BattleMoonWars 归档解压/压缩程序（砍掉重炼版）

RednaxelaFX

浏览: 3015864 次
性别:
来自: 海外

最近访客更多访客>>

suen

xckouy

limn_xmj

lingxiajiudu

博主相关

博客

微博

相册

留言

关于我

文章分类

社区版块

存档分类

博客分类：

Reverse Engineering

Ruby C#C++C EXT

以前写过BattleMoonWars的归档处理程序。可惜在JavaEye上的老帖里的代码真的非常老了；那代码我后来改过几次的，修掉以前偷懒不支持多层目录的问题，还修过对应LZSS解压的问题（但一直偷懒没实现压缩……用Java来写LZSS压缩真痛苦）。
改过的版本老早就不知道哪儿去了。茶茶也找不到那代码了，但要发新的BMW完全版的汉化补丁又等着用这打包工具……拖到今天终于重新写了个。

以前那个是用Java写的，这次用Ruby来试试。目标是用干净的代码实现归档处理程序需要的解包和打包功能，外带应用上LZSS的压缩。
写完之后，拿代码跟以前的Java版比较了一下，觉得自己貌似没啥长进。眯着眼睛看的话会觉得俩版本都差不多，至少“算法”都一样——这么小的东西根本没用上啥算法嘛orz
不过以前那版代码要想整理得干净一些的话显然是有许多办法的。印象中后来也没这么乱了，但，找不到了啊 T T
诶，用Java这么重量级的语言去实现这么简单的小工具本身就是个杯具。要是当年我就熟悉什么脚本语言的话……

Ruby的代码还是比Java的显著的少了。我倒没特意要为了短而写短，反而有些地方还特意用了比较长的名字和有点冗余的运算，因为懒得写注释，代码本身就应该能体现意图。
代码能减少的一个原因在于Java的标准I/O API用起来不方便，而且默认用big endian，在我们这little endian的平台上用起来很是费事。就为了各种转换endian、在byte[]与别的数据类型间转换之类的地方就写了很多垃圾代码。
相比之下Ruby的I/O API用起来非常方便，而且靠强大的String#unpack与Array#pack就已经可以解决很多转换问题，所以读取索引项的代码Ruby版比Java版短很多。
另外，Java代码的“噪音”还是比较大。嵌套的a.setXxx(b.getXxx())这类代码自然比不上a.xxx = b.xxx来得清爽。

OK，又到大段贴代码时间。贴代码比放附件“安全”多了，回头Google Reader把这篇备份到了之后我就不怕代码丢失了 ha ha ha
这次用的Ruby版本是去年7月自己用VC9来build的1.9.2dev。

引用

ruby 1.9.2dev (2009-07-22 trunk 24241) [i386-mswin32_90]

在GBK编码的系统上打开文件时，可能要注意用String#force_encoding('gbk')强制将ASCII-8BIT编码的文件名转换一下。

bmw_archiver.rb

#!../ruby19/bin/ruby
# coding: binary

require 'fileutils'
require File.join(File.expand_path(File.dirname(__FILE__)), 'compress/lzss.so')
include FileUtils

module BMWArchiver
  class Archive
    BENIFIT_THRESHOLD = 20
    
    SIGNATURE = 'yanepkDx'
    attr_accessor :arc_path, :dir_path, :entry_count
    attr_reader   :index
    
    class IndexEntry
      LENGTH = 256 + 4 * 3
      
      attr_accessor :file_path, :offset
      attr_accessor :original_length, :compressed_length
      attr_writer   :compressed
      
      def initialize(file_path, offset, original_length, compressed_length)
        @file_path, @offset, @original_length, @compressed_length =
          file_path, offset, original_length, compressed_length
        
