/// <summary>
/// Create a new <see cref="ForUtil"/> instance and save state into <paramref name="out"/>.
/// </summary>
internal ForUtil(float acceptableOverheadRatio, DataOutput @out)
{
@out.WriteVInt32(PackedInt32s.VERSION_CURRENT);
encodedSizes = new int[33];
encoders = new PackedInt32s.IEncoder[33];
decoders = new PackedInt32s.IDecoder[33];
iterations = new int[33];
for (int bpv = 1; bpv <= 32; ++bpv)
{
PackedInt32s.FormatAndBits formatAndBits = PackedInt32s.FastestFormatAndBits(Lucene41PostingsFormat.BLOCK_SIZE, bpv, acceptableOverheadRatio);
Debug.Assert(formatAndBits.Format.IsSupported(formatAndBits.BitsPerValue));
Debug.Assert(formatAndBits.BitsPerValue <= 32);
encodedSizes[bpv] = EncodedSize(formatAndBits.Format, PackedInt32s.VERSION_CURRENT, formatAndBits.BitsPerValue);
encoders[bpv] = PackedInt32s.GetEncoder(formatAndBits.Format, PackedInt32s.VERSION_CURRENT, formatAndBits.BitsPerValue);
decoders[bpv] = PackedInt32s.GetDecoder(formatAndBits.Format, PackedInt32s.VERSION_CURRENT, formatAndBits.BitsPerValue);
iterations[bpv] = ComputeIterations(decoders[bpv]);
@out.WriteVInt32(formatAndBits.Format.Id << 5 | (formatAndBits.BitsPerValue - 1));
}
}
/// <summary>
/// Create a new <see cref="PagedMutable"/> instance.
/// </summary>
/// <param name="size"> The number of values to store. </param>
/// <param name="pageSize"> The number of values per page. </param>
/// <param name="bitsPerValue"> The number of bits per value. </param>
/// <param name="acceptableOverheadRatio"> An acceptable overhead ratio. </param>
public PagedMutable(long size, int pageSize, int bitsPerValue, float acceptableOverheadRatio)
: this(size, pageSize, PackedInt32s.FastestFormatAndBits(pageSize, bitsPerValue, acceptableOverheadRatio))
{
FillPages();
}
internal virtual void AddNumericField(FieldInfo field, IEnumerable <long?> values, bool optimizeStorage)
{
meta.WriteVInt32(field.Number);
meta.WriteByte(MemoryDocValuesProducer.NUMBER);
meta.WriteInt64(data.GetFilePointer());
long minValue = long.MaxValue;
long maxValue = long.MinValue;
long gcd = 0;
bool missing = false;
// TODO: more efficient?
ISet <long?> uniqueValues = null;
if (optimizeStorage)
{
uniqueValues = new JCG.HashSet <long?>();
long count = 0;
foreach (var nv in values)
{
long v;
if (nv == null)
{
v = 0;
missing = true;
}
else
{
v = nv.Value;
}
if (gcd != 1)
{
if (v < long.MinValue / 2 || v > long.MaxValue / 2)
{
// in that case v - minValue might overflow and make the GCD computation return
// wrong results. Since these extreme values are unlikely, we just discard
// GCD computation for them
gcd = 1;
} // minValue needs to be set first
else if (count != 0)
{
gcd = MathUtil.Gcd(gcd, v - minValue);
}
}
minValue = Math.Min(minValue, v);
maxValue = Math.Max(maxValue, v);
if (uniqueValues != null)
{
if (uniqueValues.Add(v))
{
if (uniqueValues.Count > 256)
{
uniqueValues = null;
}
}
}
++count;
}
if (Debugging.AssertsEnabled)
{
Debugging.Assert(count == maxDoc);
}
}
if (missing)
{
long start = data.GetFilePointer();
WriteMissingBitset(values);
meta.WriteInt64(start);
meta.WriteInt64(data.GetFilePointer() - start);
}
else
{
meta.WriteInt64(-1L);
}
if (uniqueValues != null)
{
// small number of unique values
int bitsPerValue = PackedInt32s.BitsRequired(uniqueValues.Count - 1);
FormatAndBits formatAndBits = PackedInt32s.FastestFormatAndBits(maxDoc, bitsPerValue,
acceptableOverheadRatio);
if (formatAndBits.BitsPerValue == 8 && minValue >= sbyte.MinValue && maxValue <= sbyte.MaxValue)
{
meta.WriteByte(MemoryDocValuesProducer.UNCOMPRESSED); // uncompressed
foreach (var nv in values)
{
data.WriteByte((byte)nv.GetValueOrDefault());
}
}
else
{
meta.WriteByte(MemoryDocValuesProducer.TABLE_COMPRESSED); // table-compressed
long?[] decode = new long?[uniqueValues.Count];
uniqueValues.CopyTo(decode, 0);
var encode = new Dictionary <long?, int?>();
data.WriteVInt32(decode.Length);
for (int i = 0; i < decode.Length; i++)
{
data.WriteInt64(decode[i].Value);
encode[decode[i]] = i;
}
meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
data.WriteVInt32(formatAndBits.Format.Id);
data.WriteVInt32(formatAndBits.BitsPerValue);
PackedInt32s.Writer writer = PackedInt32s.GetWriterNoHeader(data, formatAndBits.Format, maxDoc,
