TreeSQL UI Console.
- Schema on the left
- TreeSQL editor and completions in the middle
- generated SQL on the top right
- JSON results on the bottom right
A server which provides:
- An HTTP endpoint which executes TreeSQL queries by translating them to SQL queries. It returns results as JSON.
- A simple console UI (pictured) which allows authoring TreeSQL queries (with schema-aware autocomplete), executing them, and seeing results.
A new query language which is halfway between SQL and GraphQL.
It's like SQL, except you write joins by starting with one table, and nesting the tables you want to join to within curly braces, forming a tree of joins. Results come back as a tree of JSON mirroring the query try.
It's like GraphQL, except that it has a direct mapping onto a DB schema (i.e. there's no arbitrary code between the query and the data storage, as there is in GraphQL servers) and it has a more SQL-like syntax (WHERE clauses, etc).
GraphQL gives developers the tree-structured queries and results they want, but is backend-agnostic. Developers struggle to provide efficient GraphQL APIs on top of existing databases, whether they're relational or NoSQL.
SQL is implemented by many sophisticated database systems, but doesn't allow developers to directly express the queries they need to fetch the data to render their UIs. ORMs attempt to bridge this gap, but have clumsy support for efficiently loading graphs of objects across multiple tables.
TreeSQL attempts to bridge this gap, by offering tree-structured queries in a way that maps directly to a conventional relational DB schema. This prototype executes queries by translating them to SQL, but TreeSQL could potentially be executed directly by the database engine -- either incorporating it into the SQL grammar, or as a separate syntax altogether.
By translating it to a correlated subquery, which uses Postgres/Cockroach's builtin JSON functions to combine all results into a single JSON datum -- a one-row, one-column result set.
A lot of things.
GROUP BY(unclear what the syntax should be for this)ORDER BY- Joining from one to one (only supports one to many)
- etc
- EdgeDB's EdgeQL is another attempt to find a hybrid syntax between GraphQL and SQL
- Hasura generates correlated subqueries from GraphQL queries, I believe