Justine Tunney
2 years ago
parent
113eeabd85
commit
04caf6f9ad
@ -1,125 +0,0 @@ |
||||
--- |
||||
Language: Cpp |
||||
# BasedOnStyle: WebKit |
||||
AccessModifierOffset: -4 |
||||
AlignAfterOpenBracket: DontAlign |
||||
AlignConsecutiveAssignments: false |
||||
AlignConsecutiveDeclarations: false |
||||
AlignEscapedNewlines: Right |
||||
AlignOperands: false |
||||
AlignTrailingComments: false |
||||
AllowAllArgumentsOnNextLine: true |
||||
AllowAllConstructorInitializersOnNextLine: true |
||||
AllowAllParametersOfDeclarationOnNextLine: true |
||||
AllowShortBlocksOnASingleLine: false |
||||
AllowShortCaseLabelsOnASingleLine: false |
||||
AllowShortFunctionsOnASingleLine: All |
||||
AllowShortLambdasOnASingleLine: All |
||||
AllowShortIfStatementsOnASingleLine: Never |
||||
AllowShortLoopsOnASingleLine: false |
||||
AlwaysBreakAfterDefinitionReturnType: None |
||||
AlwaysBreakAfterReturnType: None |
||||
AlwaysBreakBeforeMultilineStrings: false |
||||
AlwaysBreakTemplateDeclarations: MultiLine |
||||
BinPackArguments: true |
||||
BinPackParameters: true |
||||
BraceWrapping: |
||||
AfterCaseLabel: false |
||||
AfterClass: false |
||||
AfterControlStatement: false |
||||
AfterEnum: false |
||||
AfterFunction: true |
||||
AfterNamespace: false |
||||
AfterObjCDeclaration: false |
||||
AfterStruct: false |
||||
AfterUnion: false |
||||
AfterExternBlock: false |
||||
BeforeCatch: false |
||||
BeforeElse: false |
||||
IndentBraces: false |
||||
SplitEmptyFunction: true |
||||
SplitEmptyRecord: true |
||||
SplitEmptyNamespace: true |
||||
BreakBeforeBinaryOperators: All |
||||
BreakBeforeBraces: WebKit |
||||
BreakBeforeInheritanceComma: false |
||||
BreakInheritanceList: BeforeColon |
||||
BreakBeforeTernaryOperators: true |
||||
BreakConstructorInitializersBeforeComma: false |
||||
BreakConstructorInitializers: BeforeComma |
||||
BreakAfterJavaFieldAnnotations: false |
||||
BreakStringLiterals: true |
||||
ColumnLimit: 80 |
||||
CommentPragmas: '^ IWYU pragma:' |
||||
CompactNamespaces: false |
||||
ConstructorInitializerAllOnOneLineOrOnePerLine: false |
||||
ConstructorInitializerIndentWidth: 4 |
||||
ContinuationIndentWidth: 4 |
||||
Cpp11BracedListStyle: false |
||||
DerivePointerAlignment: false |
||||
DisableFormat: false |
||||
ExperimentalAutoDetectBinPacking: false |
||||
FixNamespaceComments: false |
||||
ForEachMacros: |
||||
- foreach |
||||
- Q_FOREACH |
||||
- BOOST_FOREACH |
||||
IncludeBlocks: Preserve |
||||
IncludeCategories: |
||||
- Regex: '^"(llvm|llvm-c|clang|clang-c)/' |
||||
Priority: 2 |
||||
- Regex: '^(<|"(gtest|gmock|isl|json)/)' |
||||
Priority: 3 |
||||
- Regex: '.*' |
||||
Priority: 1 |
||||
IncludeIsMainRegex: '(Test)?$' |
||||
IndentCaseLabels: false |
||||
IndentPPDirectives: None |
||||
IndentWidth: 4 |
||||
IndentWrappedFunctionNames: false |
||||
JavaScriptQuotes: Leave |
||||
JavaScriptWrapImports: true |
||||
KeepEmptyLinesAtTheStartOfBlocks: true |
||||
MacroBlockBegin: '' |
||||
MacroBlockEnd: '' |
||||
MaxEmptyLinesToKeep: 1 |
||||
NamespaceIndentation: Inner |
||||
ObjCBinPackProtocolList: Auto |
||||
ObjCBlockIndentWidth: 4 |
||||
ObjCSpaceAfterProperty: true |
||||
ObjCSpaceBeforeProtocolList: true |
||||
PenaltyBreakAssignment: 2 |
||||
PenaltyBreakBeforeFirstCallParameter: 19 |
||||
PenaltyBreakComment: 300 |
||||
PenaltyBreakFirstLessLess: 120 |
||||
PenaltyBreakString: 1000 |
||||
PenaltyBreakTemplateDeclaration: 10 |
||||
PenaltyExcessCharacter: 1000000 |
||||
PenaltyReturnTypeOnItsOwnLine: 60 |
||||
PointerAlignment: Right |
||||
ReflowComments: true |
||||
SortIncludes: true |
||||
SortUsingDeclarations: true |
||||
SpaceAfterCStyleCast: false |
||||
SpaceAfterLogicalNot: false |
||||
SpaceAfterTemplateKeyword: true |
||||
SpaceBeforeAssignmentOperators: true |
||||
SpaceBeforeCpp11BracedList: true |
||||
SpaceBeforeCtorInitializerColon: true |
||||
SpaceBeforeInheritanceColon: true |
||||
SpaceBeforeParens: ControlStatements |
||||
SpaceBeforeRangeBasedForLoopColon: true |
||||
SpaceInEmptyParentheses: false |
||||
SpacesBeforeTrailingComments: 1 |
||||
SpacesInAngles: false |
||||
SpacesInContainerLiterals: true |
||||
SpacesInCStyleCastParentheses: false |
||||
SpacesInParentheses: false |
||||
SpacesInSquareBrackets: false |
||||
Standard: Cpp11 |
||||
StatementMacros: |
||||
- Q_UNUSED |
||||
- QT_REQUIRE_VERSION |
||||
TabWidth: 4 |
||||
UseTab: Always |
||||
... |
||||
@ -1,33 +0,0 @@ |
||||
#-*-mode:makefile-gmake;indent-tabs-mode:t;tab-width:8;coding:utf-8-*-┐
|
||||
#───vi: set et ft=make ts=8 tw=8 fenc=utf-8 :vi───────────────────────┘
|
||||
# Description:
|
||||
# bzip2 is a compression format.
|
||||
|
||||