        @compressed = original_length > compressed_length
      end
      
      def compressed?
        @compressed
      end
      
      def should_compress?
        ['.yga', '.ogg'].none? {|ext| @file_path.downcase.end_with? ext }
      end
      
      def self.from_binary(str)
        IndexEntry.new(*(str.unpack 'Z256III'))
      end
      
      def to_binary
        [
          @file_path, @offset,
          @original_length, @compressed_length
        ].pack('Z256III')
      end
      
      def to_s
        <<-ENTRY_STR
IndexEntry {
  file_path:         #@file_path
  offset:            0x#{@offset.to_s 16}
  compression:       #{self.compressed? ? 'LZSS' : 'none'}
  original_length:   #@original_length
  compressed_length: #@compressed_length
}
        ENTRY_STR
      end
    end
    
    def add_index_entry(entry)
      @index ||= []
      @index << entry
    end
    
    def extract_to(dest_dir_path)
      @dir_path = dest_dir_path
      File.open(@arc_path, 'rb') do |arc_file|
        @index.each do |entry|
          arc_file.pos = entry.offset
          contents = arc_file.read(entry.compressed_length)
          if entry.compressed?
            contents = Compress::LZSS::decode(contents, entry.original_length)
          end
          dest_file_path = File.join(dest_dir_path, entry.file_path)
          mkdir_p File.dirname dest_file_path
          File.open(dest_file_path, 'wb') do |dest_file|
            dest_file << contents
          end
        end
      end
    end
    
    def save_to(dest_arc_path)
      @arc_path = dest_arc_path
      File.open(@arc_path, 'wb') do |arc_file|
        # write signature and entry count
        arc_file << SIGNATURE << [@entry_count].pack('I')
        
        # write contents first, and calculate offset
        index_section_start = SIGNATURE.bytesize + 4
        index_length        = IndexEntry::LENGTH * @entry_count
        data_section_start  = index_section_start + index_length
        current_offset      = data_section_start
        arc_file.pos        = current_offset
        
        @index.each do |entry|
          entry.offset = current_offset
          contents = File.binread File.join(@dir_path, entry.file_path)
          if entry.should_compress?
            compressed_contents = Compress::LZSS::encode(contents)
            compressed_length   = compressed_contents.bytesize
            
            # only use compressed version when it's benificial
            if compressed_length + BENIFIT_THRESHOLD < entry.original_length
              contents = compressed_contents
              entry.compressed_length = compressed_length
              entry.compressed = true
            end
          end
          arc_file << contents
          current_offset += entry.compressed_length
        end
        
        # write index
        arc_file.pos = index_section_start
        @index.each do |entry|
          arc_file << entry.to_binary
        end
      end
    end
    
    def to_s
      <<-ARCHIVE_STR
BMW Archive file: #@arc_path
No. of entries: #@entry_count
Entries:
#{@index.map{|e| e.to_s}.join}
      ARCHIVE_STR
    end
    
    class << self
      # builds an Archive object with index initialized from specified file
      def from_archive(path)
        File.open(path, 'rb') do |arc_file|
          verify_signature_of arc_file
          self.new.tap do |arc|
            arc.arc_path    = path
            arc.entry_count = read_int32_from arc_file
            arc.entry_count.times do
              arc.add_index_entry(
                IndexEntry.from_binary(arc_file.read(IndexEntry::LENGTH)))
            end
          end
        end
      end
      
      def build_from_dir(src_dir_path)
        verify_src_dir src_dir_path
        
        self.new.tap do |arc|
          arc.dir_path = src_dir_path
          
          # HACK: for the ease of using Dir.glob
          old_pwd = Dir.pwd
          cd src_dir_path
          
          Dir.glob('**/*').each do |path|
            next if Dir.exists? path  # skip directories
            
            file_length = File.size path
            arc.add_index_entry IndexEntry.new(
              path.gsub('/', '\\'),
              -1,                     # offset not calculated yet
              file_length,
              file_length)
          end
          
          arc.entry_count = arc.index.size
          
          cd old_pwd
        end
      end
      
    private
      def verify_signature_of(file)
        raise 'Wrong signature' unless file.read(8) == SIGNATURE
      end
      