formatAndBits.BitsPerValue, PackedInt32s.DEFAULT_BUFFER_SIZE);
foreach (var nv in values)
{
var v = encode[nv.GetValueOrDefault()];
writer.Add((long)v);
}
writer.Finish();
}
}
else if (gcd != 0 && gcd != 1)
{
meta.WriteByte(MemoryDocValuesProducer.GCD_COMPRESSED);
meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
data.WriteInt64(minValue);
data.WriteInt64(gcd);
data.WriteVInt32(MemoryDocValuesProducer.BLOCK_SIZE);
var writer = new BlockPackedWriter(data, MemoryDocValuesProducer.BLOCK_SIZE);
foreach (var nv in values)
{
writer.Add((nv.GetValueOrDefault() - minValue) / gcd);
}
writer.Finish();
}
else
{
meta.WriteByte(MemoryDocValuesProducer.DELTA_COMPRESSED); // delta-compressed
meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
data.WriteVInt32(MemoryDocValuesProducer.BLOCK_SIZE);
var writer = new BlockPackedWriter(data, MemoryDocValuesProducer.BLOCK_SIZE);
foreach (var nv in values)
{
writer.Add(nv.GetValueOrDefault());
}
writer.Finish();
}
}
public override void AddNumericField(FieldInfo field, IEnumerable <long?> values)
{
meta.WriteVInt32(field.Number);
meta.WriteByte((byte)NUMBER);
meta.WriteInt64(data.GetFilePointer());
long minValue = long.MaxValue;
long maxValue = long.MinValue;
long gcd = 0;
// TODO: more efficient?
HashSet <long> uniqueValues = null;
if (true)
{
uniqueValues = new HashSet <long>();
long count = 0;
foreach (long?nv in values)
{
Debug.Assert(nv != null);
long v = nv.Value;
if (gcd != 1)
{
if (v < long.MinValue / 2 || v > long.MaxValue / 2)
{
// in that case v - minValue might overflow and make the GCD computation return
// wrong results. Since these extreme values are unlikely, we just discard
// GCD computation for them
gcd = 1;
} // minValue needs to be set first
else if (count != 0)
{
gcd = MathUtil.Gcd(gcd, v - minValue);
}
}
minValue = Math.Min(minValue, v);
maxValue = Math.Max(maxValue, v);
if (uniqueValues != null)
{
if (uniqueValues.Add(v))
{
if (uniqueValues.Count > 256)
{
uniqueValues = null;
}
}
}
++count;
}
Debug.Assert(count == maxDoc);
}
if (uniqueValues != null)
{
// small number of unique values
int bitsPerValue = PackedInt32s.BitsRequired(uniqueValues.Count - 1);
FormatAndBits formatAndBits = PackedInt32s.FastestFormatAndBits(maxDoc, bitsPerValue, acceptableOverheadRatio);
if (formatAndBits.BitsPerValue == 8 && minValue >= sbyte.MinValue && maxValue <= sbyte.MaxValue)
{
meta.WriteByte((byte)UNCOMPRESSED); // uncompressed
foreach (long?nv in values)
{
data.WriteByte((byte)nv.GetValueOrDefault());
}
}
else
{
meta.WriteByte((byte)TABLE_COMPRESSED); // table-compressed
var decode = uniqueValues.ToArray();
var encode = new Dictionary <long, int>();
data.WriteVInt32(decode.Length);
for (int i = 0; i < decode.Length; i++)
{
data.WriteInt64(decode[i]);
encode[decode[i]] = i;
}
meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
data.WriteVInt32(formatAndBits.Format.Id);
data.WriteVInt32(formatAndBits.BitsPerValue);
PackedInt32s.Writer writer = PackedInt32s.GetWriterNoHeader(data, formatAndBits.Format, maxDoc, formatAndBits.BitsPerValue, PackedInt32s.DEFAULT_BUFFER_SIZE);
foreach (long?nv in values)
{
writer.Add(encode[nv.GetValueOrDefault()]);
}
writer.Finish();
}
}
else if (gcd != 0 && gcd != 1)
{
meta.WriteByte((byte)GCD_COMPRESSED);
meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
data.WriteInt64(minValue);
data.WriteInt64(gcd);
data.WriteVInt32(BLOCK_SIZE);
var writer = new BlockPackedWriter(data, BLOCK_SIZE);
foreach (long?nv in values)
{
writer.Add((nv.GetValueOrDefault() - minValue) / gcd);
}
writer.Finish();
}
else
{
meta.WriteByte((byte)DELTA_COMPRESSED); // delta-compressed
meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
data.WriteVInt32(BLOCK_SIZE);
var writer = new BlockPackedWriter(data, BLOCK_SIZE);
foreach (long?nv in values)
{
writer.Add(nv.GetValueOrDefault());
}
writer.Finish();
}
}
internal virtual void AddNumericField(FieldInfo field, IEnumerable <long?> values, bool optimizeStorage)
{
Meta.WriteVInt32(field.Number);
Meta.WriteByte((byte)Lucene42DocValuesProducer.NUMBER);
Meta.WriteInt64(Data.GetFilePointer());
long minValue = long.MaxValue;
long maxValue = long.MinValue;
long gcd = 0;
// TODO: more efficient?