PKGS += THIRD_PARTY_BZIP2
|
||||
|
||||
THIRD_PARTY_BZIP2_BINS = \
|
||||
o/$(MODE)/third_party/bzip2/µbunzip2.com \
|
||||
o/$(MODE)/third_party/bzip2/µbunzip2.com.dbg
|
||||
|
||||
THIRD_PARTY_BZIP2_OBJS = \
|
||||
o/$(MODE)/third_party/bzip2/µbunzip2.o
|
||||
|
||||
THIRD_PARTY_BZIP2_DEPS := $(call uniq, \
|
||||
$(LIBC_STR) \
|
||||
$(LIBC_STDIO))
|
||||
|
||||
$(THIRD_PARTY_BZIP2_OBJS): \
|
||||
DEFAULT_CPPFLAGS += \
|
||||
-DHAVE_CONFIG_H
|
||||
|
||||
o/$(MODE)/third_party/bzip2/µbunzip2.com.dbg: \
|
||||
$(THIRD_PARTY_BZIP2_DEPS) \
|
||||
$(THIRD_PARTY_BZIP2_OBJS) \
|
||||
$(CRT) \
|
||||
$(APE)
|
||||
@$(APELINK)
|
||||
|
||||
$(THIRD_PARTY_BZIP2_OBJS): third_party/bzip2/bzip2.mk |
||||
|
||||
.PHONY: o/$(MODE)/third_party/bzip2 |
||||
o/$(MODE)/third_party/bzip2: $(THIRD_PARTY_BZIP2_BINS) |
||||
@ -1,569 +0,0 @@ |
||||
/* micro-bunzip, a small, simple bzip2 decompression implementation.
|
||||
Copyright 2003 by Rob Landley (rob@landley.net). |
||||
|
||||
Based on bzip2 decompression code by Julian R Seward (jseward@acm.org), |
||||
which also acknowledges contributions by Mike Burrows, David Wheeler, |
||||
Peter Fenwick, Alistair Moffat, Radford Neal, Ian H. Witten, |
||||
Robert Sedgewick, and Jon L. Bentley. |
||||
|
||||
I hereby release this code under the GNU Library General Public License |
||||
(LGPL) version 2, available at http://www.gnu.org/copyleft/lgpl.html
|
||||
*/ |
||||
|
||||
#include "libc/calls/calls.h" |
||||
#include "libc/mem/mem.h" |
||||
#include "libc/runtime/runtime.h" |
||||
#include "libc/stdio/stdio.h" |
||||
#include "libc/str/str.h" |
||||
#include "libc/sysv/consts/fileno.h" |
||||
|
||||
/* Constants for huffman coding */ |
||||
#define MAX_GROUPS 6 |
||||
#define GROUP_SIZE 50 /* 64 would have been more efficient */ |
||||
#define MAX_HUFCODE_BITS 20 /* Longest huffman code allowed */ |
||||
#define MAX_SYMBOLS 258 /* 256 literals + RUNA + RUNB */ |
||||
#define SYMBOL_RUNA 0 |
||||
#define SYMBOL_RUNB 1 |
||||
|
||||
/* Status return values */ |
||||
#define RETVAL_OK 0 |
||||
#define RETVAL_LAST_BLOCK (-1) |
||||
#define RETVAL_NOT_BZIP_DATA (-2) |
||||
#define RETVAL_UNEXPECTED_INPUT_EOF (-3) |
||||
#define RETVAL_UNEXPECTED_OUTPUT_EOF (-4) |
||||
#define RETVAL_DATA_ERROR (-5) |
||||
#define RETVAL_OUT_OF_MEMORY (-6) |
||||
#define RETVAL_OBSOLETE_INPUT (-7) |
||||
|
||||
/* Other housekeeping constants */ |
||||
#define IOBUF_SIZE 4096 |
||||
|
||||
char *bunzip_errors[] = { NULL, "Bad file checksum", "Not bzip data", |
||||
"Unexpected input EOF", "Unexpected output EOF", "Data error", |
||||
"Out of memory", "Obsolete (pre 0.9.5) bzip format not supported." }; |
||||
|
||||
/* This is what we know about each huffman coding group */ |
||||
struct group_data { |
||||
int limit[MAX_HUFCODE_BITS], base[MAX_HUFCODE_BITS], permute[MAX_SYMBOLS]; |
||||
char minLen, maxLen; |
||||
}; |
||||
|
||||
/* Structure holding all the housekeeping data, including IO buffers and
|
||||
memory that persists between calls to bunzip */ |
||||
typedef struct { |
||||
/* For I/O error handling */ |
||||
jmp_buf jmpbuf; |
||||
/* Input stream, input buffer, input bit buffer */ |
||||
int64_t in_fd, inbufCount, inbufPos; |
||||
unsigned char *inbuf; |
||||
unsigned int inbufBitCount, inbufBits; |
||||
/* Output buffer */ |
||||
char outbuf[IOBUF_SIZE]; |
||||
int outbufPos; |
||||
/* The CRC values stored in the block header and calculated from the data */ |
||||
unsigned int crc32Table[256], headerCRC, dataCRC, totalCRC; |
||||
/* Intermediate buffer and its size (in bytes) */ |
||||
unsigned int *dbuf, dbufSize; |
||||
/* State for interrupting output loop */ |
||||
int writePos, writeRun, writeCount, writeCurrent; |
||||
|
||||
/* These things are a bit too big to go on the stack */ |
||||
unsigned char selectors[32768]; /* nSelectors=15 bits */ |
||||
struct group_data groups[MAX_GROUPS]; /* huffman coding tables */ |
||||
} bunzip_data; |
||||
|
||||
/* Return the next nnn bits of input. All reads from the compressed
|
||||
input are done through this function. All reads are big endian */ |
||||
static unsigned int get_bits(bunzip_data *bd, char bits_wanted) |
||||
{ |
||||
unsigned int bits = 0; |
||||
|
||||
/* If we need to get more data from the byte buffer, do so. (Loop
|
||||
getting one byte at a time to enforce endianness and avoid |
||||
unaligned access.) */ |
||||
while (bd->inbufBitCount < bits_wanted) { |
||||
/* If we need to read more data from file into byte buffer, do so */ |
||||
if (bd->inbufPos == bd->inbufCount) { |
||||
if (!(bd->inbufCount = read(bd->in_fd, bd->inbuf, IOBUF_SIZE))) |
||||
longjmp(bd->jmpbuf, RETVAL_UNEXPECTED_INPUT_EOF); |
