gibheer
/
bleve
0
0
Fork 0
Code

rewrite to used fixed size []byte instead of buffer 

removes unchecked errors in calls to buffer.Write
and also benchmarks considerably faster
This commit is contained in:
Marty Schoch 2015-03-11 15:12:13 -04:00
parent 50bd082257
commit a44a7c01af
2 changed files with 72 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
index/upside_down/row.go
View File

@ -496,10 +496,10 @@ func (br *BackIndexRow) AllStoredKeys() [][]byte {
}
func (br *BackIndexRow) Key() []byte {
buf := new(bytes.Buffer)
buf.WriteByte('b')
buf.Write(br.doc)
return buf.Bytes()
buf := make([]byte, len(br.doc)+1)
buf[0] = 'b'
copy(buf[1:], br.doc)
return buf
}
func (br *BackIndexRow) Value() []byte {
@ -562,19 +562,18 @@ type StoredRow struct {
}
func (s *StoredRow) Key() []byte {
buf := new(bytes.Buffer)
buf.WriteByte('s')
buf.Write(s.doc)
buf.WriteByte(ByteSeparator)
fieldbuf := make([]byte, 2)
binary.LittleEndian.PutUint16(fieldbuf, s.field)
buf.Write(fieldbuf)
docLen := len(s.doc)
buf := make([]byte, 1+docLen+1+2+(binary.MaxVarintLen64*len(s.arrayPositions)))
buf[0] = 's'
copy(buf[1:], s.doc)
buf[1+docLen] = ByteSeparator
binary.LittleEndian.PutUint16(buf[1+docLen+1:], s.field)
bytesUsed := 1 + docLen + 1 + 2
for _, arrayPosition := range s.arrayPositions {
arrayPositionBuffer := make([]byte, binary.MaxVarintLen64)
numBytes := binary.PutUvarint(arrayPositionBuffer, arrayPosition)
buf.Write(arrayPositionBuffer[0:numBytes])
varbytes := binary.PutUvarint(buf[bytesUsed:], arrayPosition)
bytesUsed += varbytes
}
return buf.Bytes()
return buf[0:bytesUsed]
}
func (s *StoredRow) Value() []byte {
@ -589,11 +588,12 @@ func (s *StoredRow) String() string {
}
func (s *StoredRow) ScanPrefixForDoc() []byte {
buf := new(bytes.Buffer)
buf.WriteByte('s')
buf.Write(s.doc)
buf.WriteByte(ByteSeparator)
return buf.Bytes()
docLen := len(s.doc)
buf := make([]byte, 1+docLen+1)
buf[0] = 's'
copy(buf[1:], s.doc)
buf[1+docLen] = ByteSeparator
return buf
}
func NewStoredRow(doc string, field uint16, arrayPositions []uint64, typ byte, value []byte) *StoredRow {

52
index/upside_down/row_test.go
View File

@ -88,6 +88,11 @@ func TestRows(t *testing.T) {
[]byte{'s', 'b', 'u', 'd', 'w', 'e', 'i', 's', 'e', 'r', ByteSeparator, 0, 0},
[]byte{'t', 'a', 'n', ' ', 'a', 'm', 'e', 'r', 'i', 'c', 'a', 'n', ' ', 'b', 'e', 'e', 'r'},
},
{
NewStoredRow("budweiser", 0, []uint64{2, 294, 3078}, byte('t'), []byte("an american beer")),
[]byte{'s', 'b', 'u', 'd', 'w', 'e', 'i', 's', 'e', 'r', ByteSeparator, 0, 0, 2, 166, 2, 134, 24},
[]byte{'t', 'a', 'n', ' ', 'a', 'm', 'e', 'r', 'i', 'c', 'a', 'n', ' ', 'b', 'e', 'e', 'r'},
},
{
NewInternalRow([]byte("mapping"), []byte(`{"mapping":"json content"}`)),
[]byte{'i', 'm', 'a', 'p', 'p', 'i', 'n', 'g'},
@ -266,3 +271,50 @@ func BenchmarkTermFrequencyRowDecode(b *testing.B) {
NewTermFrequencyRowKV(k, v)
}
}
func BenchmarkBackIndexRowEncode(b *testing.B) {
field := uint32(1)
t1 := "term1"
for i := 0; i < b.N; i++ {
row := NewBackIndexRow("beername",
[]*BackIndexTermEntry{
&BackIndexTermEntry{
Term: &t1,
Field: &field,
},
},
[]*BackIndexStoreEntry{
&BackIndexStoreEntry{
Field: &field,
},
})
row.Key()
row.Value()
}
}
func BenchmarkBackIndexRowDecode(b *testing.B) {
for i := 0; i < b.N; i++ {
k := []byte{0x62, 0x62, 0x65, 0x65, 0x72, 0x6e, 0x61, 0x6d, 0x65}
v := []byte{0x0a, 0x09, 0x0a, 0x05, 0x74, 0x65, 0x72, 0x6d, 0x31, 0x10, 0x01, 0x12, 0x02, 0x08, 0x01}
NewTermFrequencyRowKV(k, v)
}
}
func BenchmarkStoredRowEncode(b *testing.B) {
for i := 0; i < b.N; i++ {
row := NewStoredRow("budweiser", 0, []uint64{}, byte('t'), []byte("an american beer"))
row.Key()
row.Value()
}
}
func BenchmarkStoredRowDecode(b *testing.B) {
for i := 0; i < b.N; i++ {
k := []byte{'s', 'b', 'u', 'd', 'w', 'e', 'i', 's', 'e', 'r', ByteSeparator, 0, 0}
v := []byte{'t', 'a', 'n', ' ', 'a', 'm', 'e', 'r', 'i', 'c', 'a', 'n', ' ', 'b', 'e', 'e', 'r'}
NewTermFrequencyRowKV(k, v)
}
}