Repository: bisqwit/TinyDeflate
Branch: master
Commit: 68ced8bd5c81
Files: 2
Total size: 82.4 KB
Directory structure:
gitextract_cmy6_mlh/
├── README.md
└── gunzip.hh
================================================
FILE CONTENTS
================================================
================================================
FILE: README.md
================================================
# TinyDeflate
A deflate/gzip decompressor, as a C++17 template function,
that requires minimal amount of memory to work.
Terms of use: Zlib
Copyright © 2018 Joel Yliluoma
## Memory usage at aggressive settings (backtrackable input)
* 408 bytes of automatic storage for three huffman trees (384 elements, 5…9 bits each, 88 % space efficiency)
* 24 bytes of temporary automatic storage while a huffman tree is being generated (15 elements, 9 bits each, 66 % space efficiency)
* An assortment of automatic variables for various purposes (may be register variables, depending on the architecture and of the compiler wits)
* ABI mandated alignment losses
Total: 408 bytes minimum, 432+N bytes maximum
Theoretical minimum at 100 % efficiency: 357.1 + 15.32 ≃ 373 bytes (not yet attained by this library).
## Memory usage at aggressive settings (non-backtrackable input)
* 144 bytes of automatic storage for length tables (288 elements, 4 bits each, 100 % space efficiency)
* 384 bytes of automatic storage for two huffman trees (350 elements, 5…9 bits each, 88 % space efficiency)
* 24 bytes of temporary automatic storage while a huffman tree is being generated (15 elements, 9 bits each, 66 % space efficiency)
* An assortment of automatic variables for various purposes (may be register variables, depending on the architecture and of the compiler wits)
* ABI mandated alignment losses
Total: 528 bytes minimum, 552+N bytes maximum
Theoretical minimum at 100 % efficiency: 144 + 338.9 + 15.32 ≃ 499 bytes (not yet attained by this library).
## Memory usage at default settings (backtrackable input)
* 687 bytes of automatic storage for three huffman trees (52 % space efficiency)
* 30 bytes of temporary automatic storage while a huffman tree is being generated (53 % space efficiency)
* An assortment of automatic variables for various purposes (may be register variables, depending on the architecture and of the compiler wits)
* ABI mandated alignment losses
Total: 687 bytes minimum, 717+N bytes maximum
## Memory usage at default settings (non-backtrackable input)
* 288 bytes of automatic storage for length tables (50 % space efficiency)
* 653 bytes of automatic storage for two huffman trees (52 % space efficiency)
* 30 bytes of temporary automatic storage while a huffman tree is being generated (53 % space efficiency)
* An assortment of automatic variables for various purposes (may be register variables, depending on the architecture and of the compiler wits)
* ABI mandated alignment losses
Total: 941 bytes minimum, 971+N bytes maximum
## Tuning
To adjust the memory usage, there are three settings in gunzip.hh you can change:
| Setting name | 'false' memory use bytes | 'true' memory use bytes | 'true' performance impact
| ------------------------------------------- | ---:| ----:|--------------
| `USE_BITARRAY_TEMPORARY_IN_HUFFMAN_CREATION` | 30 | 24 | Negligible
| `USE_BITARRAY_FOR_LENGTHS` | 288 or 0 | 144 or 0 | Noticeable
| `USE_BITARRAY_FOR_HUFFNODES` | 653 or 687 | 384 or 408 | Significant
| **Total** | 971 or 717 | 552 or 432 | _Plus alignment losses, callframes and spills_
In addition, if you neither decompress into a raw memory area nor supply your own window function,
32768 bytes of automatic storage is allocated for the look-behind window.
You can also change the memory allocation scheme:
| `#define` name | Meaning
| --- | ---
| `DEFLATE_ALLOCATION_AUTOMATIC` | Automatic allocation (usually stack)
| `DEFLATE_ALLOCATION_STATIC` | Static `thread_local` allocation (memory remains allocated throughout the program, and different threads have their own copy of the data). Note that this scheme excludes running multiple decompressions in parallel, unless you do it in different threads.
| `DEFLATE_ALLOCATION_DYNAMIC` | Storage duration is the same as with automatic allocation, but the `new` keyword is explicitly used (which usually means heap/bss allocation).
There is also a constant `MAX_WINDOW_SIZE`, which is 32768 by default,
but you can reduce it to use less memory for the automatically allocated
window in situations where one is allocated (see note 9 below).
Note that this value must not be smaller than the maximum backreference
distance used by your compressed data.
## Unrequirements
* No dynamic memory is allocated under any circumstances, unless your user-supplied functors do it, or you `#define DEFLATE_ALLOCATION_DYNAMIC`.
* Aside from assert() in assert.h and some template metaprogramming tools in type_traits, no standard library functions are used.
* No global variables.
* Compatible with -fno-exceptions -fno-rtti compilation.
* Option to compile without constant arrays.
## Rationale
* Embedded platforms (Arduino, STM32 etc).
* ROM hacking
## Caveats
* Decompressor only. Deflate and GZIP streams are supported.
* Slower than your average inflate function. The template uses densely bitpacked arrays, which require plenty of bit-shifting operations for every access.
* The code obviously performs best on 32-bit or 64-bit platforms. Platforms where 32-bit entities must be synthesized from a number of 8-bit entities are at a disadvantage.
* Decompressed data integrity is not verified. Any checksum fields are totally ignored.
* On most systems, automatic storage means ‘stack allocation’. Depending on your circumstances, you may want to change the memory allocation scheme. See the Tuning chapter for details.
## Definitions
```C++
struct DeflateTrackNoSize{};
struct DeflateTrackInSize{};
struct DeflateTrackOutSize{};
struct DeflateTrackBothSize{};
int/*exit status*/ Deflate(InputParams..., OutputParams..., DeflateTrackNoSize = {});
std::pair<int/*exit status*/, std::uint_fast64_t/*number of input bytes consumed*/>
Deflate(InputParams..., OutputParams..., DeflateTrackInSize); // (11)
std::pair<int/*exit status*/, std::uint_fast64_t/*number of output bytes generated*/>
Deflate(InputParams..., OutputParams..., DeflateTrackOutSize); // (12)
std::pair<int/*exit status*/, std::pair<std::uint_fast64_t/*in size*/, std::uint_fast64_t/*out size*/>>
Deflate(InputParams..., OutputParams..., DeflateTrackBothSize); // (13)
// A counter for sizes is only allocated if explicitly requested
// by using one of the former three tracking overloads.
```
`InputParams` may be one of the following sets of parameters:
* InputFunctor input `(5)` `(14)`
* InputIterator begin `(7)` `(14)`
* InputIterator begin, InputIterator end `(6)` `(14)`
* InputIterator begin, SizeType length `(8)` `(14)`
* BidirectionalIterator begin, SizeType length `(8)` `(15)`
* ForwardIterator begin `(7)` `(14)`
* BidirectionalIterator begin `(7)` `(15)`
* RandomAccessIterator begin `(7)` `(15)`
* ForwardIterator begin, ForwardIterator end `(6)` `(15)`
* BidirectionalIterator begin, BidirectionalIterator end `(6)` `(15)`
* RandomAccessIterator begin, RandomAccessIterator end `(6)` `(15)`
`OutputParams` may be one of the following sets of parameters:
* OutputFunctor output `(1)` `(9)`
* OutputFunctor output, WindowFunctor window `(2)`
* OutputIterator target `(9)`
* RandomAccessIterator target `(10)`
* RandomAccessIterator target, SizeType target_limit `(3)` `(10)`
* RandomAccessIterator target, RandomAccessIterator target_end `(4)` `(10)`
1) If the output functor (`output`) returns a `bool`, and the returned value is `true`, the decompression aborts with return value -3
without writing any more data.
2) If the output functor (`output`) returns a `bool`, and the returned value is `true`, the decompression aborts with return value -3
without writing any more data.
If the window function returns an integer type, and the returned value is other than 0, the decompression aborts with return value -4
without writing any more data.
If either the window function returns `void`, or the output functor does not return a `bool`, aborting on output-full will not be compiled.
3) If `target_limit` bytes have been written into `target` and the decompression is not yet complete, the decompression aborts with return value -3
without writing any more data.
4) If `target_begin == target_end`, the decompression aborts with return value -3
without writing any more data.
5) If the input functor (`input`) returns an integer type other than a `char`, `signed char`, or `unsigned char`,
and the returned value is smaller than 0 or larger than 255, the decompression aborts with return value -2
without reading any more data.
6) If `begin == end`, the decompression aborts with return value -2.
7) If the input iterator deferences into a value outside the 0 — 255 range, the decompression aborts with return value -2
without reading any more data.
8) If `length` bytes have been read from `begin` and the decompression is not yet complete, the decompression aborts with return value -2
without reading any more data.
9) A separate 32768-byte sliding window will be automatically and separately allocated for the decompression.
10) The output data buffer is assumed to persist during the call and doubles as the sliding window for the decompression.
11) The `first` field in the return value has the same meaning as the `int` type return value described earlier.
The `second` field in the return value contains the number of bytes that were consumed from the input.
12) The `first` field in the return value has the same meaning as the `int` type return value described earlier.
The `second` field in the return value contains the number of bytes that were written to the output.
13) The `first` field in the return value has the same meaning as the `int` type return value described earlier.
The `second.first` field in the return value contains the number of bytes that were consumed from the input.
The `second.second` field in the return value contains the number of bytes that were written to the output.
14) This method is non-backtrackable, and uses a bit more memory than the backtrackable ones.
15) This method is backtrackable, meaning that some bytes in the input may be read twice. It uses less memory than the non-backtrackable calls.
### Tips
Some of these definitions may be ambiguous.
If you hit a compiler error, choose a different call method.
To help distinguish between (`InputIterator`,`RandomAccessIterator`,`RandomAccessIterator`)
and (`ForwardIterator`,`ForwardIterator`,`OutputIterator`), make sure the input iterators
are _const_.
If you do multiple decompression calls in your program in different spots,
it may be wise to make sure they all use the same type of parameters,
to avoid having to instantiate multiple copies of `Deflate()`.
Lambda functors are an offender in this respect, because each lambda has a
unique type even if their contents and calling conventions are identical.
In the worst case, you can use `std::function` to wrap your calls
into a common interface. Check out this video for more about this topic: https://www.youtube.com/watch?v=rUB5Hlm9AaQ
## Requirements
```C++
// An InputFunctor has the following prototype,
// wherein type1 is convertible into unsigned char:
type1 input()
// An OutputFunctor has one of the following two prototypes,
// wherein type1 can accept unsigned int parameters in range 0-255:
void output(type1 byte_to_output)
bool output(type1 byte_to_output)
// A WindowFunctor has one of the following two prototypes,
// wherein type1 can accept unsigned int parameters in range 0-258,
// and type2 can accept unsigned int parameters:
void outputcopy(type1 length, type2 offs)
type2 outputcopy(type1 length, type2 offs)
// An InputIterator must have at least the following operations defined,
// where type1 is convertible into unsigned char:
const type1& operator*() const
InputIterator& operator++()
// A OutputIterator must have at least the following operations defined,
// where type1 is convertible into unsigned char:
type1& operator*() const
OutputIterator& operator++()
// A ForwardIterator must have at least the following operations defined,
// where type1 is convertible into unsigned char:
const type1& operator*() const
ForwardIterator& operator++()
bool operator==(const ForwardIterator&) const
// A RandomAccessIterator must have at least the following operations defined,
// where type1 is convertible into unsigned char,
// and type2 is a signed integer type (may be negative):
type1& operator*()
type1& operator[] (type2)
RandomAccessIterator operator- (type2)
RandomAccessIterator& operator++()
bool operator==(const RandomAccessIterator&) const
```
## Example use:
Decompress the standard input into the standard output (uses 32 kB automatically allocated window):
```C++
Deflate([]() { return std::getchar(); },
[](unsigned char byte) { std::putchar(byte); });
// Or more simply:
Deflate(std::getchar, std::putchar);
```
Decompress an array containing gzipped data into another array that must be large enough to hold the result.
