From a1822d7d89f7133a6c5b99386242c2d82d951a6f Mon Sep 17 00:00:00 2001
From: ddcw <ddcw@localhost>
Date: Tue, 24 Sep 2024 15:02:54 +0800
Subject: [PATCH] add lz4

---
 ibd2sql/ibd2sql.py |  8 ++++-
 ibd2sql/lz4.py     | 76 ++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 83 insertions(+), 1 deletion(-)
 create mode 100644 ibd2sql/lz4.py

diff --git a/ibd2sql/ibd2sql.py b/ibd2sql/ibd2sql.py
index 4ff3a39..1a09e1a 100644
--- a/ibd2sql/ibd2sql.py
+++ b/ibd2sql/ibd2sql.py
@@ -5,6 +5,7 @@
 from ibd2sql.innodb_page_spaceORxdes import *
 from ibd2sql.innodb_page_inode import *
 from ibd2sql.innodb_page_index import *
+from ibd2sql import lz4
 import sys
 
 
@@ -69,7 +70,12 @@ def read(self):
 		data = self.f.read(self.PAGESIZE)
 		if data[24:26] == b'\x00\x0e': # 14: 压缩页, 先解压
 			FIL_PAGE_VERSION,FIL_PAGE_ALGORITHM_V1,FIL_PAGE_ORIGINAL_TYPE_V1,FIL_PAGE_ORIGINAL_SIZE_V1,FIL_PAGE_COMPRESS_SIZE_V1 = struct.unpack('>BBHHH',data[26:34])
-			data = data[:24] + struct.pack('>H',FIL_PAGE_ORIGINAL_TYPE_V1) + b'\x00'*8 + data[34:38] + zlib.decompress(data[38:38+FIL_PAGE_COMPRESS_SIZE_V1])
+			if FIL_PAGE_ALGORITHM_V1 == 1:
+				data = data[:24] + struct.pack('>H',FIL_PAGE_ORIGINAL_TYPE_V1) + b'\x00'*8 + data[34:38] + zlib.decompress(data[38:38+FIL_PAGE_COMPRESS_SIZE_V1])
+			elif FIL_PAGE_ALGORITHM_V1 == 2:
+				data = data[:24] + struct.pack('>H',FIL_PAGE_ORIGINAL_TYPE_V1) + b'\x00'*8 + data[34:38] + lz4.decompress(data[38:38+FIL_PAGE_COMPRESS_SIZE_V1],FIL_PAGE_ORIGINAL_SIZE_V1)
+			else:
+				pass
 		return data
 
 	def _init_sql_prefix(self):
diff --git a/ibd2sql/lz4.py b/ibd2sql/lz4.py
new file mode 100644
index 0000000..32fa40a
--- /dev/null
+++ b/ibd2sql/lz4.py
@@ -0,0 +1,76 @@
+# write by ddcw @https://github.com/ddcw
+# lz4解压(fast)
+# 参考: https://github.com/lz4/lz4/blob/dev/doc/lz4_Block_format.md
+
+"""
+LZ4 compressed block is composed of sequences.
+sequence = token + [length literals] +  literals + offset + [length match] + match
+token: 1bytes, first 4-bits   length of literals
+               last  4-bits   length of match
+               each field ranges from 0 to 15, when 15, read more 1 bytes for length to add
+literals: not-compressed bytes
+offset  : 从解压后的数据的某个位置开始复制 match长度的数据
+match   : 要复制的数据的长度
+"""
+
+# lz4压缩(TODO)
+def compress(bdata):
+	"""
+	input:	bdata: 要压缩的数据
+	return: data:  压缩之后的数据
+	"""
+	return bdata
+
+
+# lz4解压
+def decompress(bdata,decompress_size):
+	"""
+	input:
+		bdata: 压缩数据
+		decompress_size : 解压之后的大小
+	return: data 解压之后的数据
+	不考虑dict和prefix_size了
+	"""
+	def read_to_less255(tdata,ip):
+		length = 0
+		while True:
+			#t = struct.unpack('<B',aa[ip:ip+1])[0]
+			t = tdata[ip]
+			ip += 1
+			length += t
+			if t != 255: # 小于255时,就读完了
+				break
+		return length,ip
+		
+	ip = 0 # input pointer  (bdata的指针)
+	op = 0 # output pointer (data的指针)
+	data = bytearray(decompress_size) # 要返回的数据有这么大
+	
+	while True:
+		token = bdata[ip]
+		ip += 1
+		ll = token >> 4 # literals length
+		if ll == 15:
+			tll,ip = read_to_less255(bdata,ip)
+			ll += tll
+		data[op:op+ll] = bdata[ip:ip+ll] # literals 不可压缩的部分
+		op += ll
+		ip += ll
+		if decompress_size-op < 12:
+			if op == decompress_size: # 解压完了, 因为可能没得后面的match部分
+				break
+			else:
+				raise ValueError('Invalid lz4 compress data.')
+		#offset = struct.unpack('<H',bdata[ip:ip+2])[0]
+		offset = (bdata[ip+1]<<8) | bdata[ip] # 位移真TM好用
+		ip += 2
+		ml = token & 15 # 后4bit是match length
+		if ml == 15:
+			tml,ip = read_to_less255(bdata,ip)
+			ml += tml
+		ml += 4 # 还得加4(minmatch)
+		match = op - offset
+		data[op:op+ml] = data[match:match+ml] # match实际上是指的原始数据位置的数据,而不是压缩之后的数据位置
+		op += ml # 重复的数据只需要移动op就行,ip那没得要移动的.(match部分已经mv了)
+	return bytes(data)
+