      def verify_src_dir(dir_path)
        unless Dir.exists? dir_path
          raise "The specified source directory doesn't exist: #{dir_path}"
        end
      end
      
      def read_int32_from(file)
        file.read(4).unpack('I').first
      end
    end
  end
end


if __FILE__ == $0
  include BMWArchiver
  opt, arc_path, dir_path = ARGV
  case opt
  when 'e'
    arc = Archive.from_archive arc_path
    arc.extract_to dir_path
    puts arc
  when 'a'
    arc = Archive.build_from_dir dir_path
    arc.save_to arc_path
    puts arc
  else
    puts <<-USAGE_STR
Usage: bmw_archiver opt arc_file dir_path
options:
  e - extract arc_file to dir_path
  a - pack dir_path into arc_file
    USAGE_STR
  end
  puts 'done'
end

上面就是这归档处理程序的主要代码了。
这次多得有night_stalker同学的大力支持，帮忙在一个现成的经典LZSS实现上包了层Ruby扩展。下面是完整代码。

他已经把这个包装版发成一个gem了，地址是http://rubygems.org/gems/lzss。安装此gem需要一个C编译器，GCC、VC之类都可以。

extconf.rb

require "mkmf"

create_makefile 'lzss'

把实际实现注册给Ruby用：

lzss-ext.c

#include <ruby.h>

size_t Encode(size_t ilen, char* istr, size_t olen, char* ostr);
void Decode(size_t ilen, unsigned char* istr, size_t olen, char* ostr);

static VALUE encode(VALUE self, VALUE str) {
	size_t ilen = RSTRING_LEN(str);
	char* buff = (char*)malloc(ilen * 2);
	size_t olen = Encode(RSTRING_LEN(str), RSTRING_PTR(str), ilen * 2, buff);
	VALUE ret = rb_str_new(buff, olen);
	free(buff);
	return ret;
}

static VALUE decode(VALUE self, VALUE str, VALUE v_olen) {
	size_t ilen = RSTRING_LEN(str);
	VALUE ret = rb_str_new(0, NUM2INT(v_olen));
	Decode(ilen, RSTRING_PTR(str), NUM2INT(v_olen), RSTRING_PTR(ret));
	return ret;
}

void Init_lzss() {
	VALUE Compress = rb_define_module("Compress");
	VALUE LZSS = rb_define_module_under(Compress, "LZSS");
	rb_define_module_function(LZSS, "encode", RUBY_METHOD_FUNC(encode), 1);
	rb_define_module_function(LZSS, "decode", RUBY_METHOD_FUNC(decode), 2);
}

实际LZSS实现用的是奥村晴彦教授在80年代末写的版本。日本许多游戏里用的LZSS压缩实现都是从这个版本演化而来的，用起来非常顺手。
稍微修改了几处：
1、原始版本是用空格（0x20）来填充滑动窗口的初始值，这里修改为用0x00填充。由于原本实现主要用于压缩文本，用空格填充是个很合理的选择；这里则可能压缩任意二进制数据，所以选择用0x00填充。
2、把原本直接对FILE的I/O操作改为对unsigned char*的buffer的操作，便于把它当作库来用。
算法本身是完全没有修改的。没尝试去提高压缩速度（解压速度也没啥改善余地了）。其实这个时候还有后续的进化版，但我的首要目标是让这个归档处理程序能压缩出跟那游戏引擎里的工具完全一致的归档，自然就不太关心进一步提升压缩速度或压缩率了 =_=|||

lzss.c

/**************************************************************
    LZSS.C -- A Data Compression Program
    (tab = 4 spaces)
***************************************************************
    4/6/1989 Haruhiko Okumura
    Use, distribute, and modify this program freely.
    Please send me your improved versions.
        PC-VAN        SCIENCE
        NIFTY-Serve    PAF01022
        CompuServe    74050,1022
**************************************************************/
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>