A window buffer will not be allocated.
```C++
extern const char compressed_data[];
extern unsigned char outbuffer[131072];
Deflate(compressed_data+0, outbuffer+0);
```
Same as above, but with range checking for output, and reporting of written size:
```C++
extern const char compressed_data[];
extern unsigned char outbuffer[131072];
auto result = Deflate(compressed_data+0, outbuffer+0, sizeof(outbuffer), DeflateTrackOutSize{});
if(result.first != 0) std::fprintf(stderr, "Error %d\n", result.first);
std::fprintf(stderr, "%u bytes written\n", unsigned(result.second));
```
Same as above, but with range checking for both input and output:
```C++
extern const char compressed_data[];
extern unsigned compressed_data_length;
extern unsigned char outbuffer[131072];
int result = Deflate(compressed_data+0, compressed_data_length, outbuffer, outbuffer + sizeof(outbuffer));
if(result != 0) std::fprintf(stderr, "Error\n");
```
Decompress using a custom window function (the separate 32 kB window buffer will not be allocated):
```C++
std::vector<unsigned char> result;
Deflate(std::getchar,
[&](unsigned byte)
{
result.push_back(byte);
},
[&](unsigned length, unsigned offset)
{
if(!length)
{
// offset contains the maximum look-behind distance.
// You could use this information to allocate a buffer of a particular size.
// length=0 case is invoked exactly once before any length!=0 cases are.
}
while(length-- > 0)
{
result.push_back( result[result.size()-offset] );
}
});
```
Same as above, but stop decompressing once 4096 bytes have been written:
```C++
std::vector<unsigned char> result;
Deflate(std::getchar,
[&](unsigned byte)
{
if(result.size() >= 4096) return true;
result.push_back(byte);
return false;
},
[&](unsigned length, unsigned offset)
{
if(!length)
{
// offset contains the maximum look-behind distance.
// You could use this information to allocate a buffer of a particular size.
// length=0 case is invoked exactly once before any length!=0 cases are.
}
for(; result.size() < 4096 && length > 0; --length)
{
result.push_back( result[result.size()-offset] );
}
return length;
});
```
## Misnomer
Yes, I am aware that the project is technically named misleadingly.
This project implements the _inflate_ algorithm (decompression),
not _deflate_ (compression).
In my defense, the _compression format_ is called deflate. There is no _inflate_ format.
This library decompresses data that has been compressed with _deflate_.
Think name, not verb.
================================================
FILE: gunzip.hh
================================================
/* My tiny gzip decompressor without using zlib. - Joel Yliluoma
* http://iki.fi/bisqwit/ , http://youtube.com/user/Bisqwit
* Inspired and influenced by a 13th IOCCC winner program by Ron McFarland */
/* Further optimized based on ideas from tinf library by Joergen Ibsen */
/** @file gunzip.hh @brief TinyDeflate */
/* Fun fact: Contains zero new/delete, and no STL data structures */
/* Distributed under the terms of the Zlib license:
Copyright (C) 2018 Joel Yliluoma
This software is provided 'as-is', without any express or implied
warranty. In no event will the authors be held liable for any damages
arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not
claim that you wrote the original software. If you use this software
in a product, an acknowledgment in the product documentation would be
appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be
misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
#include <assert.h>
#include <utility> // std::forward
#include <cstdint> // integer sizes
#include <type_traits>
#include <iterator>
#if !1 //Documentation purposes only; the actual prototypes are littered with std::enable_ifs.
/// Deflate(): This is the public method declared (later) in this file.
/// Decompresses (inflates) deflate-compressed data, with a gzip or deflate header.
/// User-supplied functors:
/// @param input() returns the next byte from the (compressed) input.
/// @param output(byte) outputs one uncompressed byte.
/// @param outputcopy(length, offset) copies length uncompressed bytes from offset,
/// Offset is always >= 1.
/// offset 1 means previous byte,
/// offset 2 means previous before that and so on.
/// Note that (offset < length) is not an error and in fact happens frequently.
/// If length=0, offset indicates the largest look-behind window length that
/// you need to be prepared for. The length is a power-of-two in range 256..32768.
//
/// If you want to implement range checking in input, return a negative value
/// from input() when there is no more input.
//
/// If you want to implement range checking in output, add a return value
/// in output(): false=ok, true=abort; and a return value in outputcopy():
/// 0=ok, nonzero=one or more bytes were not writable.
//
/// @returns:
/// 0 = decompression complete
/// -1 = data error
/// -2 = input functor returned a value outside 0x00..0xFF range
/// -3 = output functor returned nonzero / bool true value
/// -4 = outputcopy functor returned nonzero remaining length value
//
template<typename InputFunctor, typename OutputFunctor, typename WindowFunctor>
int Deflate(InputFunctor&& input, OutputFunctor&& output, WindowFunctor&& outputcopy);
/// Check README.md for the full list of versions of Deflate() available.
#endif
struct DeflateTrackTagBase{};
struct DeflateTrackNoSize: public DeflateTrackTagBase{};
struct DeflateTrackInSize: public DeflateTrackTagBase{};
struct DeflateTrackOutSize: public DeflateTrackTagBase{};
struct DeflateTrackBothSize: public DeflateTrackTagBase{};
/// The rest of the file is just for the curious about implementation.
#ifndef DOXYGEN_SHOULD_SKIP_THIS
namespace gunzip_ns
{
//#define DO_DEFDB_DUMPING
// If you want more performance at the expense of RAM use,
// Turn one or more of these settings to false:
static constexpr bool USE_BITARRAY_TEMPORARY_IN_HUFFMAN_CREATION = false; /* 8 bytes save */
static constexpr bool USE_BITARRAY_FOR_LENGTHS = false; /* 160 bytes save */
static constexpr bool USE_BITARRAY_FOR_HUFFNODES = false; /* 392 bytes save */
static constexpr unsigned MAX_WINDOW_SIZE = 32768u;
static_assert(MAX_WINDOW_SIZE >= 1, "Max window size should be >= 1");
static_assert(MAX_WINDOW_SIZE <= 32768u, "Window sizes larger than 32768 are not supported by deflate standard. Edit the source code to remove this assert if you need it.");
//
#define DEFLATE_USE_DATA_TABLES
#if !defined(DEFLATE_ALLOCATION_AUTOMATIC) && !defined(DEFLATE_ALLOCATION_STATIC) && !defined(DEFLATE_ALLOCATION_DYNAMIC)
// Choose one:
#define DEFLATE_ALLOCATION_AUTOMATIC
//#define DEFLATE_ALLOCATION_STATIC
//#define DEFLATE_ALLOCATION_DYNAMIC
#endif
constexpr unsigned Flag_InputAbortable = 0x01;
constexpr unsigned Flag_OutputAbortable = 0x02;
constexpr unsigned Flag_TrackIn = 0x40;
constexpr unsigned Flag_TrackOut = 0x80;
constexpr unsigned Flag_NoTrackFlagMask = 0x03;
}
#ifdef DEFLATE_ALLOCATION_DYNAMIC
# include <memory>
#endif
// RandomAccessBitArray: An engine for arrays of data items that are not byte-aligned
template<typename U = std::uint_least64_t>
struct RandomAccessBitArrayBase
{
private:
static constexpr unsigned Ubytes = sizeof(U), Ubits = Ubytes*8;
static std::uint_fast64_t Get_Unclean(unsigned Size, const U* data, unsigned index) throw()
{
unsigned bitpos = index*Size, unitpos = bitpos / Ubits, shift = bitpos % Ubits;
std::uint_fast64_t result = data[unitpos] >> shift;
//assert(Size <= sizeof(result)*8);
unsigned acquired = Ubits - shift;
for(; acquired < Size; acquired += Ubits)
{
result += (std::uint_fast64_t)data[++unitpos] << acquired;
}
return result;
}
public:
template<unsigned Size>
static std::uint_fast64_t Get(const U* data, unsigned index) throw()
{
std::uint_fast64_t result = Get_Unclean(Size,data,index);
return (Size >= sizeof(result)*8) ? result : (result & ((std::uint64_t(1) << Size)-1));
}
template<unsigned Size, bool update = false>
static void Set(U* data, unsigned index, std::uint_fast64_t value) throw()
{
unsigned bitpos = index*Size, unitpos = bitpos / Ubits, eat = 0;
// Make sure our storage unit is at least as bit as value
//assert(Ubits >= sizeof(value)*8);
//assert(Size >= sizeof(value)*8 || value < (std::uint64_t(1) << Size));
if(Size % Ubits != 0)
{
unsigned shift = bitpos % Ubits;
eat = Ubits - shift; if(eat > Size) eat = Size;
//assert(eat < sizeof(std::uint_fast64_t)*8);
//assert(shift + eat <= Ubits);
std::uint_fast64_t vmask = (std::uint64_t(1) << eat)-1;
if(update)
data[unitpos] = (data[unitpos] & ~(vmask << shift)) | (value << shift);
else
data[unitpos] |= value << shift;
//assert(eat < sizeof(value)*8);
value >>= eat;
++unitpos;
}
if(eat < Size)
for(unsigned remain = Size-eat; ; ++unitpos)
{
eat = Ubits;
if(eat > remain)
{
eat = remain;
if(update)
{
std::uint_fast64_t vmask = ((std::uint64_t(1) << eat)-1);
data[unitpos] = (data[unitpos] & ~vmask) | value;
}
else
{
data[unitpos] |= value;
}
break;
}
else
{
data[unitpos] = value;
value >>= Ubits/2; value >>= Ubits/2;
remain -= Ubits;
if(!remain) break;
}
}
}
};
template<unsigned Nbits, typename U = std::uint_least64_t>
struct RandomAccessBitArray
{
static constexpr unsigned Ubytes = sizeof(U), Ubits = Ubytes*8, Nunits = (Nbits+Ubits-1)/Ubits;
U data[Nunits];
template<unsigned Size>
inline std::uint_fast64_t Get(unsigned index) const throw()
{
return RandomAccessBitArrayBase<U>::template Get<Size>(data, index);
}
template<unsigned Size, bool update = false>
inline void Set(unsigned index, std::uint_fast64_t value) throw()
{
RandomAccessBitArrayBase<U>::template Set<Size,update>(data, index, value);
}
};
namespace gunzip_ns
{
struct dummy{};
/// Utility: ceil(log2(n))
template<unsigned long N> struct CeilLog2_s{ static constexpr unsigned result = 1+CeilLog2_s<(N+1)/2>::result; };
template<> struct CeilLog2_s<0> { static constexpr unsigned result = 0; };
template<> struct CeilLog2_s<1> { static constexpr unsigned result = 0; };
template<unsigned long N> static constexpr unsigned CeilLog2 = CeilLog2_s<N>::result;
/// Utility: floor(log2(n))
template<unsigned long N> struct FloorLog2_s{ static constexpr unsigned result = 1+FloorLog2_s<N/2>::result; };
template<> struct FloorLog2_s<0> { static constexpr unsigned result = 0; };
template<> struct FloorLog2_s<1> { static constexpr unsigned result = 0; };
template<unsigned long N> static constexpr unsigned FloorLog2 = FloorLog2_s<N>::result;
/// Utility: smallest unsigned integer type that can store n-bit value
template<unsigned bits>
using SmallestType = std::conditional_t< (bits<=16),
std::conditional_t< (bits<= 8), std::uint_least8_t, std::uint_least16_t>,
std::conditional_t< (bits<=32), std::uint_least32_t, std::uint_least64_t>>;
/// testcases
static_assert(FloorLog2<1> == 0, "FloorLog2 fail"); static_assert(CeilLog2<1> == 0, "CeilLog2 fail");
static_assert(FloorLog2<2> == 1, "FloorLog2 fail"); static_assert(CeilLog2<2> == 1, "CeilLog2 fail");
static_assert(FloorLog2<3> == 1, "FloorLog2 fail"); static_assert(CeilLog2<3> == 2, "CeilLog2 fail");
static_assert(FloorLog2<4> == 2, "FloorLog2 fail"); static_assert(CeilLog2<4> == 2, "CeilLog2 fail");
static_assert(FloorLog2<5> == 2, "FloorLog2 fail"); static_assert(CeilLog2<5> == 3, "CeilLog2 fail");
static_assert(FloorLog2<6> == 2, "FloorLog2 fail"); static_assert(CeilLog2<6> == 3, "CeilLog2 fail");
static_assert(FloorLog2<7> == 2, "FloorLog2 fail"); static_assert(CeilLog2<7> == 3, "CeilLog2 fail");
static_assert(FloorLog2<8> == 3, "FloorLog2 fail"); static_assert(CeilLog2<8> == 3, "CeilLog2 fail");
static_assert(FloorLog2<9> == 3, "FloorLog2 fail"); static_assert(CeilLog2<9> == 4, "CeilLog2 fail");
template<bool packed, unsigned Dimension, unsigned ElementSize>
struct RandomAccessArray {};
template<unsigned Dim, unsigned Elem>
struct RandomAccessArray<true, Dim, Elem>
{
RandomAccessBitArray<Dim*Elem> impl;
inline std::uint_fast64_t Get(unsigned index) const { return impl.template Get<Elem>(index); }
inline void Set(unsigned index, std::uint_fast32_t value) { impl.template Set<Elem,true>(index, value); }
inline void QSet(unsigned index, std::uint_fast32_t value) { impl.template Set<Elem,false>(index, value); }
template<unsigned Bits>
inline void WSet(unsigned index, std::uint_fast64_t value) { impl.template Set<Bits,true>(index, value); }
};
template<unsigned Dim, unsigned Elem>
struct RandomAccessArray<false, Dim, Elem>
{
typedef SmallestType<Elem> E;
E data[Dim];
inline E Get(unsigned index) const { return data[index]; }
inline void Set(unsigned index, E value) { data[index] = value; }
inline void QSet(unsigned index, E value) { data[index] = value; }
template<unsigned Bits>
inline void WSet(unsigned index, std::uint_fast64_t value)
{
index *= Bits/Elem;
for(unsigned b=0; b<Bits; b+=Elem, value>>=Elem)
QSet(index++, (value % (1u << Elem)));
}
};
}
namespace gunzip_ns
{
//#define DEFL_DO_HUFF_STATS
template<bool Abortable, unsigned A,unsigned B, typename LengthFunctor>
bool CreateHuffmanTree(const char*
#ifdef DEFL_DO_HUFF_STATS
why
#endif
, RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES,A,B>& tree,
unsigned num_values,
LengthFunctor&& ReadLength) throw()
{
/* Lengths[] needs to be scanned exactly twice, in forward order.
