Fix decoding issue (#158) (#159)

Attributes with empty keys were already ignored, but not attributes with
no value. In this PR, I consider an attribute with no value as malformed
or invalid, so they are ignored.
Let me know if this is not the behavior you expected.

Closes: #158

go.mod

@@ -57,14 +57,3 @@ require (
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20240221002015-b0ce06bbee7c // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 )
-
-
-
-
-
-
-
-
-
-
-

pkg/benchmark/compression.go

@@ -23,6 +23,7 @@ import (
 )
 
 const CompressionTypeZstd = "zstd"
+const CompressionTypeNone = ""
 
 type CompressionAlgorithm interface {
 	fmt.Stringer

pkg/benchmark/dataset/real_logs_dataset.go

@@ -26,6 +26,7 @@ import (
 	"go.opentelemetry.io/collector/pdata/plog/plogotlp"
 
 	"github.com/klauspost/compress/zstd"
+
 	"github.com/open-telemetry/otel-arrow/pkg/benchmark"
 	"github.com/open-telemetry/otel-arrow/pkg/benchmark/stats"
 )

pkg/otel/common/arrow/attributes.go

@@ -168,10 +168,11 @@ func (b *AttributesBuilder) Append(attrs pcommon.Map) error {
 	return b.builder.Append(attrs.Len(), func() error {
 		var err error
 		attrs.Range(func(key string, v pcommon.Value) bool {
-			if key == "" {
-				// Skip entries with empty keys
+			// Attribute without key or without value are ignored.
+			if key == "" || v.Type() == pcommon.ValueTypeEmpty {
 				return true
 			}
+
 			b.kb.AppendNonEmpty(key)
 			return b.ib.Append(&v) == nil
 		})
@@ -236,8 +237,8 @@ func (c *Attributes16Accumulator) AppendWithID(parentID uint16, attrs pcommon.Ma
 	}
 
 	attrs.Range(func(key string, v pcommon.Value) bool {
-		if key == "" {
-			// Skip entries with empty keys
+		// Attribute without key or without value are ignored.
+		if key == "" || v.Type() == pcommon.ValueTypeEmpty {
 			return true
 		}
 
@@ -290,8 +291,8 @@ func (c *Attributes32Accumulator) Append(ID uint32, attrs pcommon.Map) error {
 	}
 
 	attrs.Range(func(key string, v pcommon.Value) bool {
-		if key == "" {
-			// Skip entries with empty keys
+		// Attribute without key or without value are ignored.
+		if key == "" || v.Type() == pcommon.ValueTypeEmpty {
 			return true
 		}
 

tools/logs_benchmark/main.go

@@ -81,6 +81,7 @@ func main() {
 	// Compare the performance for each input file
 	for i := range inputFiles {
 		var ds dataset.LogsDataset
+		var compression = benchmark.CompressionTypeZstd
 
 		inputFile := inputFiles[i]
 		compressionAlgo := benchmark.Zstd()
@@ -93,6 +94,7 @@ func main() {
 		// in case formatFlag was not passed
 		if strings.HasSuffix(inputFile, ".json") {
 			*formatFlag = "json"
+			compression = benchmark.CompressionTypeNone
 		} else if strings.HasSuffix(inputFile, ".pb") {
 			*formatFlag = "proto"
 		}
@@ -101,7 +103,7 @@ func main() {
 		if strings.HasSuffix(inputFile, ".csv") {
 			ds = CsvToLogsDataset(inputFile)
 		} else {
-			rds := dataset.NewRealLogsDataset(inputFiles[i], benchmark.CompressionTypeZstd, *formatFlag)
+			rds := dataset.NewRealLogsDataset(inputFiles[i], compression, *formatFlag)
 			//rds.Resize(10)
 			ds = rds
 		}