#define N         4096    /* size of ring buffer */
#define F           18    /* upper limit for match_length */
#define THRESHOLD    2    /* encode string into position and length if match_length is greater than this */
#define NIL          N    /* index for root of binary search trees */

/* of longest match.  These are set by the InsertNode() procedure. */
static int match_position;
static int match_length;

static void InsertNode(unsigned char* text_buf, int* lson, int* rson, int* dad, int r)
    /* Inserts string of length F, text_buf[r..r+F-1], into one of the
       trees (text_buf[r]'th tree) and returns the longest-match position
       and length via the global variables match_position and match_length.
       If match_length = F, then removes the old node in favor of the new
       one, because the old one will be deleted sooner.
       Note r plays double role, as tree node and position in buffer. */
{
    int  i, p, cmp;
    unsigned char  *key;

    cmp = 1;  key = &text_buf[r];  p = N + 1 + key[0];
    rson[r] = lson[r] = NIL;  match_length = 0;
    for ( ; ; ) {
        if (cmp >= 0) {
            if (rson[p] != NIL) p = rson[p];
            else {  rson[p] = r;  dad[r] = p;  return;  }
        } else {
            if (lson[p] != NIL) p = lson[p];
            else {  lson[p] = r;  dad[r] = p;  return;  }
        }
        for (i = 1; i < F; i++)
            if ((cmp = key[i] - text_buf[p + i]) != 0)  break;
        if (i > match_length) {
            match_position = p;
            if ((match_length = i) >= F)  break;
        }
    }
    dad[r] = dad[p];  lson[r] = lson[p];  rson[r] = rson[p];
    dad[lson[p]] = r;  dad[rson[p]] = r;
    if (rson[dad[p]] == p) rson[dad[p]] = r;
    else                  lson[dad[p]] = r;
    dad[p] = NIL;  /* remove p */
}

static void DeleteNode(int* lson, int* rson, int* dad, int p)  /* deletes node p from tree */
{
    int  q;
    
    if (dad[p] == NIL) return;  /* not in tree */
    if (rson[p] == NIL) q = lson[p];
    else if (lson[p] == NIL) q = rson[p];
    else {
        q = lson[p];
        if (rson[q] != NIL) {
            do {  q = rson[q];  } while (rson[q] != NIL);
            rson[dad[q]] = lson[q];  dad[lson[q]] = dad[q];
            lson[q] = lson[p];  dad[lson[p]] = q;
        }
        rson[q] = rson[p];  dad[rson[p]] = q;
    }
    dad[q] = dad[p];
    if (rson[dad[p]] == p) rson[dad[p]] = q;  else lson[dad[p]] = q;
    dad[p] = NIL;
}

#define _get(c) \
	if (! ilen) {\
		c = EOF;\
		break;\
	}\
	c = *istr;\
	++istr;\
	--ilen

#define _put(c) \
	*ostr = c;\
	++ostr;\
	--olen

size_t Encode(size_t ilen, char* istr, size_t olen, char* ostr)
{
    int  i, c, len, r, s, last_match_length, code_buf_ptr;
    unsigned char  code_buf[17], mask;
	size_t codesize = 0;
	int lson[N + 1], rson[N + 257], dad[N + 1]; /* left & right children & parents -- These constitute binary search trees. */
	unsigned char text_buf[N + F - 1];          /* ring buffer of size N, with extra F-1 bytes to facilitate string comparison */
  
  match_position = 0;
  match_length = 0;

	if (ilen == 0) return 0;
    