* Technically we could use a functor instead of a table,
* but this would require that the dynamic tree generator
* can read sections of the compressed data twice,
* which we currently do not support.
*/
constexpr unsigned ElemBits = CeilLog2<A-15>; // ceil(log2(A-15)) where A-15 is max value of num_values
static_assert((1u << B) >= (A-15), "B is too small");
assert(num_values <= (A-15));
RandomAccessArray<USE_BITARRAY_TEMPORARY_IN_HUFFMAN_CREATION, 15, ElemBits> offs{}; // 24 or 16 bytes.
// Theoretically 15.32 bytes for 288 num_values, 9.375 for 32 num_values, 7.97 for 19 num_values.
// Clear code length count table
tree.template WSet<(15*B + 63) & ~63>(0, 0); // First 15 needed, but round to nice unit
// Scan symbol length, and sum code length counts
#ifdef DEFL_DO_HUFF_STATS
unsigned largest_treetable_value = 0, largest_offs = 0, smallest_treetable_value = ~0u;
unsigned largest_treetrans_index=0, largest_treetrans_value=0;
unsigned longest_length = 0;
#endif
for(unsigned a = 0; a < num_values; ++a)
{
int length = ReadLength(a); // Note: Can be zero.
if(Abortable && length < 0) return true;
if(length)
{
unsigned v = tree.Get(0 + length-1)+1;
#ifdef DEFL_DO_HUFF_STATS
largest_treetable_value = std::max(largest_treetable_value, v);
longest_length = std::max(longest_length, unsigned(length));
#endif
//fprintf(stderr, " [%d]%3d CLL (val: %d)\n", length, v, v);
tree.Set(0 + length-1, v);
}
else
{
//fprintf(stderr, " [_]%3d CLL (val: 0)\n", 0);
}
}
// Compute offset table for distribution sort
for(unsigned sum=0, a = 1; a < 16; ++a)
{
offs.QSet(a-1, sum); // starting offset for values that have length "a"
sum += tree.Get(0 + a-1); // number of values that have length "a"
}
#ifdef DEFL_DO_HUFF_STATS
for(unsigned a=1; a<=longest_length; ++a)
smallest_treetable_value = std::min(smallest_treetable_value, (unsigned)tree.Get(0 + a-1));
#endif
// Create code->symbol translation table (symbols sorted by code)
for(unsigned value = 0; value < num_values; ++value)
{
int length = ReadLength(value); // Note: Can be zero.
if(Abortable && length < 0) return true;
if(length)
{
unsigned q = offs.Get(length-1); offs.Set(length-1, q+1); // q = offset[length]++;
#ifdef DEFL_DO_HUFF_STATS
largest_offs = std::max(q, largest_offs);
largest_treetrans_index = std::max(largest_treetrans_index, q);
largest_treetrans_value = std::max(largest_treetrans_value, value);
#endif
assert(q < num_values /*&& value < num_values*/);
//fprintf(stderr, " [x]%3d CLL %d\n", 15+q, value);
tree.Set(15 + q, value);
}
}
#ifdef DEFL_DO_HUFF_STATS
std::fprintf(stderr, "Largest \"%12s\"(treetable_value=%4u..%4u, offs=%4u, treetrans_index=%4u, treetrans_value=%4u)\n",
why, smallest_treetable_value,largest_treetable_value,
largest_offs, largest_treetrans_index, largest_treetrans_value);
#endif
// Largest values observed in the wild:
// Dyn Lengths: Max treetable_value =255, max offs =285, max treetrans_index =285, max treetrans_value =285
// Stat Lengths:Max treetable_value =152, max offs =287, max treetrans_index =287, max treetrans_value =287
// Len Lengths: Max treetable_value = 13, max offs = 18, max treetrans_index = 18, max treetrans_value = 18
// Dyn Dists: Max treetable_value = 19, max offs = 29, max treetrans_index = 29, max treetrans_value = 29
// Stat Dists: Max treetable_value = 32, max offs = 31, max treetrans_index = 31, max treetrans_value = 31
return false;
}
#ifdef DEFLATE_USE_DATA_TABLES
template<bool=0> // Using a dummy template parameter makes this function and its data weak,
inline const std::uint_least8_t* GetBTable() throw() // removing linker problems in multi-module use
{
static const std::uint_least8_t data[] {
// Length bases (0-31)
0,1,2,3,4,5,6,7,8,10,12,14,16,20,24,28,32,40,48,56,64,80,96,112,128,160,192,224,255, 0,0,0,
// Length bits and distance bits (29-60) (overlap 3 bytes)
// 0x00,0x01,0x01,0x02,0x02,0x13,0x13,0x14,0x14,0x25,0x25,0x26,0x26,
//0x37,0x37,0x38,0x38,0x49,0x49,0x4A,0x4A,0x5B,0x5B,0x5C,0x5C,0x0D,0x0D,0x00,0x00
// Reverse-order table
3*3,17*3,15*3,13*3,11*3,9*3,7*3,5*3,4*3,6*3,8*3,10*3,12*3,14*3,16*3,18*3,0*3,1*3,2*3
};
return data;
}
//template<bool=0>
//inline const std::uint_least16_t* GetWTable() throw()
//{
// static const std::uint_least16_t data[32] {
// 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577, 0,0 };
// return data;
//}
//inline unsigned dbase(unsigned distcode) { return GetWTable<>()[distcode]; }
inline unsigned lbase(unsigned lencode) { return GetBTable<>()[lencode-257+0] + 3; }
//inline unsigned dbits(unsigned distcode) { return GetBTable<>()[distcode+29] & 0xF; }
//inline unsigned lbits(unsigned lencode) { return GetBTable<>()[lencode-257+29] >> 4; }
inline unsigned rshift(unsigned a) { return GetBTable<>()[a + 32]; }
#else
inline unsigned lbase(unsigned lencode) { return (lencode > 285 ? 3 : ((lencode >= 265) ? (((lencode-257)%4+4) << ((lencode-257)/4-1)) + (lencode==285 ? 2 : 3) : (lencode-254))); }
inline unsigned rshift(unsigned a) { if(!a) return 3*3; else if(a>=16) return (a-16)*3; int r = 12 + 7*(a<8) - a*2; return (r<0 ? -r : r)*3; }
#endif
inline unsigned dbase(unsigned distcode) { return (1 + (distcode>=4 ? ((2+distcode%2) << (distcode/2-1)) : distcode)); }
inline unsigned dbits(unsigned distcode) { return distcode>=4 ? distcode/2-1 : 0; }
inline unsigned lbits(unsigned lencode) { return ((lencode>=265 && lencode<285) ? ((lencode-265)/4+1) : 0); }
//inline unsigned order(unsigned index) { return index<3 ? (index+16) : ((index%2) ? (1-index/2)&7 : (6+index/2)); }
// Cortex-M0+ Cortex-M4 x86_64
// dbase with table 12+64 = 76 12+64 = 76 14+64 = 78
// dbase with func 24 22 26
// lbase with table 12+32 = 48 12+32 = 48 21+64 = 76
// lbase with func 54 56 64
// dbits+lbits with table 12+16+29 = 57 12+16+29 = 57 17+21+29 = 67
// dbits+lbits with func 14+18 = 32 14+18 = 32 13+20 = 33
// Support for pre-c++20
template<typename T>
using remove_cvref_t = std::remove_reference_t<std::remove_cv_t<T>>;
// Support for pre-c++20 (result_of is removed in c++20, invoke_result is added in c++17, so neither can be used exclusively)
template <class T>
struct result_of { // explain usage
static_assert((T)false, "result_of<CallableType> is invalid; use "
"result_of<CallableType(zero or more argument types)> instead.");
};
#if __cplusplus > 202000UL
template <typename F, typename... Args>
struct result_of<F(Args...)> : std::invoke_result<F, Args...> {};
#else
template <typename F, typename... Args>
struct result_of<F(Args...)> : std::result_of<F(Args...)> {};
#endif
template <class T>
using result_of_t = typename result_of<T>::type;
#define BEGIN_COND(name) \
template<typename T, typename=void> struct name : public std::false_type {}; \
template<typename T> struct name<T, std::enable_if_t<
#define END_COND(name) \
, void>> : public std::true_type {}; \
template<typename T> \
inline constexpr bool name ## _v = name<T>::value; \
// Input parameter condition testers:
BEGIN_COND(is_input_functor)
std::is_convertible_v<result_of_t<remove_cvref_t<T>()>,int>
END_COND(is_input_functor)
BEGIN_COND(is_input_iterator)
std::is_convertible_v<typename std::iterator_traits<remove_cvref_t<T>>::value_type, unsigned char>
&& std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::input_iterator_tag>
END_COND(is_input_iterator)
BEGIN_COND(is_bidir_input)
std::is_convertible_v<typename std::iterator_traits<remove_cvref_t<T>>::value_type, unsigned char>
&& (std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::forward_iterator_tag>
|| std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::bidirectional_iterator_tag>
|| std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::random_access_iterator_tag>)
END_COND(is_bidir_input)
BEGIN_COND(is_size_type)
std::is_convertible_v<remove_cvref_t<T>, std::size_t> && !std::is_pointer_v<remove_cvref_t<T>>
END_COND(is_size_type)
// Output parameter condition testers:
BEGIN_COND(is_random_iterator)
std::is_convertible_v<typename std::iterator_traits<remove_cvref_t<T>>::value_type, unsigned char>
&& !std::is_const_v<typename std::iterator_traits<remove_cvref_t<T>>::reference>
&& std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::random_access_iterator_tag>
