You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
docs: add 'Choosing a Protocol from Python' to the server guide, and pin the PG wire {cypher} claim with a test #253
The repo-wide best-practice sweep, round 2, over the server docs. The bindings docs show HTTP with requests as the way to reach a served database from Python, and say nothing about which protocol or client to use. ArcadeDB's own documentation now has a section, "Python: choosing a protocol" (https://docs.arcadedb.com/arcadedb/how-to/connectivity/drivers/python-http#python-choosing-protocol), and our own protocol measurements agree with it on the points below.
Add a short "Choosing a Protocol from Python" section to docs/guide/server.md and link it from docs/index.md and docs/api-access-methods.md: the PostgreSQL wire with the {cypher} prefix is the fastest single-row route (about half the time of a persistent HTTP connection), a persistent HTTP connection is next, upstream's official arcadedb-driver (HTTP, httpx) and arcadedb-driver-grpc are convenient but add client cost (about 3x a lean client on a one-row read), the Python neo4j Bolt driver is slower than a lean HTTP client on one-row reads and much slower on large results unless fetch_size is raised, HTTP is the route for large result sets and bulk loads. Every number is labelled as laptop, relative, one client thread, loopback.
The section says that the wheel's server has no gRPC plugin, so arcadedb-driver-grpc talks to the official server distribution only, and it does not contradict upstream's page: where upstream measured gRPC streaming fastest on very large results with a big batch size, the section says so and says that at the default batch size our own 10,000-row read was slower than HTTP.
A new test pins the one claim the wheel can check itself: a {cypher} statement with a bound parameter over the Postgres wire returns the right row.
Activity
docs: add "Choosing a Protocol from Python" to the server guide and p…