    /* initialize trees */
	/* For i = 0 to N - 1, rson[i] and lson[i] will be the right and
    left children of node i.  These nodes need not be initialized.
    Also, dad[i] is the parent of node i.  These are initialized to
    NIL (= N), which stands for 'not used.'
    For i = 0 to 255, rson[N + i + 1] is the root of the tree
    for strings that begin with character i.  These are initialized
    to NIL.  Note there are 256 trees. */
    for (i = N + 1; i <= N + 256; i++) rson[i] = NIL;
    for (i = 0; i < N; i++) dad[i] = NIL;

    code_buf[0] = 0;  /* code_buf[1..16] saves eight units of code, and
        code_buf[0] works as eight flags, "1" representing that the unit
        is an unencoded letter (1 byte), "0" a position-and-length pair
        (2 bytes).  Thus, eight units require at most 16 bytes of code. */
    code_buf_ptr = mask = 1;
    s = 0;  r = N - F;
    for (i = s; i < r; i++) text_buf[i] = 0;  /* Clear the buffer with
        any character that will appear often. */
    for (len = 0; len < F && ilen; len++) {
		_get(c);
        text_buf[r + len] = c;
		/* Read F bytes into the last F bytes of the buffer */
	}
    for (i = 1; i <= F; i++) InsertNode(text_buf, lson, rson, dad, r - i);  /* Insert the F strings,
        each of which begins with one or more 'space' characters.  Note
        the order in which these strings are inserted.  This way,
        degenerate trees will be less likely to occur. */
    InsertNode(text_buf, lson, rson, dad, r);  /* Finally, insert the whole string just read.  The
        global variables match_length and match_position are set. */
    do {
        if (match_length > len) match_length = len;  /* match_length
            may be spuriously long near the end of text. */
        if (match_length <= THRESHOLD) {
            match_length = 1;  /* Not long enough match.  Send one byte. */
            code_buf[0] |= mask;  /* 'send one byte' flag */
            code_buf[code_buf_ptr++] = text_buf[r];  /* Send uncoded. */
        } else {
            code_buf[code_buf_ptr++] = (unsigned char) match_position;
            code_buf[code_buf_ptr++] = (unsigned char)
                (((match_position >> 4) & 0xf0)
             | (match_length - (THRESHOLD + 1)));  /* Send position and
                    length pair. Note match_length > THRESHOLD. */
        }
        if ((mask <<= 1) == 0) {  /* Shift mask left one bit. */
            for (i = 0; i < code_buf_ptr; i++) {  /* Send at most 8 units of */
                _put(code_buf[i]);    /* code together */
			}
            codesize += code_buf_ptr;
            code_buf[0] = 0;  code_buf_ptr = mask = 1;
        }
        last_match_length = match_length;
        for (i = 0; i < last_match_length && ilen; i++) {
			_get(c);
            DeleteNode(lson, rson, dad, s);        /* Delete old strings and */
            text_buf[s] = c;    /* read new bytes */
            if (s < F - 1) text_buf[s + N] = c;  /* If the position is
                near the end of buffer, extend the buffer to make
                string comparison easier. */
            s = (s + 1) & (N - 1);  r = (r + 1) & (N - 1);
                /* Since this is a ring buffer, increment the position
                   modulo N. */
            InsertNode(text_buf, lson, rson, dad, r);    /* Register the string in text_buf[r..r+F-1] */
        }
        while (i++ < last_match_length) {    /* After the end of text, */
            DeleteNode(lson, rson, dad, s);                    /* no need to read, but */
            s = (s + 1) & (N - 1);  r = (r + 1) & (N - 1);
            if (--len) InsertNode(text_buf, lson, rson, dad, r);        /* buffer may not be empty. */
        }
    } while (len > 0);    /* until length of string to be processed is zero */
    if (code_buf_ptr > 1) {        /* Send remaining code. */
        for (i = 0; i < code_buf_ptr; i++) {
			_put(code_buf[i]);
		}
        codesize += code_buf_ptr;
    }

	return codesize;
}

void Decode(size_t ilen, unsigned char* istr, size_t olen, char* ostr)    /* Just the reverse of Encode(). */
{
	unsigned char text_buf[N + F - 1];          /* ring buffer of size N, with extra F-1 bytes to facilitate string comparison */
    int  i, j, k, r, c;
    unsigned int  flags;
    
