Files
Bogdan DrutuandGitHub 8a94f0eb89 Clarify that data cannot be accessed after Consume* func is called (#12284)
The problem here is that we have components like batchprocessor or
others which are keeping a reference of the data and eventually
modifying them without any concurrent access protection. Hence, it is
required for the callers to not access any data that were passed
downstream.

This is not a backwards incompatible change, since pdata is not
thread-safe this is a MUST to have here (consider a bug for not being
documented) otherwise things will not work in the processing pipelines.

Signed-off-by: Bogdan Drutu <[email protected]>
2025-02-06 16:50:29 +00:00

45 lines
1.3 KiB
Go

// Copyright The OpenTelemetry Authors
// SPDX-License-Identifier: Apache-2.0
package consumer // import "go.opentelemetry.io/collector/consumer"
import (
"context"
"go.opentelemetry.io/collector/consumer/internal"
"go.opentelemetry.io/collector/pdata/plog"
)
// Logs is an interface that receives plog.Logs, processes it
// as needed, and sends it to the next processing node if any or to the destination.
type Logs interface {
internal.BaseConsumer
// ConsumeLogs processes the logs. After the function returns, the logs are no longer accessible,
// and accessing them is considered undefined behavior.
ConsumeLogs(ctx context.Context, ld plog.Logs) error
}
// ConsumeLogsFunc is a helper function that is similar to ConsumeLogs.
type ConsumeLogsFunc func(ctx context.Context, ld plog.Logs) error
// ConsumeLogs calls f(ctx, ld).
func (f ConsumeLogsFunc) ConsumeLogs(ctx context.Context, ld plog.Logs) error {
return f(ctx, ld)
}
type baseLogs struct {
*internal.BaseImpl
ConsumeLogsFunc
}
// NewLogs returns a Logs configured with the provided options.
func NewLogs(consume ConsumeLogsFunc, options ...Option) (Logs, error) {
if consume == nil {
return nil, errNilFunc
}
return &baseLogs{
BaseImpl: internal.NewBaseImpl(options...),
ConsumeLogsFunc: consume,
}, nil
}