END_COND(is_random_iterator)
BEGIN_COND(is_output_iterator)
std::is_convertible_v<typename std::iterator_traits<remove_cvref_t<T>>::value_type, unsigned char>
&& !std::is_const_v<typename std::iterator_traits<remove_cvref_t<T>>::reference>
&& !std::is_pointer_v<remove_cvref_t<T>>
&& (std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::output_iterator_tag>
|| std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::forward_iterator_tag>
|| std::is_same_v<typename std::iterator_traits<remove_cvref_t<T>>::iterator_category, std::bidirectional_iterator_tag>)
END_COND(is_output_iterator)
// Output functor & window functor: Returns void or something that can be converted to bool
BEGIN_COND(is_output_functor)
std::is_same_v<result_of_t<remove_cvref_t<T>(int)>,void> || std::is_convertible_v<result_of_t<remove_cvref_t<T>(int)>,bool>
END_COND(is_output_functor)
BEGIN_COND(is_window_functor)
std::is_same_v<result_of_t<remove_cvref_t<T>(int,int)>,void> || std::is_convertible_v<result_of_t<remove_cvref_t<T>(int,int)>,int>
END_COND(is_window_functor)
BEGIN_COND(is_abortable_input_type)
!(std::is_same_v<T, unsigned char> || std::is_same_v<T, signed char> || std::is_same_v<T, char>)
END_COND(is_abortable_input_type)
#undef END_COND
#undef BEGIN_COND
template<typename T>
constexpr bool DeflAbortable_InFun = is_abortable_input_type_v<remove_cvref_t<result_of_t<T()>>>;
template<typename T>
constexpr bool DeflAbortable_OutFun = std::is_same_v<result_of_t<T(int)>, bool>;
template<typename T>
constexpr bool DeflAbortable_WinFun = std::is_convertible_v<remove_cvref_t<result_of_t<T(int,int)>>, int>;
template<bool Abortable>
struct OutputHelper
{
template<typename OutputFunctor>
static inline bool output(OutputFunctor&& output, unsigned char byte)
{
output(byte);
return false;
}
template<typename WindowFunctor>
static inline bool outputcopy(WindowFunctor&& outputcopy, std::uint_least16_t length, std::uint_fast32_t offset)
{
outputcopy(length, offset);
return false;
}
};
template<>
struct OutputHelper<true>
{
template<typename OutputFunctor>
static inline bool output(OutputFunctor&& output, unsigned char byte)
{
return output(byte);
}
template<typename WindowFunctor>
static inline bool outputcopy(WindowFunctor&& outputcopy, std::uint_least16_t& length, std::uint_fast32_t offset)
{
length = outputcopy(length, offset);
return length != 0;
}
};
struct SizeTracker_NoOutput
{
inline void OutByte() { }
inline void OutBytes(std::uint_fast64_t) { }
// Dummy forwarders. Do the same as std::forward.
template<typename T>
static inline constexpr T&& ForwardOutput(std::remove_reference_t<T>& fun) { return static_cast<T&&>(fun); }
template<typename T>
static inline constexpr T&& ForwardOutput(std::remove_reference_t<T>&& fun) { return static_cast<T&&>(fun); }
template<typename T>
static inline constexpr T&& ForwardWindow(std::remove_reference_t<T>& fun) { return static_cast<T&&>(fun); }
template<typename T>
static inline constexpr T&& ForwardWindow(std::remove_reference_t<T>&& fun) { return static_cast<T&&>(fun); }
};
struct SizeTracker_NoInput
{
inline void InByte() { }
inline void InBytes(std::uint_fast64_t) { }
template<typename T>
static inline constexpr T&& ForwardInput(std::remove_reference_t<T>& fun) { return static_cast<T&&>(fun); }
template<typename T>
static inline constexpr T&& ForwardInput(std::remove_reference_t<T>&& fun) { return static_cast<T&&>(fun); }
};
struct SizeTracker_DoInput
{
std::uint_fast64_t insize=0;
inline void InByte() { ++insize; }
inline void InBytes(std::uint_fast64_t n) { insize += n; }
template<typename InputFunctor, std::enable_if_t<!DeflAbortable_InFun<InputFunctor>,gunzip_ns::dummy>...>
auto ForwardInput(const InputFunctor& input)
{
return [&]() { InByte(); return input(); };
}
template<typename InputFunctor, std::enable_if_t<DeflAbortable_InFun<InputFunctor>,gunzip_ns::dummy>...>
auto ForwardInput(const InputFunctor& input)
{
return [&]() { auto r = input(); if(!(r & ~0xFF)) { InByte(); } return r; };
}
};
struct SizeTracker_DoOutput
{
std::uint_fast64_t outsize=0;
inline void OutByte() { ++outsize; }
inline void OutBytes(std::uint_fast64_t n) { outsize += n; }
template<typename OutputFunctor, std::enable_if_t<!DeflAbortable_OutFun<OutputFunctor>,gunzip_ns::dummy>...>
auto ForwardOutput(const OutputFunctor& output)
{
return [&](unsigned char byte) { OutByte(); return output(byte); };
}
template<typename OutputFunctor, std::enable_if_t<DeflAbortable_OutFun<OutputFunctor>,gunzip_ns::dummy>...>
auto ForwardOutput(const OutputFunctor& output)
{
return [&](unsigned char byte) { auto r = output(byte); if(!r) { OutByte(); } return r; };
}
template<typename WindowFunctor, std::enable_if_t<!DeflAbortable_WinFun<WindowFunctor>,gunzip_ns::dummy>...>
auto ForwardWindow(const WindowFunctor& outputcopy)
{
return [&](std::uint_least16_t length, std::uint_fast32_t offset)
{
OutBytes(length);
return outputcopy(length, offset);
};
}
template<typename WindowFunctor, std::enable_if_t<DeflAbortable_WinFun<WindowFunctor>,gunzip_ns::dummy>...>
auto ForwardWindow(const WindowFunctor& outputcopy)
{
return [&](std::uint_least16_t length, std::uint_fast32_t offset)
{
auto remain = outputcopy(length, offset);
OutBytes(length - remain);
return remain;
};
}
};
template<typename TrackType>
struct SizeTracker: public SizeTracker_NoOutput, public SizeTracker_NoInput
{
inline int operator() (int returncode) const { return returncode; }
};
template<>
struct SizeTracker<DeflateTrackOutSize>: public SizeTracker_NoInput, public SizeTracker_DoOutput
{
typedef std::pair<int,std::uint_fast64_t> result;
inline result operator() (int returncode) const { return result{returncode,outsize}; }
};
template<>
struct SizeTracker<DeflateTrackInSize>: public SizeTracker_NoOutput, public SizeTracker_DoInput
{
typedef std::pair<int,std::uint_fast64_t> result;
inline result operator() (int returncode) const { return result{returncode,insize}; }
};
template<>
struct SizeTracker<DeflateTrackBothSize>: public SizeTracker_DoInput, public SizeTracker_DoOutput
{
typedef std::pair<int, std::pair<std::uint_fast64_t,std::uint_fast64_t>> result;
inline result operator() (int returncode) const { return result{returncode,std::make_pair(insize,outsize)}; }
};
#ifdef DO_DEFDB_DUMPING
unsigned bitcounter=0;
#endif
struct DeflateBitCache
{
std::uint_least8_t BitCache = 0, BitCount = 0;
template<bool Abortable, typename InputFunctor>
std::uint_least64_t GetBits(InputFunctor&& input, unsigned numbits)
{
#ifdef DO_DEFDB_DUMPING
bitcounter += numbits;
#endif
std::uint_fast64_t result = BitCache;
if(numbits <= BitCount)
{
BitCount -= numbits;
BitCache >>= numbits;
result &= ((1u << numbits)-1); // 0-8
return result;
}
for(unsigned acquired = BitCount; ; acquired += 8)
{
unsigned byte = input();
if(Abortable && (byte & ~0xFFu))
{
// Note: Throws away bits already eaten from BitCache
return ~std::uint_least64_t(0); // error
}
unsigned eat = numbits-acquired;
byte &= 0xFF;
if(eat <= 8)
{
result |= ((std::uint_fast64_t)(byte & ((1u << eat)-1))) << acquired;
BitCount = 8-eat;
BitCache = byte >> eat;
return result;
}
result |= ((std::uint_fast64_t)(byte)) << acquired;
}
}
template<bool Abortable, typename InputFunctor, unsigned A,unsigned B>
std::uint_least32_t HuffRead(InputFunctor&& input,
RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES,A,B>& tree)
{
int sum=0, cur=0, len=0;
#ifdef DEFL_DO_HUFF_STATS
static int maxlen = 0;
#endif
// Get more bits while code value is above sum
do {
auto p = GetBits<Abortable>(std::forward<InputFunctor>(input), 1);
if(Abortable && !~p)
{
// Note: Throws away progress already made traversing the tree
return ~std::uint_least32_t(0); // error flag
}
cur = (unsigned(cur) << 1) | bool(p);
#ifdef DEFL_DO_HUFF_STATS
if(len > maxlen)
{
maxlen = len;
std::fprintf(stderr, "maxlen access: %d (%d)\n", maxlen, (int)tree.Get(0 + len));
}
#endif
auto v = tree.Get(0 + len++);
sum += v;
cur -= v;
} while(cur >= 0);
return tree.Get(15 + sum + cur);
}
};
template<bool Backtrackable>
struct DeflateState: public DeflateBitCache
{
std::uint_least8_t lencode = 0, num = 0; // used in DynTreeFunc
// Lengths are in 0..15 range.
