the month's total in Settings adds up. Clear Cache in Settings → Cache removes it with the rest of the cache, and the month's total starts again from nothing.

Next: Azure Blob Storage.