2
0
mirror of https://github.com/esiur/esiur-dotnet.git synced 2026-04-29 06:48:41 +00:00

Gvwie updated

This commit is contained in:
2026-04-09 16:36:05 +03:00
parent 15479288cc
commit ae524f577a
9 changed files with 575 additions and 509 deletions
+65 -66
View File
@@ -6,16 +6,6 @@ namespace Esiur.Data.Gvwie;
public static class GroupUInt64Codec
{
// Header layout:
// 1 | cccc | www
//
// MSB = 1 => grouped form
// cccc = 0..14 => short count = cccc + 1 (1..15)
// cccc = 15 => extended count, followed by varint(count - 16)
// www = 0..7 => width = www + 1 (1..8)
//
// MSB = 0 => literal fast path for values in 7 bits
// ----------------- Encoder -----------------
public static byte[] Encode(IList<ulong> values, bool aligned = false)
{
@@ -35,7 +25,7 @@ public static class GroupUInt64Codec
}
int start = i;
int width = WidthFromValue(v, aligned);
int width = WidthFromUInt64(v, aligned);
int count = 1;
// Build a run of same-width non-literal values
@@ -47,17 +37,18 @@ public static class GroupUInt64Codec
if (v2 <= 0x7Ful)
break;
int w2 = WidthFromValue(v2, aligned);
int w2 = WidthFromUInt64(v2, aligned);
if (w2 != width)
break;
count++;
}
if (count <= 15)
if (count <= 12)
{
// Short group:
// Header: 1 | (count-1)[4 bits] | (width-1)[3 bits]
// count field 0000..1011 => count 1..12
byte header = 0x80;
header |= (byte)(((count - 1) & 0x0F) << 3);
header |= (byte)((width - 1) & 0x07);
@@ -66,13 +57,33 @@ public static class GroupUInt64Codec
else
{
// Extended group:
// Header: 1 | 1111 | (width-1)[3 bits]
// Followed by varint(count - 16)
// Header: 1 | g[4 bits] | (width-1)[3 bits]
//
// g = 1100 => LoL = 1 byte
// g = 1101 => LoL = 2 bytes
// g = 1110 => LoL = 3 bytes
// g = 1111 => LoL = 4 bytes
//
// LoL stores (count - 13) in little-endian form.
uint extra = checked((uint)(count - 13));
int lol = LengthOfLength(extra);
byte groupBits = lol switch
{
1 => 0b1100,
2 => 0b1101,
3 => 0b1110,
4 => 0b1111,
_ => throw new InvalidOperationException("Invalid LoL.")
};
byte header = 0x80;
header |= 0x78; // count bits = 1111
header |= (byte)(groupBits << 3);
header |= (byte)((width - 1) & 0x07);
dst.Add(header);
WriteVarUInt32(dst, checked((uint)(count - 16)));
WriteLE(dst, extra, lol);
}
// Payload: 'count' values, LE, 'width' bytes each
@@ -97,30 +108,36 @@ public static class GroupUInt64Codec
if ((h & 0x80) == 0)
{
// Fast path: literal 7-bit unsigned value
// Fast path: 7-bit literal in low bits
result.Add((ulong)(h & 0x7F));
continue;
}
int countField = (h >> 3) & 0x0F;
int width = (h & 0x07) + 1; // 1..8
int width = (h & 0x07) + 1;
int count;
if (countField == 15)
if (countField <= 11)
{
uint extra = ReadVarUInt32(src, ref pos);
count = checked(16 + (int)extra);
// Short group: 0..11 => count 1..12
count = countField + 1;
}
else
{
count = countField + 1;
// Extended group:
// 12 => LoL=1
// 13 => LoL=2
// 14 => LoL=3
// 15 => LoL=4
int lol = countField - 11;
uint extra = checked((uint)ReadLE(src, ref pos, lol));
count = checked(13 + (int)extra);
}
for (int j = 0; j < count; j++)
{
ulong raw = ReadLE(src, ref pos, width);
result.Add(raw);
}
result.Add(ReadLE(src, ref pos, width));
}
return result.ToArray();
@@ -129,18 +146,27 @@ public static class GroupUInt64Codec
// ----------------- Helpers -----------------
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static int WidthFromValue(ulong v, bool aligned = false)
private static int WidthFromUInt64(ulong v, bool aligned = false)
{
if (v <= 0xFFul) return 1;
if (v <= 0xFFFFul) return 2;
if (v <= 0xFFFFFFul) return aligned ? 4 : 3;
if (v <= 0xFFFFFFFFul) return 4;
if (v <= 0xFFFFFFFFFFul) return aligned ? 8: 5;
if (v <= 0xFFFFFFFFFFFFul) return aligned ? 8: 6;
if (v <= 0xFFFFFFFFFFul) return aligned ? 8 : 5;
if (v <= 0xFFFFFFFFFFFFul) return aligned ? 8 : 6;
if (v <= 0xFFFFFFFFFFFFFFul) return aligned ? 8 : 7;
return 8;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static int LengthOfLength(uint value)
{
if (value <= 0xFFu) return 1;
if (value <= 0xFFFFu) return 2;
if (value <= 0xFFFFFFu) return 3;
return 4;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static void WriteLE(List<byte> dst, ulong value, int width)
{
@@ -148,50 +174,23 @@ public static class GroupUInt64Codec
dst.Add((byte)((value >> (8 * i)) & 0xFF));
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static void WriteLE(List<byte> dst, uint value, int width)
{
for (int i = 0; i < width; i++)
dst.Add((byte)((value >> (8 * i)) & 0xFF));
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static ulong ReadLE(ReadOnlySpan<byte> src, ref int pos, int width)
{
if ((uint)(pos + width) > (uint)src.Length)
throw new ArgumentException("Buffer underflow while reading group payload.");
throw new ArgumentException("Buffer underflow while reading payload.");
ulong v = 0;
for (int i = 0; i < width; i++)
v |= (ulong)src[pos++] << (8 * i);
return v;
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static void WriteVarUInt32(List<byte> dst, uint value)
{
while (value >= 0x80)
{
dst.Add((byte)((value & 0x7F) | 0x80));
value >>= 7;
}
dst.Add((byte)value);
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static uint ReadVarUInt32(ReadOnlySpan<byte> src, ref int pos)
{
uint result = 0;
int shift = 0;
while (true)
{
if (pos >= src.Length)
throw new ArgumentException("Buffer underflow while reading varint.");
byte b = src[pos++];
result |= (uint)(b & 0x7F) << shift;
if ((b & 0x80) == 0)
return result;
shift += 7;
if (shift >= 35)
throw new ArgumentException("Varint is too long for UInt32.");
}
}
}