||||
bd->inbufPos = 0; |
||||
} |
||||
/* Avoid 32-bit overflow (dump bit buffer to top of output) */ |
||||
if (bd->inbufBitCount >= 24) { |
||||
bits = bd->inbufBits & ((1u << bd->inbufBitCount) - 1); |
||||
bits_wanted -= bd->inbufBitCount; |
||||
bits <<= bits_wanted; |
||||
bd->inbufBitCount = 0; |
||||
} |
||||
/* Grab next 8 bits of input from buffer. */ |
||||
bd->inbufBits = (bd->inbufBits << 8) | bd->inbuf[bd->inbufPos++]; |
||||
bd->inbufBitCount += 8; |
||||
} |
||||
/* Calculate result */ |
||||
bd->inbufBitCount -= bits_wanted; |
||||
bits |= (bd->inbufBits >> bd->inbufBitCount) & ((1u << bits_wanted) - 1); |
||||
|
||||
return bits; |
||||
} |
||||
|
||||
/* Decompress a block of text to into intermediate buffer */ |
||||
|
||||
static int read_bunzip_data(bunzip_data *bd) |
||||
{ |
||||
struct group_data *hufGroup; |
||||
int dbufCount, nextSym, dbufSize, origPtr, groupCount, *base, *limit, |
||||
selector, i, j, k, t, runPos, symCount, symTotal, nSelectors, |
||||
byteCount[256]; |
||||
unsigned char uc, symToByte[256], mtfSymbol[256], *selectors; |
||||
unsigned int *dbuf; |
||||
|
||||
/* Read in header signature (borrowing mtfSymbol for temp space). */ |
||||
for (i = 0; i < 6; i++) |
||||
mtfSymbol[i] = get_bits(bd, 8); |
||||
mtfSymbol[6] = 0; |
||||
/* Read CRC (which is stored big endian). */ |
||||
bd->headerCRC = get_bits(bd, 32); |
||||
/* Is this the last block (with CRC for file)? */ |
||||
if (!strcmp((char *)mtfSymbol, "\x17\x72\x45\x38\x50\x90")) |
||||
return RETVAL_LAST_BLOCK; |
||||
/* If it's not a valid data block, barf. */ |
||||
if (strcmp((char *)mtfSymbol, "\x31\x41\x59\x26\x53\x59")) |
||||
return RETVAL_NOT_BZIP_DATA; |
||||
|
||||
dbuf = bd->dbuf; |
||||
dbufSize = bd->dbufSize; |
||||
selectors = bd->selectors; |
||||
/* We can add support for blockRandomised if anybody complains.
|
||||
There was some code for this in busybox 1.0.0-pre3, but nobody |
||||
ever noticed that it didn't actually work. */ |
||||
if (get_bits(bd, 1)) |
||||
return RETVAL_OBSOLETE_INPUT; |
||||
if ((origPtr = get_bits(bd, 24)) > dbufSize) |
||||
return RETVAL_DATA_ERROR; |
||||
/* mapping table: if some byte values are never used (encoding
|
||||
things like ascii text), the compression code removes the gaps to |
||||
have fewer symbols to deal with, and writes a sparse bitfield |
||||
indicating which values were present. We make a translation table |
||||
to convert the symbols back to the corresponding bytes. */ |
||||
t = get_bits(bd, 16); |
||||
memset(symToByte, 0, 256); |
||||
symTotal = 0; |
||||
for (i = 0; i < 16; i++) { |
||||
if (t & (1u << (15 - i))) { |
||||
k = get_bits(bd, 16); |
||||
for (j = 0; j < 16; j++) |
||||
if (k & (1u << (15 - j))) |
||||
symToByte[symTotal++] = (16 * i) + j; |
||||
} |
||||
} |
||||
/* How many different huffman coding groups does this block use? */ |
||||
groupCount = get_bits(bd, 3); |
||||
if (groupCount < 2 || groupCount > MAX_GROUPS) |
||||
return RETVAL_DATA_ERROR; |
||||
/* nSelectors: Every GROUP_SIZE many symbols we select a new huffman
|
||||
coding group. Read in the group selector list, which is stored as |
||||
MTF encoded bit runs. */ |
||||
if (!(nSelectors = get_bits(bd, 15))) |
||||
return RETVAL_DATA_ERROR; |
||||
for (i = 0; i < groupCount; i++) |
||||
mtfSymbol[i] = i; |
||||
for (i = 0; i < nSelectors; i++) { |
||||
/* Get next value */ |
||||
for (j = 0; get_bits(bd, 1); j++) |
||||
if (j >= groupCount) |
||||
return RETVAL_DATA_ERROR; |
||||
/* Decode MTF to get the next selector */ |
||||
uc = mtfSymbol[j]; |
||||
memmove(mtfSymbol + 1, mtfSymbol, j); |
||||
mtfSymbol[0] = selectors[i] = uc; |
||||
} |
||||
/* Read the huffman coding tables for each group, which code for symTotal
|
||||
literal symbols, plus two run symbols (RUNA, RUNB) */ |
||||
symCount = symTotal + 2; |
||||
for (j = 0; j < groupCount; j++) { |
||||
unsigned char length[MAX_SYMBOLS], temp[MAX_HUFCODE_BITS + 1]; |
||||
int minLen, maxLen, pp; |
||||
/* Read lengths */ |
||||
t = get_bits(bd, 5); |
||||
for (i = 0; i < symCount; i++) { |
||||
for (;;) { |
||||
if (t < 1 || t > MAX_HUFCODE_BITS) |
||||
return RETVAL_DATA_ERROR; |
||||
if (!get_bits(bd, 1)) |
||||
break; |
||||
if (!get_bits(bd, 1)) |
||||
t++; |
||||
else |
||||
t--; |
||||
} |
||||
length[i] = t; |
||||
} |
||||
/* Find largest and smallest lengths in this group */ |
||||
minLen = maxLen = length[0]; |
||||
for (i = 1; i < symCount; i++) { |
||||
if (length[i] > maxLen) |
||||
maxLen = length[i]; |
||||
else if (length[i] < minLen) |
||||
minLen = length[i]; |
||||
} |
||||
/* Calculate permute[], base[], and limit[] tables from length[].
