Skip to content

Latest commit

 

History

History
132 lines (96 loc) · 4.22 KB

README.md

File metadata and controls

132 lines (96 loc) · 4.22 KB
Raw

greptime-proto

GreptimeDB protobuf files.

Build

Requirement

Command

  • Compile for Rust

    make rust

  • Compile for Go

    make go

  • Compile for Java

    make java

    The compilation for Go and Java will use builder container namely/protoc-all.

Usage

Rust

# Add this repository as dependency to your Cargo.toml file:
greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git" }
// To use the GreptimeDB's gRPC service:
use greptime_proto::v1::*;

// To talk to GreptimeDB Meta service:
use greptime_proto::v1::meta::*;

// To request GreptimeDB as Prometheus remote read/write:
use greptime_proto::prometheus::remote::*;

Go

Download the go module:

go get github.com/GreptimeTeam/greptime-proto@main

Then use greptime-proto as the normal Go module:

import (
        greptimev1 "github.com/GreptimeTeam/greptime-proto/go/greptime/v1"
)
...

For SDK developers

GreptimeDB's gRPC service is built on top of Arrow Flight RPC. You can find the Arrow's official implementation status of each programming language flight rpc.

If you can't find the language you are using, you can always generate the Arrow Flight gRPC service from the raw protobuf definition flight protobuf definitions. Or call into other language binding like C++.

Once the Arrow Flight client is ready, you only need to care about the following protobuf files to accomplish our SDK writing:

.
├── greptime
│   └── v1
│       ├── column.proto
│       ├── common.proto
│       ├── database.proto
│       ├── ddl.proto
│       ├── health.proto
│       └── prom.proto

You can find all protobuf files in the directory "proto" under the project's root.

Right now the GreptimeDB only responds to Arrow Flight's do_get interface. All the reads and writes (that are from clients) are handled there. do_get needs a "ticket" as the input request, you need to serialize the GreptimeRequest message defined in "database.proto" for it.

There are 3 kinds of GreptimeRequest, which are:

  • InsertRequest, carries the data to be ingested. It's a little verbose to assemble, especially the "column"s that define the schema and value of the input data. You can find the definition of "column" in "column.proto".
  • QueryRequest has the SQL in it. Note that you can INSERT INTO GreptimeDB as well as SELECT it.
  • DdlRequest defines the "Data Definition Language" request, like table creation or deletion. It's sometimes more representative than the normal SQL. DdlRequests are defined in "ddl.proto".

There's also RequestHeader in the GreptimeRequest to specify the "catalog" and "schema" to be used in this request. If either one is not set (or left empty), GreptimeDB will use the default catalog "greptime" and default schema "public".

The response of do_get is handled in Arrow Flight's client. It's a stream of FlightData. You can find its definition in Arrow Flight's protobuf file. Special care must be taken when dealing with insertion, that GreptimeDB puts insertion result in FlightData's metadata. Insertion results, either from InsertRequest or INSERT INTO SQL, are both have the same format, that "Affected Rows: x". "x" represents the rows that are successfully inserted. When dealing with this special FlightData, please ignore its data_body field, but directly strip the app_metadata field from it, and deserialize the bytes as FlightMetadata message. You will find the "affected rows" result in it.

We already have our SDK written in Java, Rust and Go, feel free to take any of them as an example.