Verisyntax — User Manual (prototype 0.1)

Use imported data

A dataset is a read-only result definition:

dataset paid_orders:
  from orders as order_item
  select order_item.id as order_id
  select order_item.total as total
  where order_item.status = "paid"

To combine two tables:

dataset customer_orders:
  from customers as customer
  inner join orders as order_item on customer.id = order_item.customer_id
  select customer.name as customer_name
  select order_item.total as total

Use left join instead of inner join when all rows from the first source must remain, even without a match. The compiler checks source names, aliases, columns, backend compatibility, and join-key type/unit/period compatibility. The standalone desktop composer supports ordered chains across two to eight same-database sources; every additional source must join to an alias already introduced.

To run a dataset in the standalone IDE:

1. Save and Check the source.
2. Click Compile & build; compilation and linking run automatically.
3. Select the dataset in the run menu.
4. Optionally select Result file… and a database audit destination.
5. Click Run and approve the exact executable hash.

The IDE automatically opens a large result dialog. It says Run completed when numerical diagnostics are acceptable and Run completed with warnings when R-hat or ESS requires review. Logic answers, dataset rows, or Bayesian posterior cards appear first. Run documentation and raw JSON stay collapsed under Technical details and raw output. Closing the dialog does not discard anything: the complete audit text remains in Results, and Open result view reopens the presentation.

Each main Bayesian metric has Explain + / Close −, providing an inline explanation of posterior mean, HDI, R-hat, or effective samples without opening raw output. A separate Explain sampling settings and acceptance rates section explains chains, retained samples, seed, warmup adaptation, and acceptance before and after warmup.