RandomAccessArray<USE_BITARRAY_FOR_LENGTHS, 288, CeilLog2<16>> Lengths; // 144 bytes
// Length tree
// Values up to 288 in indexes 0-14. (Table) (255 is max observed in wild)
// Values up to 287 in indexes 15-302. (Trans)
RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES, 15+288, CeilLog2<289>> ltree; // 341->344 bytes
// Distance tree
// Values up to 32 in indexes 0-14. (Table)
// Values up to 31 in indexes 15-46. (Trans)
RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES, 15+32, CeilLog2<33>> dtree; // 36->40 bytes
RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES, 15+32, CeilLog2<33>>& lltree = dtree;
// Theoretical minimum memory use:
// (15*log2(289) + 288*log2(288))/8 = 309.45 bytes for ltree
// (15*log2(33) + 32 *log2(32))/8 = 29.46 bytes for dtree
// 144.00 bytes for lengths
// total 482.91 bytes
template<bool Abortable, typename InputFunctor, typename BacktrackFunctor>
auto DynTreeFunc(InputFunctor&& input, std::uint_fast16_t length, BacktrackFunctor&& /*backtrack*/,
bool
#ifdef DO_DEFDB_DUMPING
dists
#endif
)
{
Lengths = {}; // clear at least length nibbles; easiest to clear it all
bool error = false;
for(std::uint_fast16_t code = 0; ; )
{
#ifdef DO_DEFDB_DUMPING
unsigned bits_before=bitcounter;
#endif
if(!num)
{
auto p = HuffRead<Abortable>(std::forward<InputFunctor>(input), lltree);
if(Abortable && !~p) { error = true; goto done; }
std::uint_least8_t what = p; // 0-18
if(!(what & 16)) { lencode = what * 0x11u; what = 0x01; } // 1 times (what < 16) (use what, set prev)
else if(what < 17) { lencode = (lencode >> 4) * 0x11u; what = 0x23; } // 3..6 (use prev)
else if(what == 17) { lencode = 0; what = 0x33; } // 3..10 (use 0, set prev)
else { lencode = 0; what = 0x7B; } // 11..138 (use 0, set prev)
p = GetBits<Abortable>(std::forward<InputFunctor>(input), what >> 4); // 0, 2, 3 or 7 bits
#ifdef DO_DEFDB_DUMPING
if(what>>4)
{
char tmp[64]="[_]"; sprintf(tmp, "[%d]", int(bitcounter-bits_before));
fprintf(stderr, "%4s %cREP (%d times)\n", tmp, (lencode&0xF)?'L':'Z', p+(what&0xF));
}
#endif
if(Abortable && !~p) { error = true; goto done; }
num = p + (what & 0xF); // 1..138
}
#ifdef DO_DEFDB_DUMPING
bool rep=num>1;
#endif
do {
#ifdef DO_DEFDB_DUMPING
char tmp[64]="[_]"; if(!rep) sprintf(tmp, "[%d]", int(bitcounter-bits_before));
if(code == 0x100)
fprintf(stderr, "%4s EofB CL (val:%2d)\n", tmp, int(lencode&0xF));
else if(dists)
fprintf(stderr, "%4s d_%02d CL (val:%2d)\n", tmp, int(code), int(lencode&0xF));
else if(code > 0x100)
fprintf(stderr, "%4s l_%02d CL (val:%2d)\n", tmp, int(code- 0x101), int(lencode&0xF));
else
fprintf(stderr, "%4s 0x%02X CL (val:%2d)\n", tmp, (int)code, int(lencode&0xF));
#endif
--num;
Lengths.QSet(code++, lencode & 0xF);
if(code == length) { goto done; }
} while(num > 0);
}
done:;
return [this,error](unsigned index) -> std::conditional_t<Abortable, int, unsigned char>
{
if(Abortable && error) return -1;
return Lengths.Get(index);
};
}
};
template<>
struct DeflateState<true>: public DeflateBitCache
{
// Length tree
// Values up to 288 in indexes 0-14. (Table) (255 is max observed in wild)
// Values up to 287 in indexes 15-302. (Trans)
RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES, 15+288, CeilLog2<289>> ltree; // 341->344 bytes
// Distance tree
// Values up to 32 in indexes 0-14. (Table)
// Values up to 31 in indexes 15-46. (Trans)
RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES, 15+32, CeilLog2<33>> dtree; // 36->40 bytes
// Length-lengths tree
// Values up to 19 in indexes 0-14. (Table) (13 is max observed in wild)
// Values up to 18 in indexes 15-33. (Trans)
RandomAccessArray<USE_BITARRAY_FOR_HUFFNODES, 15+19, CeilLog2<20>> lltree; // 22->24 bytes
// Theoretical minimum memory use:
// (15*log2(289) + 288*log2(288))/8 = 309.45 bytes for ltree
// (15*log2(33) + 32 *log2(32))/8 = 29.46 bytes for dtree
// (15*log2(20) + 19 *log2(19))/8 = 18.19 bytes for lltree
// total 357.10 bytes
std::uint_least8_t lencode, num; // used in DynTreeFunc
std::uint_least8_t checkpoint_lencode, checkpoint_num;
std::uint_least8_t checkpoint_BitCache, checkpoint_BitCount;
template<bool Abortable, typename InputFunctor, typename BacktrackFunctor>
auto DynTreeFunc(InputFunctor&& input, std::uint_fast16_t /*length*/, BacktrackFunctor&& backtrack,
bool
#ifdef DO_DEFDB_DUMPING
dists
#endif
)
{
// Create checkpoint
checkpoint_lencode = 0;
checkpoint_num = 0;
checkpoint_BitCache = BitCache;
checkpoint_BitCount = BitCount;
backtrack(false);
return [this,&input,&backtrack](unsigned index) -> std::conditional_t<Abortable, int, unsigned char>
{
if(index == 0)
{
// Restore checkpoint
lencode = checkpoint_lencode;
num = checkpoint_num;
BitCache = checkpoint_BitCache;
BitCount = checkpoint_BitCount;
backtrack(true);
}
if(Abortable && (num==0xFF)) return -1;
if(!num)
{
auto p = HuffRead<Abortable>(std::forward<InputFunctor>(input), lltree);
if(Abortable && !~p) { num = 0xFF; return -1; } // If p== ~uint64_t()
std::uint_least8_t what = p; // 0-18
if(!(what & 16)) { lencode = what * 0x11u; what = 0x01; } // 1 times (what < 16) (use what, set prev)
else if(what < 17) { lencode = (lencode >> 4) * 0x11u; what = 0x23; } // 3..6 (use prev)
else if(what == 17) { lencode = 0; what = 0x33; } // 3..10 (use 0, set prev)
else { lencode = 0; what = 0x7B; } // 11..138 (use 0, set prev)
p = GetBits<Abortable>(std::forward<InputFunctor>(input), what >> 4); // 0, 2, 3 or 7 bits
if(Abortable && !~p) { num = 0xFF; return -1; } // If p== ~uint64_t()
num = p + (what & 0xF); // 1..138
}
--num;
return (lencode & 0xF);
};
}
};
struct DeflateWindow
{
unsigned char Data[gunzip_ns::MAX_WINDOW_SIZE];
SmallestType<CeilLog2<gunzip_ns::MAX_WINDOW_SIZE>> Head = 0;
};
#ifdef DEFLATE_ALLOCATION_STATIC
template<typename ObjectType>
ObjectType& GetStaticObj()
{
static thread_local ObjectType obj;
obj.~ObjectType();
new(&obj) ObjectType();
return obj;
}
#endif
/* Values of Abortable:
* Input abortable = &1
* Output abortable = &2
* Resumable = &4
*
* Input abortable Output abortable Resumable Value
* no no no 0
* yes no no 1
* no yes no 2
* yes yes no 3
* 4 = invalid
* yes no yes 5
* no yes yes 6
* yes yes yes 7
*/
template<unsigned char Abortable,
typename InputFunctor, typename OutputFunctor, typename WindowFunctor,
typename BacktrackFunctor>
int Gunzip(InputFunctor&& input,
OutputFunctor&& output,
WindowFunctor&& outputcopy,
BacktrackFunctor&& backtrack)
{
using namespace gunzip_ns;
typedef DeflateState<!std::is_same_v<remove_cvref_t<BacktrackFunctor>,dummy>> StateType;
#ifdef DEFLATE_ALLOCATION_AUTOMATIC
StateType state;
#elif defined(DEFLATE_ALLOCATION_STATIC)
auto& state = gunzip_ns::GetStaticObj<StateType>();
#elif defined(DEFLATE_ALLOCATION_DYNAMIC)
std::unique_ptr<StateType> stateptr(new StateType);
auto& state = *stateptr;
#endif
// The following routines are macros rather than e.g. lambda functions,
// in order to make them inlined in the function structure, and breakable/resumable.
// Bit-by-bit input routine
#define DummyGetBits(numbits) do { \
auto p = state.template GetBits<bool(Abortable&Flag_InputAbortable)>(std::forward<InputFunctor>(input), numbits); \
if((Abortable & Flag_InputAbortable) && !~p) return -2; \
} while(0)
#define GetBits(numbits, target) \
auto p = state.template GetBits<bool(Abortable&Flag_InputAbortable)>(std::forward<InputFunctor>(input), numbits); \
if((Abortable & Flag_InputAbortable) && !~p) return -2; \
target = p
// Huffman tree read routine.
#define HuffRead(tree, target) \
auto p = state.template HuffRead<bool(Abortable&Flag_InputAbortable)>(std::forward<InputFunctor>(input), tree); \
if((Abortable & Flag_InputAbortable) && !~p) return -2; \
target = p
#define Fail_If(condition) do { \
/*assert(!(condition));*/ \
if(condition) return -1; \
} while(0)
// Read stream header
GetBits(16, std::uint_least16_t header);
// ^ Read deflate header: method[4] ; winsize[4] ; checksum[8]
if(header == 0x8B1F) // Is it actually a gzip header?
{
// Get format identifier, flags, MTIME, XFL and OS
GetBits(64, header); Fail_If((header & 0xFF) != 8); // Format identifier should be 8
if(header&0x0400) // Skip extra fields as indicated by FEXTRA
{ GetBits(16, std::uint_fast16_t q); DummyGetBits(8*q); }
if(header&0x0800) for(;;) { GetBits(8, bool q); if(!q) break; } // NAME: Skip filename if FNAME was present
if(header&0x1000) for(;;) { GetBits(8, bool q); if(!q) break; } // COMMENT: Skip comment if FCOMMENT was present
if(header&0x0200) { DummyGetBits(16); } // HCRC: Skip FCRC if was present
outputcopy(0, 32768u); // GZIP always uses 32k window
}
else // No. Deflate header?