    for (i = 0; i < N - F; i++) text_buf[i] = 0;
    r = N - F;  flags = 0;
    for ( ; ; ) {
        if (((flags >>= 1) & 256) == 0) {
			_get(c);
            flags = c | 0xff00;        /* uses higher byte cleverly */
        }                            /* to count eight */
        if (flags & 1) {
            _get(c);
            _put(c);  text_buf[r++] = c;  r &= (N - 1);
        } else {
            _get(i);
            _get(j);
            i |= ((j & 0xf0) << 4);  j = (j & 0x0f) + THRESHOLD;
            for (k = 0; k <= j; k++) {
                c = text_buf[(i + k) & (N - 1)];
                _put(c);  text_buf[r++] = c;  r &= (N - 1);
            }
        }
    }
}

#undef _get
#undef _put

最后把需要的Ruby和那归档处理程序打个包放附件。备份完成 XDD

bmw_archiver.zip (3.9 MB)
下载次数: 10

分享到：

HotSpot 17.0-b12的逃逸分析/标量替换的一 ... | 关于Java程序的执行的一次分享

2010-05-03 21:40
浏览 2231
评论(8)
分类:非技术
查看更多

8 楼 night_stalker 2010-05-08

唔，原来 github 可以加多个 ssh key，终于可以把代码扔上去了 ……
http://github.com/luikore/lzss

7 楼 RednaxelaFX 2010-05-04

Sword-Breaker 写道

咦,我一直以为只有ppg和sosg的在做...(当年在ppg论坛看到的补丁,貌似是sosg+ppg的),补丁出来的话请务必在推特上说一句(我FO你了=,=)...以前订你的blog都是为了技术文=,=没想到突然跑个BMW银的文章出来=,=

咳咳，这边就是PPG组……而且这篇也勉强算是技术文吧orz
有兴趣的话可以看一下左边栏里reverse engineering类里面的东西 >_<

哈哈，想当年还有宝马撞车事件，然后多交了几个碰友 XDD

6 楼 Sword-Breaker 2010-05-04

RednaxelaFX 写道

Sword-Breaker 写道

无意中发现关于BMW银的汉化信息=,=搜了一下发现sosg在去年11月就有一个补丁,不过似乎是一个有bug而且针对先行版的版本...按照你的说法=,=看来很快就有完整版了的汉化补丁了么

应该……快了吧？我也不太清楚现在进度如何，好久没跟大家联系过了。
茶茶一直在催我写这玩儿，说再不写就直接不打成DAT包发掉。
年初我也已经把改过的exe给了，应该就剩打包了吧？

Umm，这边不是SOSG。
我都不记得我们这组的BMW汉化最初是针对BMW2还是BMW3做的了，发也已经发过好几版补丁了。好久远啊……终于要完全结束了。

5 楼 RednaxelaFX 2010-05-04

Sword-Breaker 写道

4 楼 Sword-Breaker 2010-05-04

3 楼 lwwin 2010-05-04

当然－－我觉得你可以称作语言不依赖的koder了^^

然后特化到ruby么= = 好吧……我看不懂 OTL

2 楼 RednaxelaFX 2010-05-03

lwwin 写道

无论如何－－都不用C来写吗－－|||

不啊以前写的有几个工具是用C或者C++写的。没记错的话ef和恋楯的都是C++写的，还有啥是用C写的来着……忘了。太久没写这种小玩儿了 =_=|||
现在的话写这种东西我是会首选Ruby的，关键是自己写起来舒服就好。

1 楼 lwwin 2010-05-03

无论如何－－都不用C来写吗－－|||

发表评论

您还没有登录,请您登录后再发表评论

最近访客更多访客>>

博主相关

文章分类

社区版块

存档分类

最新评论