message_convert_plugin.h 文件参考

Plugin interface for VLink/webviz message conversion across visualization backends. 更多...

#include <string>
#include "../base/bytes.h"
#include "../base/plugin.h"
#include "../impl/types.h"

message_convert_plugin.h 的引用(Include)关系图:

浏览该文件的源代码.

类
struct	vlink::WebChannel
	Describes a frontend-advertised publish channel. 更多...
struct	vlink::VlinkPublish
	VLink publish destination resolved for an inbound frontend message. 更多...
class	vlink::MessageConvertPlugin
	Abstract interface for VLink/webviz message conversion plugins supporting multiple visualization backends. 更多...

命名空间
namespace	vlink

枚举
enum class	vlink::ConvertTarget : uint8_t { vlink::kFoxglove = 0 , vlink::kRerun = 1 }
	Identifies the visualization backend that the plugin is converting for. 更多...

详细描述

Plugin interface for VLink/webviz message conversion across visualization backends.

MessageConvertPlugin allows users to implement custom VLink/webviz message conversion logic in a shared library plugin. The plugin receives raw VLink message bytes (Protobuf, FlatBuffers, CDR, POD, or any custom serialisation) and produces a backend-specific payload along with schema/type metadata.

This interface has zero third-party dependencies – it only uses VLink base types (Bytes) and standard C++ types, making it easy to implement in external projects without linking against Protobuf, FlatBuffers, Rerun SDK, or JSON libraries.

A single plugin can support multiple visualization backends by checking the ConvertTarget parameter in each method. The plugin coexists with JSON-based VLink-to-webviz mapping files. When both are present, the plugin is tried first; if it returns false for a given type, the JSON mapping pipeline takes over.

Supported targets

Target	Payload format	type_name meaning
kFoxglove	FlatBuffer / Protobuf binary	Foxglove schema name
kRerun	JSON string describing components	Rerun archetype name

Rerun JSON payload format

When targeting Rerun, the plugin should produce a UTF-8 JSON string as the payload. The JSON object describes the Rerun archetype components. Each archetype has its own expected fields:

// Points3D example:
{
  "positions": [[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],
  "colors": [[255, 0, 0, 255], [0, 255, 0, 255]],
  "radii": [0.1, 0.2]
}
 
// EncodedImage example (binary data is base64-encoded):
{
  "media_type": "image/jpeg",
  "data_base64": "<base64-encoded image bytes>"
}
 
// GeoPoints example:
{
  "lat_deg": [37.7749, 37.7750],
  "lon_deg": [-122.4194, -122.4195]
}
 
// TextLog example:
{
  "text": "Hello world",
  "level": "INFO"
}
 
// Scalars example:
{
  "value": 3.14
}
 
// Transform3D example:
{
  "translation": [1.0, 2.0, 3.0],
  "rotation_quat": [0.0, 0.0, 0.0, 1.0]
}
 
// Boxes3D example:
{
  "half_sizes": [[1.0, 2.0, 3.0]],
  "centers": [[0.0, 0.0, 0.0]],
  "quaternions": [[0.0, 0.0, 0.0, 1.0]],
  "colors": [[255, 0, 0, 255]],
  "labels": ["box1"]
}
 
// Pinhole example:
{
  "image_from_camera": [[fx, 0, cx], [0, fy, cy], [0, 0, 1]],
  "resolution": [1920, 1080]
}

Binary archetypes should carry their raw bytes in a data_base64 string. Currently the built-in Rerun JSON bridge supports this for EncodedImage, Image, DepthImage, SegmentationImage, EncodedDepthImage, Asset3D, AssetVideo, and Tensor. Image, DepthImage, and SegmentationImage also require width/height (or resolution). Tensor additionally requires shape and may provide dim_names. Direct VLink-to-Rerun mappings still cover a broader set of archetypes than the plugin JSON bridge.

Plugin lifecycle

1. init() is called once after loading, with an optional config string.
2. can_convert() is called for each discovered VLink serialisation type (with the target backend) to determine if the plugin handles it.
3. get_schema_info() is called once per type to register the channel/archetype.
4. convert() is called for every incoming message on matched types.
5. The plugin is destroyed when the server shuts down.

Example implementation (supports both Foxglove and Rerun)

#include <vlink/extension/message_convert_plugin.h>
 
class MyConvertPlugin : public vlink::MessageConvertPlugin {
  VLINK_PLUGIN_REGISTER(MessageConvertPlugin)
 
 public:
  bool init(const std::string& config) override {
    // Parse config, load schemas, etc.
    return true;
  }
 
  bool can_convert(const std::string& vlink_ser, ConvertTarget target) override {
    if (vlink_ser != "my_pkg.MyMessage") return false;
    // Support both backends
    return target == ConvertTarget::kFoxglove || target == ConvertTarget::kRerun;
  }
 
  bool get_schema_info(const std::string& vlink_ser, ConvertTarget target,
                       std::string& type_name, std::string& encoding,
                       std::string& schema_encoding,
                       std::string& schema_data) override {
    if (target == ConvertTarget::kFoxglove) {
      type_name = "foxglove.LocationFix";
      encoding = "flatbuffers";
      schema_encoding = "flatbuffers";
      // schema_data = binary FBS schema bytes
    } else if (target == ConvertTarget::kRerun) {
      type_name = "GeoPoints";
      encoding = "json";
      // schema_encoding and schema_data are unused for Rerun
    }
    return true;
  }
 
  bool convert(const std::string& vlink_ser, const vlink::Bytes& raw,
               ConvertTarget target, vlink::Bytes& payload) override {
    if (target == ConvertTarget::kFoxglove) {
      // Build Foxglove FlatBuffer payload
    } else if (target == ConvertTarget::kRerun) {
      // Build JSON: {"lat_deg":[37.77], "lon_deg":[-122.41]}
      std::string json = R"({"lat_deg":[37.77],"lon_deg":[-122.41]})";
      payload = vlink::Bytes::deep_copy(json.data(), json.size());
    }
    return true;
  }
};
VLINK_PLUGIN_DECLARE(MyConvertPlugin, 4, 0)