|
||||
* |
||||
* permute[] is the lookup table for converting huffman coded symbols |
||||
* into decoded symbols. base[] is the amount to subtract from the |
||||
* value of a huffman symbol of a given length when using permute[]. |
||||
* |
||||
* limit[] indicates the largest numerical value a symbol with a given |
||||
* number of bits can have. It lets us know when to stop reading. |
||||
* |
||||
* To use these, keep reading bits until value<=limit[bitcount] or |
||||
* you've read over 20 bits (error). Then the decoded symbol |
||||
* equals permute[hufcode_value-base[hufcode_bitcount]]. |
||||
*/ |
||||
hufGroup = bd->groups + j; |
||||
hufGroup->minLen = minLen; |
||||
hufGroup->maxLen = maxLen; |
||||
/* Note that minLen can't be smaller than 1, so we adjust the
|
||||
base and limit array pointers so we're not always wasting the |
||||
first entry. We do this again when using them (during symbol |
||||
decoding).*/ |
||||
base = hufGroup->base - 1; |
||||
limit = hufGroup->limit - 1; |
||||
/* Calculate permute[] */ |
||||
pp = 0; |
||||
for (i = minLen; i <= maxLen; i++) |
||||
for (t = 0; t < symCount; t++) |
||||
if (length[t] == i) |
||||
hufGroup->permute[pp++] = t; |
||||
/* Count cumulative symbols coded for at each bit length */ |
||||
for (i = minLen; i <= maxLen; i++) |
||||
temp[i] = limit[i] = 0; |
||||
for (i = 0; i < symCount; i++) |
||||
temp[length[i]]++; |
||||
/* Calculate limit[] (the largest symbol-coding value at each
|
||||
bit length, which is (previous limit<<1)+symbols at this |
||||
level), and base[] (number of symbols to ignore at each bit |
||||
length, which is limit-cumulative count of symbols coded for |
||||
already). */ |
||||
pp = t = 0; |
||||
for (i = minLen; i < maxLen; i++) { |
||||
pp += temp[i]; |
||||
limit[i] = pp - 1; |
||||
pp <<= 1; |
||||
base[i + 1] = pp - (t += temp[i]); |
||||
} |
||||
limit[maxLen] = pp + temp[maxLen] - 1; |
||||
base[minLen] = 0; |
||||
} |
||||
/* We've finished reading and digesting the block header. Now read this
|
||||
block's huffman coded symbols from the file and undo the huffman |
||||
coding and run length encoding, saving the result into |
||||
dbuf[dbufCount++]=uc */ |
||||
|
||||
/* Initialize symbol occurrence counters and symbol mtf table */ |
||||
memset(byteCount, 0, 256 * sizeof(int)); |
||||
for (i = 0; i < 256; i++) |
||||
mtfSymbol[i] = (unsigned char)i; |
||||
/* Loop through compressed symbols */ |
||||
runPos = dbufCount = symCount = selector = 0; |
||||
for (;;) { |
||||
/* Determine which huffman coding group to use. */ |
||||
if (!(symCount--)) { |
||||
symCount = GROUP_SIZE - 1; |
||||
if (selector >= nSelectors) |
||||
return RETVAL_DATA_ERROR; |
||||
hufGroup = bd->groups + selectors[selector++]; |
||||
base = hufGroup->base - 1; |
||||
limit = hufGroup->limit - 1; |
||||
} |
||||
/* Read next huffman-coded symbol */ |
||||
i = hufGroup->minLen; |
||||
j = get_bits(bd, i); |
||||
for (;;) { |
||||
if (i > hufGroup->maxLen) |
||||
return RETVAL_DATA_ERROR; |
||||
if (j <= limit[i]) |
||||
break; |
||||
i++; |
||||
|
||||
j = (j << 1) | get_bits(bd, 1); |
||||
} |
||||
/* Huffman decode nextSym (with bounds checking) */ |
||||
j -= base[i]; |
||||
if (j < 0 || j >= MAX_SYMBOLS) |
||||
return RETVAL_DATA_ERROR; |
||||
nextSym = hufGroup->permute[j]; |
||||
/* If this is a repeated run, loop collecting data */ |
||||
if (nextSym == SYMBOL_RUNA || nextSym == SYMBOL_RUNB) { |
||||
/* If this is the start of a new run, zero out counter */ |
||||
if (!runPos) { |
||||
runPos = 1; |
||||
t = 0; |
||||
} |
||||
/* Neat trick that saves 1 symbol: instead of or-ing 0 or 1
|
||||
at each bit position, add 1 or 2 instead. For example, |
||||
1011 is 1<<0 + 1<<1 + 2<<2. 1010 is 2<<0 + 2<<1 + 1<<2. |
||||
You can make any bit pattern that way using 1 less symbol |
||||
than the basic or 0/1 method (except all bits 0, which |
||||
would use no symbols, but a run of length 0 doesn't mean |
||||
anything in this context). Thus space is saved. */ |
||||
if (nextSym == SYMBOL_RUNA) |
||||
t += runPos; |
||||
else |
||||
t += 2 * runPos; |
||||
runPos <<= 1; |
||||
continue; |
||||
} |
||||
/* When we hit the first non-run symbol after a run, we now know
|
||||
how many times to repeat the last literal, so append that |
||||
many copies to our buffer of decoded symbols (dbuf) now. (The |
||||
last literal used is the one at the head of the mtfSymbol |
||||
array.) */ |
||||
if (runPos) { |
||||
runPos = 0; |
||||
if (dbufCount + t >= dbufSize) |
||||
return RETVAL_DATA_ERROR; |
||||
|
||||
uc = symToByte[mtfSymbol[0]]; |
||||
byteCount[uc] += t; |
||||
while (t--) |
||||
dbuf[dbufCount++] = uc; |
||||
} |
||||
/* Is this the terminating symbol? */ |
||||
if (nextSym > symTotal) |
||||
break; |
||||
/* At this point, the symbol we just decoded indicates a new
|
||||
literal character. Subtract one to get the position in the |
||||
MTF array at which this literal is currently to be found. |
||||
(Note that the result can't be -1 or 0, because 0 and 1 are |
||||
RUNA and RUNB. Another instance of the first symbol in the |
||||
mtf array, position 0, would have been handled as part of a |
||||
run.) */ |
||||
if (dbufCount >= dbufSize) |
||||
return RETVAL_DATA_ERROR; |
||||
i = nextSym - 1; |
||||
uc = mtfSymbol[i]; |
||||
memmove(mtfSymbol + 1, mtfSymbol, i); |
||||
mtfSymbol[0] = uc; |
||||
uc = symToByte[uc]; |
||||
/* We have our literal byte. Save it into dbuf. */ |
||||
byteCount[uc]++; |
||||
dbuf[dbufCount++] = (unsigned int)uc; |
||||
} |
||||
/* At this point, we've finished reading huffman-coded symbols and
|
||||
compressed runs from the input stream. There are dbufCount many |
||||
of them in dbuf[]. Now undo the Burrows-Wheeler transform on |
||||
dbuf. See http://dogma.net/markn/articles/bwt/bwt.htm */
|
||||
|
||||