{
Fail_If((header & 0x208F) != 0x0008 || ((((header<<8)+(header>>8))&0xFFFF)%31) != 0);
outputcopy(0, 256 << ((header >> 4) & 0xF)); // Preset dictionary (0x2000) is not supported
}
// Read compressed blocks
for(;;)
{
GetBits(3, header);
//fprintf(stderr, "header=%d\n", header);
if(header & 4) // Dynamic block
{
Fail_If(header & 2);
std::uint_least16_t nlen_ndist_ncode;
GetBits(14, nlen_ndist_ncode);
#define nlen (((nlen_ndist_ncode >> 0u) & 0x1Fu) + 257u) // 257..288
#define ndist (((nlen_ndist_ncode >> 5u) & 0x1Fu) + 1u) // 1..32
std::uint_least8_t ncode = ((nlen_ndist_ncode >> 10u) + 4u); // 4..19
{std::uint_fast64_t lenlens; GetBits(ncode*3, lenlens); // Max: 19*3 = 57 bits
#ifdef DO_DEFDB_DUMPING
fprintf(stderr, " [5] HLIT%5d (val:%d)\n [5] HDIST%4d (val:%d)\n [4] HCLEN%4d (val:%d)\n",
nlen,nlen-257, ndist,ndist-1, ncode,ncode-4);
for(unsigned a=0; a<19; ++a)
for(unsigned b=0; b<19; ++b)
if(rshift(b) == 3*a)
{
if(a < ncode)
fprintf(stderr, " [3]%3d CLL (val: %d)\n", b, int((lenlens >> rshift(b)) & 7));
else
fprintf(stderr, " [_]%3d CLL (val: %d)\n", b, int((lenlens >> rshift(b)) & 7));
}
#endif
auto lltree_fun = [=](unsigned a) -> unsigned char { return (lenlens >> rshift(a)) & 7; };
while(CreateHuffmanTree<bool(Abortable&Flag_InputAbortable)>("Len Lengths", state.lltree, 19, lltree_fun)) { return -2; }}
{auto ltree_fun = state.template DynTreeFunc<bool(Abortable&Flag_InputAbortable)>(std::forward<InputFunctor>(input), nlen, std::forward<BacktrackFunctor>(backtrack), false);
while(CreateHuffmanTree<bool(Abortable&Flag_InputAbortable)>("Dyn Lengths", state.ltree, nlen, ltree_fun)) { return -2; }}
{auto dtree_fun = state.template DynTreeFunc<bool(Abortable&Flag_InputAbortable)>(std::forward<InputFunctor>(input), ndist, std::forward<BacktrackFunctor>(backtrack), true);
while(CreateHuffmanTree<bool(Abortable&Flag_InputAbortable)>("Dyn Dists", state.dtree, ndist, dtree_fun)) { return -2; }}
#undef nlen
#undef ndist
}
else // Fixed block
{
if(header < 2) // Copy stored block data
{
DummyGetBits(state.BitCount%8); // Go to byte boundary (discard a few bits)
GetBits(32, std::uint_least32_t a);
Fail_If(((a ^ (a >> 16)) & 0xFFFF) != 0xFFFF);
#ifdef DO_DEFDB_DUMPING
fprintf(stderr, "raw block of %d bytes (0x%X)\n", (unsigned short)a, a);
#endif
// Note: It is valid for (lower 16 bits of) "a" to be 0 here.
// It is sometimes used for aligning the stream to byte boundary.
while(a-- & 0xFFFF)
{
GetBits(8, unsigned char octet);
while(OutputHelper<bool(Abortable&Flag_OutputAbortable)>::output(output, octet)) { return -3; }
}
goto skipdef;
}
unsigned char (*ltree_fun)(unsigned) = [](unsigned n)->unsigned char{return (n<0x90 || n>=0x118) ? 8u : (n<0x100 ? 9u : 7u); };
unsigned char (*dtree_fun)(unsigned) = [](unsigned )->unsigned char{return 5u;};
while(CreateHuffmanTree<false>("Stat Lengths", state.ltree, 288, ltree_fun)) { return -2; }
while(CreateHuffmanTree<false>("Stat Dists", state.dtree, 32, dtree_fun)) { return -2; }
}
// Do actual deflating.
for(;;)
{
#ifdef DO_DEFDB_DUMPING
unsigned a=bitcounter;
#endif
HuffRead(state.ltree, std::uint_least16_t lencode); // 0..287
if(!(lencode & -256)) // 0..255: literal byte
{
#ifdef DO_DEFDB_DUMPING
{char tmp[64];sprintf(tmp,"[%d]",bitcounter-a); fprintf(stderr, "%4s %02X\n", tmp, lencode);}
#endif
while(OutputHelper<bool(Abortable&Flag_OutputAbortable)>::output(output, lencode)) { return -3; }
}
else if(!(lencode & 0xFF)) break; // 256=end
else // 257..287: length code for backwards reference
{
GetBits(lbits(lencode), std::uint_least16_t length); length += lbase(lencode);
{HuffRead(state.dtree, std::uint_least8_t distcode); // Read distance code (0..31)
{GetBits(dbits(distcode), std::uint_least32_t offset); offset += dbase(distcode);
#ifdef DO_DEFDB_DUMPING
{char tmp[64];sprintf(tmp,"[%d]",bitcounter-a); fprintf(stderr, "%4s (%d,%d)\n", tmp,length,offset);}
#endif
while(OutputHelper<bool(Abortable&Flag_OutputAbortable)>::outputcopy(outputcopy,length,offset)) { return -4; }}}
}
}
skipdef:if(header & 1) break; // last block flag
}
// Note: after this, may come a checksum, and a trailer. Ignoring them.
#undef GetBits
#undef DummyGetBits
#undef Fail_If
#undef HuffRead
return 0;
}
}//ns
/*
`InputParams` may be one of the following sets of parameters:
* InputFunctor input `(5)` `(14)`
* InputIterator begin `(7)` `(14)`
* InputIterator begin, InputIterator end `(6)` `(14)`
* InputIterator begin, SizeType length `(8)` `(14)`
* BidirectionalIterator begin, SizeType length `(8)` `(15)`
* ForwardIterator begin `(7)` `(14)`
* BidirectionalIterator begin `(7)` `(15)`
* RandomAccessIterator begin `(7)` `(15)`
* ForwardIterator begin, ForwardIterator end `(6)` `(15)`
* BidirectionalIterator begin, BidirectionalIterator end `(6)` `(15)`
* RandomAccessIterator begin, RandomAccessIterator end `(6)` `(15)`
`OutputParams` may be one of the following sets of parameters:
* OutputFunctor output `(1)` `(9)`
* OutputFunctor output, WindowFunctor window `(2)`
* OutputIterator target `(9)`
* RandomAccessIterator target `(10)`
* RandomAccessIterator target, SizeType target_limit `(3)` `(10)`
* RandomAccessIterator target, RandomAccessIterator target_end `(4)` `(10)`
*/
namespace gunzip_ns
{
#ifdef DEFLATE_ALLOCATION_AUTOMATIC
#define DeflDeclWindow gunzip_ns::DeflateWindow window;
#elif defined(DEFLATE_ALLOCATION_STATIC)
#define DeflDeclWindow auto& window = gunzip_ns::GetStaticObj<gunzip_ns::DeflateWindow>();
#elif defined(DEFLATE_ALLOCATION_DYNAMIC)
#define DeflDeclWindow std::unique_ptr<gunzip_ns::DeflateWindow> winptr(new gunzip_ns::DeflateWindow); \
auto& window = *winptr;
#endif
template<unsigned char code, typename I,typename O,typename C,typename B>
auto DeflateDispatchFinal(I&& i, O&& o, C&& c, B&& b)
{
if constexpr(code & (Flag_TrackIn | Flag_TrackOut))
{
//fprintf(stderr, "both track flag\n");
SizeTracker<DeflateTrackBothSize> tracker;
return tracker(Gunzip<code & Flag_NoTrackFlagMask>
(tracker.template ForwardInput<I>(i), tracker.template ForwardOutput<O>(o), tracker.template ForwardWindow<C>(c), std::forward<B>(b)));
}
else if constexpr(code & Flag_TrackIn)
{
//fprintf(stderr, "in track flag\n");
SizeTracker<DeflateTrackInSize> tracker;
return tracker(Gunzip<code & Flag_NoTrackFlagMask>
(tracker.template ForwardInput<I>(i),std::forward<O>(o),std::forward<C>(c),std::forward<B>(b)));
}
else if constexpr(code & Flag_TrackOut)
{
//fprintf(stderr, "out track flag\n");
SizeTracker<DeflateTrackOutSize> tracker;
return tracker(Gunzip<code & Flag_NoTrackFlagMask>
(std::forward<I>(i), tracker.template ForwardOutput<O>(o), tracker.template ForwardWindow<C>(c), std::forward<B>(b)));
}
else
{
//fprintf(stderr, "no track flag\n");
return Gunzip<code & Flag_NoTrackFlagMask>(std::forward<I>(i),std::forward<O>(o),std::forward<C>(c),std::forward<B>(b));
}
}
// One-parameter output dispatch:
template<unsigned char code, typename BtFun, typename InFun, typename T1>
auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1)
{
// Is param1 a random access iterator?
if constexpr(is_random_iterator_v<T1>)
{
//fprintf(stderr, "random iterator\n");
auto output = [&](unsigned char l) { *param1 = l; ++param1; };
auto outputcopy = [&](std::uint_least16_t length, std::uint_fast32_t offs)
{
/* length=0 means that offs is the size of the window. */
for(; length--; ++param1) { *param1 = *(param1-offs); }
};
return DeflateDispatchFinal<code>(std::forward<InFun>(infun), output, outputcopy, std::forward<BtFun>(bt));
}
// Is param1 an output iterator?
else if constexpr(is_output_iterator_v<T1>)
{
//fprintf(stderr, "output iterator\n");
DeflDeclWindow
auto output = [&](unsigned char l)
{
window.Data[window.Head++ % MAX_WINDOW_SIZE] = l;
*param1 = l; ++param1;
};
auto outputcopy = [&](std::uint_least16_t length, std::uint_fast32_t offs)
{
/* length=0 means that offs is the size of the window. */
for(; length>0; --length)
{
unsigned char byte = window.Data[(window.Head - offs) % MAX_WINDOW_SIZE];
output(byte);
}
return false;
};
return DeflateDispatchFinal<code>(std::forward<InFun>(infun), output, outputcopy, std::forward<BtFun>(bt));
}
// param1 must be an output functor, then.
else if constexpr(is_output_functor_v<T1>)
{
//fprintf(stderr, "output functor\n");
DeflDeclWindow
auto output = [&](unsigned char l)
{
window.Data[window.Head++ % MAX_WINDOW_SIZE] = l;
return param1(l);
};
auto outputcopy = [&](std::uint_least16_t length, std::uint_fast32_t offs)
{
/* length=0 means that offs is the size of the window. */
for(; length>0; --length)
{
unsigned char byte = window.Data[(window.Head - offs) % MAX_WINDOW_SIZE];
if(OutputHelper<DeflAbortable_OutFun<T1>>::output(output, byte))
break;
}
return length;
};
return DeflateDispatchFinal
<code | (DeflAbortable_OutFun<T1> ? Flag_OutputAbortable : 0)>
(std::forward<InFun>(infun), output, outputcopy, std::forward<BtFun>(bt));
}
else
{
//fprintf(stderr, "unreached code 1\n");
static_assert(code==0xFF, "Deflate: Unknown output parameter type");
}
}
// Two-parameter output dispatch:
template<unsigned char code, typename BtFun, typename InFun, typename T1, typename T2>
auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1, T2&& param2)
{
if constexpr(std::is_same_v<remove_cvref_t<T2>, DeflateTrackNoSize>)
{
//fprintf(stderr, "no track flag...\n");
return DeflateOutputDispatch<code> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(param1));
}
else if constexpr(std::is_same_v<remove_cvref_t<T2>, DeflateTrackInSize>)
{
//fprintf(stderr, "in track flag...\n");
return DeflateOutputDispatch<code | Flag_TrackIn> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(param1));
}
else if constexpr(std::is_same_v<remove_cvref_t<T2>, DeflateTrackOutSize>)
{
//fprintf(stderr, "out track flag...\n");
return DeflateOutputDispatch<code | Flag_TrackOut> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(param1));
}
else if constexpr(std::is_same_v<remove_cvref_t<T2>, DeflateTrackBothSize>)
{
//fprintf(stderr, "both track flag...\n");
return DeflateOutputDispatch<code | Flag_TrackIn | Flag_TrackOut> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(param1));
}
// Are param1 and param2 both random access iterators?
else if constexpr(std::is_same_v<T1,T2> && is_random_iterator_v<T1>)
{
//fprintf(stderr, "random iterator + random iterator\n");
auto output = [&](unsigned char l)
{
if(param1 == param2) return true;
*param1 = l; ++param1;
return false;
};
auto outputcopy = [&](std::uint_least16_t length, std::uint_fast32_t offs)
{
/* length=0 means that offs is the size of the window. */
for(; length > 0 && !(param1 == param2); --length, ++param1)
{
*param1 = *(param1 - offs);
}
return length;
};
return DeflateDispatchFinal<code | Flag_OutputAbortable>(std::forward<InFun>(infun), output, outputcopy, std::forward<BtFun>(bt));
}
// Is param1 a random access iterator and param2 a size?