HashSet <long> uniqueValues = null;
if (optimizeStorage)
{
uniqueValues = new HashSet <long>();
long count = 0;
foreach (long?nv in values)
{
// TODO: support this as MemoryDVFormat (and be smart about missing maybe)
long v = nv.GetValueOrDefault();
if (gcd != 1)
{
if (v < long.MinValue / 2 || v > long.MaxValue / 2)
{
// in that case v - minValue might overflow and make the GCD computation return
// wrong results. Since these extreme values are unlikely, we just discard
// GCD computation for them
gcd = 1;
} // minValue needs to be set first
else if (count != 0)
{
gcd = MathUtil.Gcd(gcd, v - minValue);
}
}
minValue = Math.Min(minValue, v);
maxValue = Math.Max(maxValue, v);
if (uniqueValues != null)
{
if (uniqueValues.Add(v))
{
if (uniqueValues.Count > 256)
{
uniqueValues = null;
}
}
}
++count;
}
Debug.Assert(count == MaxDoc);
}
if (uniqueValues != null)
{
// small number of unique values
int bitsPerValue = PackedInt32s.BitsRequired(uniqueValues.Count - 1);
FormatAndBits formatAndBits = PackedInt32s.FastestFormatAndBits(MaxDoc, bitsPerValue, AcceptableOverheadRatio);
if (formatAndBits.BitsPerValue == 8 && minValue >= sbyte.MinValue && maxValue <= sbyte.MaxValue)
{
Meta.WriteByte((byte)Lucene42DocValuesProducer.UNCOMPRESSED); // uncompressed
foreach (long?nv in values)
{
Data.WriteByte((byte)nv.GetValueOrDefault());
}
}
else
{
Meta.WriteByte((byte)Lucene42DocValuesProducer.TABLE_COMPRESSED); // table-compressed
long[] decode = uniqueValues.ToArray(/*new long?[uniqueValues.Count]*/);
var encode = new Dictionary <long, int>();
Data.WriteVInt32(decode.Length);
for (int i = 0; i < decode.Length; i++)
{
Data.WriteInt64(decode[i]);
encode[decode[i]] = i;
}
Meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
Data.WriteVInt32(formatAndBits.Format.Id);
Data.WriteVInt32(formatAndBits.BitsPerValue);
PackedInt32s.Writer writer = PackedInt32s.GetWriterNoHeader(Data, formatAndBits.Format, MaxDoc, formatAndBits.BitsPerValue, PackedInt32s.DEFAULT_BUFFER_SIZE);
foreach (long?nv in values)
{
writer.Add(encode[nv.GetValueOrDefault()]);
}
writer.Finish();
}
}
else if (gcd != 0 && gcd != 1)
{
Meta.WriteByte((byte)Lucene42DocValuesProducer.GCD_COMPRESSED);
Meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
Data.WriteInt64(minValue);
Data.WriteInt64(gcd);
Data.WriteVInt32(Lucene42DocValuesProducer.BLOCK_SIZE);
BlockPackedWriter writer = new BlockPackedWriter(Data, Lucene42DocValuesProducer.BLOCK_SIZE);
foreach (long?nv in values)
{
writer.Add((nv.GetValueOrDefault() - minValue) / gcd);
}
writer.Finish();
}
else
{
Meta.WriteByte((byte)Lucene42DocValuesProducer.DELTA_COMPRESSED); // delta-compressed
Meta.WriteVInt32(PackedInt32s.VERSION_CURRENT);
Data.WriteVInt32(Lucene42DocValuesProducer.BLOCK_SIZE);
BlockPackedWriter writer = new BlockPackedWriter(Data, Lucene42DocValuesProducer.BLOCK_SIZE);
foreach (long?nv in values)
{
writer.Add(nv.GetValueOrDefault());
}
writer.Finish();
}
}