/* Now we know what dbufCount is, do a better sanity check on origPtr. */ |
||||
if (origPtr < 0 || origPtr >= dbufCount) |
||||
return RETVAL_DATA_ERROR; |
||||
/* Turn byteCount into cumulative occurrence counts of 0 to n-1. */ |
||||
j = 0; |
||||
for (i = 0; i < 256; i++) { |
||||
k = j + byteCount[i]; |
||||
byteCount[i] = j; |
||||
j = k; |
||||
} |
||||
/* Figure out what order dbuf would be in if we sorted it. */ |
||||
for (i = 0; i < dbufCount; i++) { |
||||
uc = (unsigned char)(dbuf[i] & 0xff); |
||||
dbuf[byteCount[uc]] |= (i << 8); |
||||
byteCount[uc]++; |
||||
} |
||||
/* blockRandomised support would go here. */ |
||||
|
||||
/* Using i as position, j as previous character, t as current character,
|
||||
and uc as run count */ |
||||
bd->dataCRC = 0xffffffffL; |
||||
/* Decode first byte by hand to initialize "previous" byte. Note
|
||||
that it doesn't get output, and if the first three characters are |
||||
identical it doesn't qualify as a run (hence uc=255, which will |
||||
either wrap to 1 or get reset). */ |
||||
if (dbufCount) { |
||||
bd->writePos = dbuf[origPtr]; |
||||
bd->writeCurrent = (unsigned char)(bd->writePos & 0xff); |
||||
bd->writePos >>= 8; |
||||
bd->writeRun = -1; |
||||
} |
||||
bd->writeCount = dbufCount; |
||||
|
||||
return RETVAL_OK; |
||||
} |
||||
|
||||
/* Flush output buffer to disk */ |
||||
static void flush_bunzip_outbuf(bunzip_data *bd, int64_t out_fd) |
||||
{ |
||||
if (bd->outbufPos) { |
||||
if (write(out_fd, bd->outbuf, bd->outbufPos) != bd->outbufPos) |
||||
longjmp(bd->jmpbuf, RETVAL_UNEXPECTED_OUTPUT_EOF); |
||||
bd->outbufPos = 0; |
||||
} |
||||
} |
||||
|
||||
/* Undo burrows-wheeler transform on intermediate buffer to produce output.
|
||||
If !len, write up to len bytes of data to buf. Otherwise write to out_fd. |
||||
Returns len ? bytes written : RETVAL_OK. Notice all errors negative #'s. */ |
||||
static int write_bunzip_data( |
||||
bunzip_data *bd, int64_t out_fd, char *outbuf, int len) |
||||
{ |
||||
unsigned int *dbuf = bd->dbuf; |
||||
int count, pos, current, run, copies, outbyte, previous, gotcount = 0; |
||||
|
||||
for (;;) { |
||||
/* If last read was short due to end of file, return last block now */ |
||||
if (bd->writeCount < 0) |
||||
return bd->writeCount; |
||||
/* If we need to refill dbuf, do it. */ |
||||
if (!bd->writeCount) { |
||||
int i = read_bunzip_data(bd); |
||||
if (i) { |
||||
if (i == RETVAL_LAST_BLOCK) { |
||||
bd->writeCount = i; |
||||
return gotcount; |
||||
} else |
||||
return i; |
||||
} |
||||
} |
||||
/* Loop generating output */ |
||||
count = bd->writeCount; |
||||
pos = bd->writePos; |
||||
current = bd->writeCurrent; |
||||
run = bd->writeRun; |
||||
while (count) { |
||||
/* If somebody (like busybox tar) wants a certain number of
|
||||
bytes of data from memory instead of written to a file, |
||||
humor them */ |
||||
if (len && bd->outbufPos >= len) |
||||
goto dataus_interruptus; |
||||
count--; |
||||
/* Follow sequence vector to undo Burrows-Wheeler transform */ |
||||
previous = current; |
||||
pos = dbuf[pos]; |
||||
current = pos & 0xff; |
||||
pos >>= 8; |
||||
/* Whenever we see 3 consecutive copies of the same byte,
|
||||
the 4th is a repeat count */ |
||||
if (run++ == 3) { |
||||
copies = current; |
||||
outbyte = previous; |
||||
current = -1; |
||||
} else { |
||||
copies = 1; |
||||
outbyte = current; |
||||
} |
||||
/* Output bytes to buffer, flushing to file if necessary */ |
||||
while (copies--) { |
||||
if (bd->outbufPos == IOBUF_SIZE) |
||||
flush_bunzip_outbuf(bd, out_fd); |
||||
bd->outbuf[bd->outbufPos++] = outbyte; |
||||
bd->dataCRC = (bd->dataCRC << 8) |
||||
^ bd->crc32Table[(bd->dataCRC >> 24) ^ outbyte]; |
||||
} |
||||
if (current != previous) |
||||
run = 0; |
||||
} |
||||
/* Decompression of this block completed successfully */ |
||||
bd->dataCRC = ~(bd->dataCRC); |
||||
bd->totalCRC |
||||
= ((bd->totalCRC << 1) | (bd->totalCRC >> 31)) ^ bd->dataCRC; |
||||
/* If this block had a CRC error, force file level CRC error. */ |
||||
if (bd->dataCRC != bd->headerCRC) { |
||||
bd->totalCRC = bd->headerCRC + 1; |
||||
return RETVAL_LAST_BLOCK; |
||||
} |
||||
dataus_interruptus: |
||||
bd->writeCount = count; |
||||
if (len) { |
||||
gotcount += bd->outbufPos; |
||||
memcpy(outbuf, bd->outbuf, len); |
||||
/* If we got enough data, checkpoint loop state and return */ |
||||
if ((len -= bd->outbufPos) < 1) { |
||||
bd->outbufPos -= len; |
||||
if (bd->outbufPos) |
||||
memmove(bd->outbuf, bd->outbuf + len, bd->outbufPos); |
||||
bd->writePos = pos; |
||||
bd->writeCurrent = current; |
||||
bd->writeRun = run; |
||||
return gotcount; |
||||
} |
||||
} |
||||
} |
||||
} |
||||
|
||||
/* Allocate the structure, read file header. If !len, src_fd contains
|
||||
filehandle to read from. Else inbuf contains data. */ |
||||
static int start_bunzip(bunzip_data **bdp, int64_t src_fd, char *inbuf, int len) |
||||
{ |
||||
bunzip_data *bd; |
||||
unsigned int i, j, c; |
||||
|
||||
/* Figure out how much data to allocate */ |
||||
i = sizeof(bunzip_data); |
||||
if (!len) |
||||
i += IOBUF_SIZE; |
||||
/* Allocate bunzip_data. Most fields initialize to zero. */ |
||||
if (!(bd = *bdp = malloc(i))) |
||||
return RETVAL_OUT_OF_MEMORY; |
||||
memset(bd, 0, sizeof(bunzip_data)); |
||||
if (len) { |
||||
bd->inbuf = (unsigned char *)inbuf; |
||||
bd->inbufCount = len; |
||||
bd->in_fd = -1; |
||||
} else { |
||||
bd->inbuf = (unsigned char *)(bd + 1); |
||||
bd->in_fd = src_fd; |
||||
} |
||||
/* Init the CRC32 table (big endian) */ |
||||
for (i = 0; i < 256; i++) { |
||||
c = i << 24; |
||||
for (j = 8; j; j--) |
||||
c = c & 0x80000000 ? (c << 1) ^ 0x04c11db7 : (c << 1); |
||||
bd->crc32Table[i] = c; |
||||
} |
||||
/* Setup for I/O error handling via longjmp */ |
||||
i = setjmp(bd->jmpbuf); |