else if constexpr(is_size_type_v<T2> && is_random_iterator_v<T1>)
{
//fprintf(stderr, "random iterator + size\n");
typename std::iterator_traits<remove_cvref_t<T1>>::difference_type used{}, cap=param2;
auto output = [&](unsigned char l)
{
if(used >= cap) return true;
param1[used] = l; ++used;
return false;
};
auto outputcopy = [&](std::uint_least16_t length, std::uint_fast32_t offs)
{
/* length=0 means that offs is the size of the window. */
for(; length > 0 && used < cap; ++used, --length)
{
param1[used] = param1[used - offs];
}
return length;
};
return DeflateDispatchFinal<code | Flag_OutputAbortable>(std::forward<InFun>(infun), output, outputcopy, std::forward<BtFun>(bt));
}
// Then, param1 must be an output functor and param2 a window functor.
else if constexpr(is_output_functor_v<T1> && is_window_functor_v<T2>)
{
//fprintf(stderr, "output functor + window functor\n");
return DeflateDispatchFinal
<code | ( (DeflAbortable_OutFun<T1> && DeflAbortable_WinFun<T2>) ? Flag_OutputAbortable : 0 ) >
(std::forward<InFun>(infun), std::forward<T1>(param1), std::forward<T2>(param2), std::forward<BtFun>(bt));
}
else
{
//fprintf(stderr, "unreached code 2\n");
static_assert(code==0xFF, "Deflate: Unknown output parameter type");
}
}
// Three-parameter output dispatch:
template<unsigned char code, typename BtFun, typename InFun, typename T1, typename T2, typename T3>
auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& p1, T2&& p2, T3)
{
if constexpr(std::is_same_v<remove_cvref_t<T3>, DeflateTrackNoSize>)
{
//fprintf(stderr, "no track flag...\n");
return DeflateOutputDispatch<code> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(p1), std::forward<T2>(p2));
}
else if constexpr(std::is_same_v<remove_cvref_t<T3>, DeflateTrackInSize>)
{
//fprintf(stderr, "in track flag...\n");
return DeflateOutputDispatch<code | Flag_TrackIn> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(p1), std::forward<T2>(p2));
}
else if constexpr(std::is_same_v<remove_cvref_t<T3>, DeflateTrackOutSize>)
{
//fprintf(stderr, "out track flag...\n");
return DeflateOutputDispatch<code | Flag_TrackOut> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(p1), std::forward<T2>(p2));
}
else if constexpr(std::is_same_v<remove_cvref_t<T3>, DeflateTrackBothSize>)
{
//fprintf(stderr, "both track flag...\n");
return DeflateOutputDispatch<code | Flag_TrackIn | Flag_TrackOut> (std::forward<BtFun>(bt), std::forward<InFun>(infun), std::forward<T1>(p1), std::forward<T2>(p2));
}
else
{
//fprintf(stderr, "unreached code 3\n");
static_assert(code==0xFF, "Deflate: Mismatched parameters. Expected last parameter to be a DeflateTrack option.");
}
}
// One or two parameter input dispatch:
template<unsigned char code, typename BtFun, typename T1, typename T2, typename... T>
auto DeflateInputDispatch(BtFun&& bt, T1&& param1, T2&& param2, T&&... args)
{
using namespace gunzip_ns;
// Are param1 and param2 an input iterator pair?
if constexpr(std::is_same_v<T1, T2> && is_input_iterator_v<T1>)
{
//fprintf(stderr, "input iterator + input iterator\n");
auto inputfun = [&]() -> std::common_type_t<int, decltype(*param1)>
{ if(param1 == param2) { return -1; } int r = *param1; ++param1; return r; };
return DeflateOutputDispatch<code|Flag_InputAbortable>(std::forward<BtFun>(bt), inputfun, std::forward<T>(args)...);
}
// Are param1 and param2 a pair of bidirectional input iterators (forward, bidir, random)?
else if constexpr(std::is_same_v<T1, T2> && is_bidir_input_v<T1>)
{
//fprintf(stderr, "bidir input + bidir input\n");
remove_cvref_t<T1> saved{param1};
auto btfun = [&](bool act) { if(act) param1 = saved; else saved = std::move(param1); };
auto inputfun = [&]() -> std::common_type_t<int, decltype(*param1)>
{ if(param1 == param2) { return -1; } int r = *param1; ++param1; return r; };
return DeflateOutputDispatch<code|Flag_InputAbortable>(btfun, inputfun, std::forward<T>(args)...);
}
// Is param1 an input iterator and param2 a size?
else if constexpr(is_size_type_v<T2> && is_input_iterator_v<T1>)
{
//fprintf(stderr, "input iterator + size\n");
typename std::iterator_traits<remove_cvref_t<T1>>::difference_type remain{param2};
auto inputfun = [&]() -> std::common_type_t<int, decltype(*param1)>
{ if(!remain) return -1; --remain; int r = *param1; ++param1; return r; };
return DeflateOutputDispatch<code|Flag_InputAbortable>(std::forward<BtFun>(bt), inputfun, std::forward<T>(args)...);
}
// Is param1 a bidirectional input iterator (forward, bidir, random) and param2 a size?
else if constexpr(is_size_type_v<T2> && is_bidir_input_v<T1>)
{
//fprintf(stderr, "bidir input + size\n");
typename std::iterator_traits<remove_cvref_t<T1>>::difference_type remain{param2}, savestate{};
auto btfun = [&](bool act) { if(act) { param1 -= (savestate-remain); remain = savestate; } else savestate = remain; };
auto inputfun = [&]() -> std::common_type_t<int, decltype(*param1)>
{ if(!remain) return -1; --remain; int r = *param1; ++param1; return r; };
return DeflateOutputDispatch<code|Flag_InputAbortable>(btfun, inputfun, std::forward<T>(args)...);
}
// Is param1 an input iterator?
else if constexpr(is_input_iterator_v<T1>)
{
//fprintf(stderr, "input iterator\n");
auto inputfun = [&]() -> std::remove_cv_t<decltype(*param1)> { auto r = *param1; ++param1; return r; };
return DeflateOutputDispatch
<code | ( is_abortable_input_type_v<remove_cvref_t<decltype(*param1)>> ? Flag_InputAbortable : 0 ) >
(std::forward<BtFun>(bt), inputfun, std::forward<T2>(param2), std::forward<T>(args)...);
}
// Is param1 a bidirectional input iterator (forward, bidir, random)?
else if constexpr(is_bidir_input_v<T1>)
{
//fprintf(stderr, "bidir input\n");
remove_cvref_t<T1> saved{param1};
auto btfun = [&](bool act) { if(act) param1 = saved; else saved = std::move(param1); };
auto inputfun = [&]() -> std::remove_cv_t<decltype(*param1)> { auto r = *param1; ++param1; return r; };
return DeflateOutputDispatch<code>(btfun, inputfun, std::forward<T2>(param2), std::forward<T>(args)...);
}
// param1 must be an input functor, then. Let's move on to param2 testing!
else if constexpr(is_input_functor_v<T1>)
{
//fprintf(stderr, "input functor\n");
return DeflateOutputDispatch
<code | ( DeflAbortable_InFun<T1> ? Flag_InputAbortable : 0 ) >
(std::forward<BtFun>(bt), std::forward<T1>(param1), std::forward<T2>(param2), std::forward<T>(args)...);
}
else
{
//fprintf(stderr, "unreached code 0\n");
static_assert(code==0xFF, "Deflate: Mismatched parameters. Expected something for an input.");
}
}
#undef DeflDeclWindow
}
template<typename... T>
auto Deflate(T&&... args)
{
return gunzip_ns::DeflateInputDispatch<0>(gunzip_ns::dummy{}, std::forward<T>(args)...);
}
#endif /* #ifndef DOXYGEN_SHOULD_SKIP_THIS */
gitextract_cmy6_mlh/ ├── README.md └── gunzip.hh
SYMBOL INDEX (96 symbols across 1 files)
FILE: gunzip.hh
type DeflateTrackTagBase (line 69) | struct DeflateTrackTagBase{}
type DeflateTrackNoSize (line 70) | struct DeflateTrackNoSize: public DeflateTrackTagBase{}
type DeflateTrackInSize (line 71) | struct DeflateTrackInSize: public DeflateTrackTagBase{}
type DeflateTrackOutSize (line 72) | struct DeflateTrackOutSize: public DeflateTrackTagBase{}
type DeflateTrackBothSize (line 73) | struct DeflateTrackBothSize: public DeflateTrackTagBase{}
type gunzip_ns (line 78) | namespace gunzip_ns
type dummy (line 215) | struct dummy{}
type CeilLog2_s (line 218) | struct CeilLog2_s{ static constexpr unsigned result = 1+CeilLog2_s<(N+...
type CeilLog2_s<0> (line 219) | struct CeilLog2_s<0> { static constexpr unsigned result = 0; }
type CeilLog2_s<1> (line 220) | struct CeilLog2_s<1> { static constexpr unsigned result = 0; }
type FloorLog2_s (line 224) | struct FloorLog2_s{ static constexpr unsigned result = 1+FloorLog2_s<N...
type FloorLog2_s<0> (line 225) | struct FloorLog2_s<0> { static constexpr unsigned result = 0; }
type FloorLog2_s<1> (line 226) | struct FloorLog2_s<1> { static constexpr unsigned result = 0; }
type RandomAccessArray (line 247) | struct RandomAccessArray {}
type RandomAccessArray<true, Dim, Elem> (line 250) | struct RandomAccessArray<true, Dim, Elem>
method Get (line 253) | inline std::uint_fast64_t Get(unsigned index) const { return impl.te...
method Set (line 254) | inline void Set(unsigned index, std::uint_fast32_t value) { impl.te...
method QSet (line 255) | inline void QSet(unsigned index, std::uint_fast32_t value) { impl.te...
method WSet (line 257) | inline void WSet(unsigned index, std::uint_fast64_t value) { impl.te...
type RandomAccessArray<false, Dim, Elem> (line 261) | struct RandomAccessArray<false, Dim, Elem>
method E (line 265) | inline E Get(unsigned index) const { return data[index]; }
method Set (line 266) | inline void Set(unsigned index, E value) { data[index] = value; }
method QSet (line 267) | inline void QSet(unsigned index, E value) { data[index] = value; }
method WSet (line 269) | inline void WSet(unsigned index, std::uint_fast64_t value)
function DeflateDispatchFinal (line 1133) | auto DeflateDispatchFinal(I&& i, O&& o, C&& c, B&& b)
function DeflateOutputDispatch (line 1165) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1)
function DeflateOutputDispatch (line 1235) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1, T2&...
function DeflateOutputDispatch (line 1317) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& p1, T2&& p2...
function DeflateInputDispatch (line 1348) | auto DeflateInputDispatch(BtFun&& bt, T1&& param1, T2&& param2, T&&......
type RandomAccessBitArrayBase (line 116) | struct RandomAccessBitArrayBase
method Get_Unclean (line 121) | static std::uint_fast64_t Get_Unclean(unsigned Size, const U* data, un...
method Get (line 135) | static std::uint_fast64_t Get(const U* data, unsigned index) throw()
method Set (line 142) | static void Set(U* data, unsigned index, std::uint_fast64_t value) thr...