||||
if (i) |
||||
return i; |
||||
/* Ensure that file starts with "BZh" */ |
||||
for (i = 0; i < 3; i++) |
||||
if (get_bits(bd, 8) != "BZh"[i]) |
||||
return RETVAL_NOT_BZIP_DATA; |
||||
/* Next byte ascii '1'-'9', indicates block size in units of 100k of
|
||||
uncompressed data. Allocate intermediate buffer for block. */ |
||||
i = get_bits(bd, 8); |
||||
if (i < '1' || i > '9') |
||||
return RETVAL_NOT_BZIP_DATA; |
||||
bd->dbufSize = 100000 * (i - '0'); |
||||
if (!(bd->dbuf = malloc(bd->dbufSize * sizeof(int)))) |
||||
return RETVAL_OUT_OF_MEMORY; |
||||
return RETVAL_OK; |
||||
} |
||||
|
||||
/* Example usage: decompress src_fd to dst_fd. (Stops at end of bzip data,
|
||||
not end of file.) */ |
||||
static char *uncompressStream(int64_t src_fd, int64_t dst_fd) |
||||
{ |
||||
bunzip_data *bd; |
||||
int i; |
||||
if (!(i = start_bunzip(&bd, src_fd, 0, 0))) { |
||||
i = write_bunzip_data(bd, dst_fd, 0, 0); |
||||
if (i == RETVAL_LAST_BLOCK && bd->headerCRC == bd->totalCRC) |
||||
i = RETVAL_OK; |
||||
} |
||||
flush_bunzip_outbuf(bd, dst_fd); |
||||
if (bd->dbuf) |
||||
free(bd->dbuf); |
||||
free(bd); |
||||
return bunzip_errors[-i]; |
||||
} |
||||
|
||||
int main(int argc, char *argv[]) |
||||
{ |
||||
char *err; |
||||
if (!(err = uncompressStream(STDIN_FILENO, STDOUT_FILENO))) { |
||||
return 0; |
||||
} else { |
||||
fprintf(stderr, "\n%s\n", err); |
||||
return 1; |
||||
} |
||||
} |
||||
@ -1,26 +0,0 @@ |
||||
Copyright 2006-2011, the V8 project authors. All rights reserved. |
||||
Redistribution and use in source and binary forms, with or without |
||||
modification, are permitted provided that the following conditions are |
||||
met: |
||||
|
||||
* Redistributions of source code must retain the above copyright |
||||
notice, this list of conditions and the following disclaimer. |
||||
* Redistributions in binary form must reproduce the above |
||||
copyright notice, this list of conditions and the following |
||||
disclaimer in the documentation and/or other materials provided |
||||
with the distribution. |
||||
* Neither the name of Google Inc. nor the names of its |
||||
contributors may be used to endorse or promote products derived |
||||
from this software without specific prior written permission. |
||||
|
||||
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
||||
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
||||
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
||||
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
||||
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
||||
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
||||
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
||||
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
||||
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
||||
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
||||
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
||||
@ -1 +0,0 @@ |
||||
google/double-conversion@1dce44c4313a6f356fcfa4b3e8887f037ac0bf23 |
||||
@ -1,621 +0,0 @@ |
||||
// Copyright 2010 the V8 project authors. All rights reserved.
|
||||
// Redistribution and use in source and binary forms, with or without
|
||||
// modification, are permitted provided that the following conditions are
|
||||
// met:
|
||||
//
|
||||
// * Redistributions of source code must retain the above copyright
|
||||
// notice, this list of conditions and the following disclaimer.
|
||||
// * Redistributions in binary form must reproduce the above
|
||||
// copyright notice, this list of conditions and the following
|
||||
// disclaimer in the documentation and/or other materials provided
|
||||
// with the distribution.
|
||||
// * Neither the name of Google Inc. nor the names of its
|
||||
// contributors may be used to endorse or promote products derived
|
||||
// from this software without specific prior written permission.
|
||||
//
|
||||
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
||||
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
||||
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
||||
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
|
||||
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
||||
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
||||
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
||||
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
||||
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
|
||||
#include "libc/macros.h" |
||||
#include "libc/math.h" |
||||
#include "third_party/double-conversion/bignum-dtoa.h" |
||||
#include "third_party/double-conversion/bignum.h" |
||||
#include "third_party/double-conversion/ieee.h" |
||||
|
||||
asm(".ident\t\"\\n\\n\
|
||||
double-conversion (BSD-3)\\n\
|
||||
Copyright 2010 the V8 project authors\""); |
||||
asm(".include \"libc/disclaimer.inc\""); |
||||
|
||||
namespace double_conversion { |
||||
|
||||
static int NormalizedExponent(uint64_t significand, int exponent) { |
||||
DOUBLE_CONVERSION_ASSERT(significand != 0); |
||||
while ((significand & Double::kHiddenBit) == 0) { |
||||
significand = significand << 1; |
||||
exponent = exponent - 1; |
||||
} |
||||
return exponent; |
||||
} |
||||
|
||||
// Forward declarations:
|
||||
// Returns an estimation of k such that 10^(k-1) <= v < 10^k.
|
||||
static int EstimatePower(int exponent); |
||||
// Computes v / 10^estimated_power exactly, as a ratio of two bignums, numerator
|
||||
// and denominator.
|
||||
static void InitialScaledStartValues(uint64_t significand, int exponent, |
||||
bool lower_boundary_is_closer, |
||||
int estimated_power, |
||||
bool need_boundary_deltas, |
||||
Bignum* numerator, Bignum* denominator, |
||||
Bignum* delta_minus, Bignum* delta_plus); |
||||
// Multiplies numerator/denominator so that its values lies in the range 1-10.
|
||||
// Returns decimal_point s.t.
|
||||
// v = numerator'/denominator' * 10^(decimal_point-1)
|
||||
// where numerator' and denominator' are the values of numerator and
|
||||
// denominator after the call to this function.
|
||||
static void FixupMultiply10(int estimated_power, bool is_even, |
||||
int* decimal_point, Bignum* numerator, |
||||
Bignum* denominator, Bignum* delta_minus, |
||||
Bignum* delta_plus); |
||||
// Generates digits from the left to the right and stops when the generated
|
||||
// digits yield the shortest decimal representation of v.
|
||||
static void GenerateShortestDigits(Bignum* numerator, Bignum* denominator, |
||||
Bignum* delta_minus, Bignum* delta_plus, |
||||
bool is_even, Vector<char> buffer, |
||||
int* length); |
||||
// Generates 'requested_digits' after the decimal point.
|
||||
static void BignumToFixed(int requested_digits, int* decimal_point, |
||||
Bignum* numerator, Bignum* denominator, |
||||
Vector<char> buffer, int* length); |
||||
// Generates 'count' digits of numerator/denominator.
|
||||
// Once 'count' digits have been produced rounds the result depending on the
|
||||
// remainder (remainders of exactly .5 round upwards). Might update the
|
||||
// decimal_point when rounding up (for example for 0.9999).
|
||||
static void GenerateCountedDigits(int count, int* decimal_point, |
||||
Bignum* numerator, Bignum* denominator, |
||||
Vector<char> buffer, int* length); |
||||
|
||||
void BignumDtoa(double v, BignumDtoaMode mode, int requested_digits, |
||||
Vector<char> buffer, int* length, int* decimal_point) { |
||||
DOUBLE_CONVERSION_ASSERT(v > 0); |
||||
DOUBLE_CONVERSION_ASSERT(!Double(v).IsSpecial()); |
||||
uint64_t significand; |
||||
int exponent; |
||||
bool lower_boundary_is_closer; |
||||
if (mode == BIGNUM_DTOA_SHORTEST_SINGLE) { |
||||
float f = static_cast<float>(v); |
||||
DOUBLE_CONVERSION_ASSERT(f == v); |
||||
significand = Single(f).Significand(); |
||||
exponent = Single(f).Exponent(); |
||||
lower_boundary_is_closer = Single(f).LowerBoundaryIsCloser(); |
||||
} else { |
||||
significand = Double(v).Significand(); |
||||
exponent = Double(v).Exponent(); |
||||
lower_boundary_is_closer = Double(v).LowerBoundaryIsCloser(); |
||||
} |
||||
bool need_boundary_deltas = |
||||
(mode == BIGNUM_DTOA_SHORTEST || mode == BIGNUM_DTOA_SHORTEST_SINGLE); |
||||
|
||||
bool is_even = (significand & 1) == 0; |
||||
int normalized_exponent = NormalizedExponent(significand, exponent); |
||||
// estimated_power might be too low by 1.
|
||||
int estimated_power = EstimatePower(normalized_exponent); |
||||
|
||||
// Shortcut for Fixed.
|
||||
// The requested digits correspond to the digits after the point. If the
|
||||
// number is much too small, then there is no need in trying to get any
|
||||
// digits.
|
||||
if (mode == BIGNUM_DTOA_FIXED && -estimated_power - 1 > requested_digits) { |
||||
buffer[0] = '\0'; |
||||
*length = 0; |
||||
// Set decimal-point to -requested_digits. This is what Gay does.
|
||||
// Note that it should not have any effect anyways since the string is
|
||||
// empty.
|
||||
*decimal_point = -requested_digits; |
||||
return; |
||||
} |
||||
|
||||
Bignum numerator; |
||||
Bignum denominator; |
||||
Bignum delta_minus; |
||||
Bignum delta_plus; |
||||
// Make sure the bignum can grow large enough. The smallest double equals
|
||||
// 4e-324. In this case the denominator needs fewer than 324*4 binary digits.
|
||||
// The maximum double is 1.7976931348623157e308 which needs fewer than
|
||||
// 308*4 binary digits.
|
||||
DOUBLE_CONVERSION_ASSERT(Bignum::kMaxSignificantBits >= 324 * 4); |
||||
InitialScaledStartValues(significand, exponent, lower_boundary_is_closer, |
||||
estimated_power, need_boundary_deltas, &numerator, |
||||
&denominator, &delta_minus, &delta_plus); |
||||
// We now have v = (numerator / denominator) * 10^estimated_power.
|
||||
FixupMultiply10(estimated_power, is_even, decimal_point, &numerator, |
||||
&denominator, &delta_minus, &delta_plus); |
||||
// We now have v = (numerator / denominator) * 10^(decimal_point-1), and
|
||||
// 1 <= (numerator + delta_plus) / denominator < 10
|
||||
switch (mode) { |
||||
case BIGNUM_DTOA_SHORTEST: |
||||
case BIGNUM_DTOA_SHORTEST_SINGLE: |
||||
GenerateShortestDigits(&numerator, &denominator, &delta_minus, |
||||
&delta_plus, is_even, buffer, length); |
||||
break; |
||||
case BIGNUM_DTOA_FIXED: |
||||
BignumToFixed(requested_digits, decimal_point, &numerator, &denominator, |
||||
buffer, length); |
||||
break; |
||||
case BIGNUM_DTOA_PRECISION: |
||||
GenerateCountedDigits(requested_digits, decimal_point, &numerator, |
||||
&denominator, buffer, length); |
||||
break; |
||||
default: |
||||
DOUBLE_CONVERSION_UNREACHABLE(); |
||||
} |
||||
buffer[*length] = '\0'; |
||||
} |
||||
|
||||
// The procedure starts generating digits from the left to the right and stops
|
||||
// when the generated digits yield the shortest decimal representation of v. A
|
||||
// decimal representation of v is a number lying closer to v than to any other
|
||||
// double, so it converts to v when read.
|
||||
//
|
||||
// This is true if d, the decimal representation, is between m- and m+, the
|
||||
// upper and lower boundaries. d must be strictly between them if !is_even.
|
||||
// m- := (numerator - delta_minus) / denominator
|
||||
// m+ := (numerator + delta_plus) / denominator
|
||||
//
|
||||
// Precondition: 0 <= (numerator+delta_plus) / denominator < 10.
|
||||
// If 1 <= (numerator+delta_plus) / denominator < 10 then no leading 0 digit
|
||||
// will be produced. This should be the standard precondition.