type RandomAccessBitArray (line 195) | struct RandomAccessBitArray
method Get (line 201) | inline std::uint_fast64_t Get(unsigned index) const throw()
method Set (line 207) | inline void Set(unsigned index, std::uint_fast64_t value) throw()
type gunzip_ns (line 213) | namespace gunzip_ns
type dummy (line 215) | struct dummy{}
type CeilLog2_s (line 218) | struct CeilLog2_s{ static constexpr unsigned result = 1+CeilLog2_s<(N+...
type CeilLog2_s<0> (line 219) | struct CeilLog2_s<0> { static constexpr unsigned result = 0; }
type CeilLog2_s<1> (line 220) | struct CeilLog2_s<1> { static constexpr unsigned result = 0; }
type FloorLog2_s (line 224) | struct FloorLog2_s{ static constexpr unsigned result = 1+FloorLog2_s<N...
type FloorLog2_s<0> (line 225) | struct FloorLog2_s<0> { static constexpr unsigned result = 0; }
type FloorLog2_s<1> (line 226) | struct FloorLog2_s<1> { static constexpr unsigned result = 0; }
type RandomAccessArray (line 247) | struct RandomAccessArray {}
type RandomAccessArray<true, Dim, Elem> (line 250) | struct RandomAccessArray<true, Dim, Elem>
method Get (line 253) | inline std::uint_fast64_t Get(unsigned index) const { return impl.te...
method Set (line 254) | inline void Set(unsigned index, std::uint_fast32_t value) { impl.te...
method QSet (line 255) | inline void QSet(unsigned index, std::uint_fast32_t value) { impl.te...
method WSet (line 257) | inline void WSet(unsigned index, std::uint_fast64_t value) { impl.te...
type RandomAccessArray<false, Dim, Elem> (line 261) | struct RandomAccessArray<false, Dim, Elem>
method E (line 265) | inline E Get(unsigned index) const { return data[index]; }
method Set (line 266) | inline void Set(unsigned index, E value) { data[index] = value; }
method QSet (line 267) | inline void QSet(unsigned index, E value) { data[index] = value; }
method WSet (line 269) | inline void WSet(unsigned index, std::uint_fast64_t value)
function DeflateDispatchFinal (line 1133) | auto DeflateDispatchFinal(I&& i, O&& o, C&& c, B&& b)
function DeflateOutputDispatch (line 1165) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1)
function DeflateOutputDispatch (line 1235) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1, T2&...
function DeflateOutputDispatch (line 1317) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& p1, T2&& p2...
function DeflateInputDispatch (line 1348) | auto DeflateInputDispatch(BtFun&& bt, T1&& param1, T2&& param2, T&&......
function lbase (line 403) | inline unsigned lbase(unsigned lencode) { return GetBTable<>()[lencode-...
function rshift (line 406) | inline unsigned rshift(unsigned a) { return GetBTable<>()[a + 32]; }
function lbase (line 408) | inline unsigned lbase(unsigned lencode) { return (lencode > 285 ? 3 : (...
function rshift (line 409) | inline unsigned rshift(unsigned a) { if(!a) return 3*3; else if(a>...
function dbase (line 411) | inline unsigned dbase(unsigned distcode) { return (1 + (distcode>=4 ? ((...
function dbits (line 412) | inline unsigned dbits(unsigned distcode) { return distcode>=4 ? distcode...
function lbits (line 413) | inline unsigned lbits(unsigned lencode) { return ((lencode>=265 && lenc...
type result_of (line 429) | struct result_of { // explain usage
type result_of<F(Args...)> (line 435) | struct result_of<F(Args...)> : std::invoke_result<F, Args...> {}
type result_of<F(Args...)> (line 438) | struct result_of<F(Args...)> : std::result_of<F(Args...)> {}
type OutputHelper (line 505) | struct OutputHelper
method output (line 508) | static inline bool output(OutputFunctor&& output, unsigned char byte)
method outputcopy (line 514) | static inline bool outputcopy(WindowFunctor&& outputcopy, std::uint_le...
type OutputHelper<true> (line 522) | struct OutputHelper<true>
method output (line 525) | static inline bool output(OutputFunctor&& output, unsigned char byte)
method outputcopy (line 530) | static inline bool outputcopy(WindowFunctor&& outputcopy, std::uint_le...
type SizeTracker_NoOutput (line 537) | struct SizeTracker_NoOutput
method OutByte (line 539) | inline void OutByte() { }
method OutBytes (line 540) | inline void OutBytes(std::uint_fast64_t) { }
method T (line 544) | static inline constexpr T&& ForwardOutput(std::remove_reference_t<T>& ...
method T (line 546) | static inline constexpr T&& ForwardOutput(std::remove_reference_t<T>&&...
method T (line 549) | static inline constexpr T&& ForwardWindow(std::remove_reference_t<T>& ...
method T (line 551) | static inline constexpr T&& ForwardWindow(std::remove_reference_t<T>&&...
type SizeTracker_NoInput (line 553) | struct SizeTracker_NoInput
method InByte (line 555) | inline void InByte() { }
method InBytes (line 556) | inline void InBytes(std::uint_fast64_t) { }
method T (line 559) | static inline constexpr T&& ForwardInput(std::remove_reference_t<T>& f...
method T (line 561) | static inline constexpr T&& ForwardInput(std::remove_reference_t<T>&& ...
type SizeTracker_DoInput (line 563) | struct SizeTracker_DoInput
method InByte (line 567) | inline void InByte() { ++insize; }
method InBytes (line 568) | inline void InBytes(std::uint_fast64_t n) { insize += n; }
method ForwardInput (line 571) | auto ForwardInput(const InputFunctor& input)
method ForwardInput (line 577) | auto ForwardInput(const InputFunctor& input)
type SizeTracker_DoOutput (line 582) | struct SizeTracker_DoOutput
method OutByte (line 586) | inline void OutByte() { ++outsize; }
method OutBytes (line 587) | inline void OutBytes(std::uint_fast64_t n) { outsize += n; }
method ForwardOutput (line 590) | auto ForwardOutput(const OutputFunctor& output)
method ForwardOutput (line 596) | auto ForwardOutput(const OutputFunctor& output)
method ForwardWindow (line 602) | auto ForwardWindow(const WindowFunctor& outputcopy)
method ForwardWindow (line 612) | auto ForwardWindow(const WindowFunctor& outputcopy)
type SizeTracker (line 624) | struct SizeTracker: public SizeTracker_NoOutput, public SizeTracker_NoInput
type SizeTracker<DeflateTrackOutSize> (line 629) | struct SizeTracker<DeflateTrackOutSize>: public SizeTracker_NoInput, pub...
method result (line 632) | inline result operator() (int returncode) const { return result{return...
type SizeTracker<DeflateTrackInSize> (line 635) | struct SizeTracker<DeflateTrackInSize>: public SizeTracker_NoOutput, pub...
method result (line 638) | inline result operator() (int returncode) const { return result{return...
type SizeTracker<DeflateTrackBothSize> (line 641) | struct SizeTracker<DeflateTrackBothSize>: public SizeTracker_DoInput, pu...
method result (line 644) | inline result operator() (int returncode) const { return result{return...
type DeflateBitCache (line 650) | struct DeflateBitCache
method GetBits (line 655) | std::uint_least64_t GetBits(InputFunctor&& input, unsigned numbits)
method HuffRead (line 690) | std::uint_least32_t HuffRead(InputFunctor&& input,
type DeflateState (line 722) | struct DeflateState: public DeflateBitCache
method DynTreeFunc (line 746) | auto DynTreeFunc(InputFunctor&& input, std::uint_fast16_t length, Back...
type DeflateState<true> (line 815) | struct DeflateState<true>: public DeflateBitCache
method DynTreeFunc (line 843) | auto DynTreeFunc(InputFunctor&& input, std::uint_fast16_t /*length*/, ...
type DeflateWindow (line 893) | struct DeflateWindow
function ObjectType (line 901) | ObjectType& GetStaticObj()
function Gunzip (line 928) | int Gunzip(InputFunctor&& input,
type gunzip_ns (line 1121) | namespace gunzip_ns
type dummy (line 215) | struct dummy{}
type CeilLog2_s (line 218) | struct CeilLog2_s{ static constexpr unsigned result = 1+CeilLog2_s<(N+...
type CeilLog2_s<0> (line 219) | struct CeilLog2_s<0> { static constexpr unsigned result = 0; }
type CeilLog2_s<1> (line 220) | struct CeilLog2_s<1> { static constexpr unsigned result = 0; }
type FloorLog2_s (line 224) | struct FloorLog2_s{ static constexpr unsigned result = 1+FloorLog2_s<N...
type FloorLog2_s<0> (line 225) | struct FloorLog2_s<0> { static constexpr unsigned result = 0; }
type FloorLog2_s<1> (line 226) | struct FloorLog2_s<1> { static constexpr unsigned result = 0; }
type RandomAccessArray (line 247) | struct RandomAccessArray {}
type RandomAccessArray<true, Dim, Elem> (line 250) | struct RandomAccessArray<true, Dim, Elem>
method Get (line 253) | inline std::uint_fast64_t Get(unsigned index) const { return impl.te...
method Set (line 254) | inline void Set(unsigned index, std::uint_fast32_t value) { impl.te...
method QSet (line 255) | inline void QSet(unsigned index, std::uint_fast32_t value) { impl.te...
method WSet (line 257) | inline void WSet(unsigned index, std::uint_fast64_t value) { impl.te...
type RandomAccessArray<false, Dim, Elem> (line 261) | struct RandomAccessArray<false, Dim, Elem>
method E (line 265) | inline E Get(unsigned index) const { return data[index]; }
method Set (line 266) | inline void Set(unsigned index, E value) { data[index] = value; }
method QSet (line 267) | inline void QSet(unsigned index, E value) { data[index] = value; }
method WSet (line 269) | inline void WSet(unsigned index, std::uint_fast64_t value)
function DeflateDispatchFinal (line 1133) | auto DeflateDispatchFinal(I&& i, O&& o, C&& c, B&& b)
function DeflateOutputDispatch (line 1165) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1)
function DeflateOutputDispatch (line 1235) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& param1, T2&...
function DeflateOutputDispatch (line 1317) | auto DeflateOutputDispatch(BtFun&& bt, InFun&& infun, T1&& p1, T2&& p2...
function DeflateInputDispatch (line 1348) | auto DeflateInputDispatch(BtFun&& bt, T1&& param1, T2&& param2, T&&......
function Deflate (line 1425) | auto Deflate(T&&... args)
Condensed preview — 2 files, each showing path, character count, and a content snippet. Download the .json file or copy for the full structured content (87K chars).
[
{
"path": "README.md",
"chars": 16553,
"preview": "# TinyDeflate\n\nA deflate/gzip decompressor, as a C++17 template function,\nthat requires minimal amount of memory to work"
},
{
"path": "gunzip.hh",
"chars": 67848,
"preview": "/* My tiny gzip decompressor without using zlib. - Joel Yliluoma\n * http://iki.fi/bisqwit/ , http://youtube.com/user/Bis"
}
]
About this extraction
This page contains the full source code of the bisqwit/TinyDeflate GitHub repository, extracted and formatted as plain text for AI agents and large language models (LLMs). The extraction includes 2 files (82.4 KB), approximately 21.2k tokens, and a symbol index with 96 extracted functions, classes, methods, constants, and types. Use this with OpenClaw, Claude, ChatGPT, Cursor, Windsurf, or any other AI tool that accepts text input. You can copy the full output to your clipboard or download it as a .txt file.
Extracted by GitExtract — free GitHub repo to text converter for AI. Built by Nikandr Surkov.