|
||||
static void GenerateShortestDigits(Bignum* numerator, Bignum* denominator, |
||||
Bignum* delta_minus, Bignum* delta_plus, |
||||
bool is_even, Vector<char> buffer, |
||||
int* length) { |
||||
// Small optimization: if delta_minus and delta_plus are the same just reuse
|
||||
// one of the two bignums.
|
||||
if (Bignum::Equal(*delta_minus, *delta_plus)) { |
||||
delta_plus = delta_minus; |
||||
} |
||||
*length = 0; |
||||
for (;;) { |
||||
uint16_t digit; |
||||
digit = numerator->DivideModuloIntBignum(*denominator); |
||||
DOUBLE_CONVERSION_ASSERT( |
||||
digit <= 9); // digit is a uint16_t and therefore always positive.
|
||||
// digit = numerator / denominator (integer division).
|
||||
// numerator = numerator % denominator.
|
||||
buffer[(*length)++] = static_cast<char>(digit + '0'); |
||||
|
||||
// Can we stop already?
|
||||
// If the remainder of the division is less than the distance to the lower
|
||||
// boundary we can stop. In this case we simply round down (discarding the
|
||||
// remainder).
|
||||
// Similarly we test if we can round up (using the upper boundary).
|
||||
bool in_delta_room_minus; |
||||
bool in_delta_room_plus; |
||||
if (is_even) { |
||||
in_delta_room_minus = Bignum::LessEqual(*numerator, *delta_minus); |
||||
} else { |
||||
in_delta_room_minus = Bignum::Less(*numerator, *delta_minus); |
||||
} |
||||
if (is_even) { |
||||
in_delta_room_plus = |
||||
Bignum::PlusCompare(*numerator, *delta_plus, *denominator) >= 0; |
||||
} else { |
||||
in_delta_room_plus = |
||||
Bignum::PlusCompare(*numerator, *delta_plus, *denominator) > 0; |
||||
} |
||||
if (!in_delta_room_minus && !in_delta_room_plus) { |
||||
// Prepare for next iteration.
|
||||
numerator->Times10(); |
||||
delta_minus->Times10(); |
||||
// We optimized delta_plus to be equal to delta_minus (if they share the
|
||||
// same value). So don't multiply delta_plus if they point to the same
|
||||
// object.
|
||||
if (delta_minus != delta_plus) { |
||||
delta_plus->Times10(); |
||||
} |
||||
} else if (in_delta_room_minus && in_delta_room_plus) { |
||||
// Let's see if 2*numerator < denominator.
|
||||
// If yes, then the next digit would be < 5 and we can round down.
|
||||
int compare = Bignum::PlusCompare(*numerator, *numerator, *denominator); |
||||
if (compare < 0) { |
||||
// Remaining digits are less than .5. -> Round down (== do nothing).
|
||||
} else if (compare > 0) { |
||||
// Remaining digits are more than .5 of denominator. -> Round up.
|
||||
// Note that the last digit could not be a '9' as otherwise the whole
|
||||
// loop would have stopped earlier.
|
||||
// We still have an assert here in case the preconditions were not
|
||||
// satisfied.
|
||||
DOUBLE_CONVERSION_ASSERT(buffer[(*length) - 1] != '9'); |
||||
buffer[(*length) - 1]++; |
||||
} else { |
||||
// Halfway case.
|
||||
// TODO(floitsch): need a way to solve half-way cases.
|
||||
// For now let's round towards even (since this is what Gay seems to
|
||||
// do).
|
||||
|
||||
if ((buffer[(*length) - 1] - '0') % 2 == 0) { |
||||
// Round down => Do nothing.
|
||||
} else { |
||||
DOUBLE_CONVERSION_ASSERT(buffer[(*length) - 1] != '9'); |
||||
buffer[(*length) - 1]++; |
||||
} |
||||
} |
||||
return; |
||||
} else if (in_delta_room_minus) { |
||||
// Round down (== do nothing).
|
||||
return; |
||||
} else { // in_delta_room_plus
|
||||
// Round up.
|
||||
// Note again that the last digit could not be '9' since this would have
|
||||
// stopped the loop earlier.
|
||||
// We still have an DOUBLE_CONVERSION_ASSERT here, in case the
|
||||
// preconditions were not satisfied.
|
||||
DOUBLE_CONVERSION_ASSERT(buffer[(*length) - 1] != '9'); |
||||
buffer[(*length) - 1]++; |
||||
return; |
||||
} |
||||
} |
||||
} |
||||
|
||||
// Let v = numerator / denominator < 10.
|
||||
// Then we generate 'count' digits of d = x.xxxxx... (without the decimal point)
|
||||
// from left to right. Once 'count' digits have been produced we decide wether
|
||||
// to round up or down. Remainders of exactly .5 round upwards. Numbers such
|
||||
// as 9.999999 propagate a carry all the way, and change the
|
||||
// exponent (decimal_point), when rounding upwards.
|
||||
static void GenerateCountedDigits(int count, int* decimal_point, |
||||
Bignum* numerator, Bignum* denominator, |
||||
Vector<char> buffer, int* length) { |
||||
DOUBLE_CONVERSION_ASSERT(count >= 0); |
||||
for (int i = 0; i < count - 1; ++i) { |
||||
uint16_t digit; |
||||
digit = numerator->DivideModuloIntBignum(*denominator); |
||||
DOUBLE_CONVERSION_ASSERT( |
||||
digit <= 9); // digit is a uint16_t and therefore always positive.
|
||||
// digit = numerator / denominator (integer division).
|
||||
// numerator = numerator % denominator.
|
||||
buffer[i] = static_cast<char>(digit + '0'); |
||||
// Prepare for next iteration.
|
||||
numerator->Times10(); |
||||
} |
||||
// Generate the last digit.
|
||||
uint16_t digit; |
||||
digit = numerator->DivideModuloIntBignum(*denominator); |
||||
if (Bignum::PlusCompare(*numerator, *numerator, *denominator) >= 0) { |
||||
digit++; |
||||
} |
||||
DOUBLE_CONVERSION_ASSERT(digit <= 10); |
||||
buffer[count - 1] = static_cast<char>(digit + '0'); |
||||
// Correct bad digits (in case we had a sequence of '9's). Propagate the
|
||||
// carry until we hat a non-'9' or til we reach the first digit.
|
||||
for (int i = count - 1; i > 0; --i) { |
||||
if (buffer[i] != '0' + 10) break; |
||||
buffer[i] = '0'; |
||||
buffer[i - 1]++; |
||||
} |
||||
if (buffer[0] == '0' + 10) { |
||||
// Propagate a carry past the top place.
|
||||
buffer[0] = '1'; |
||||
(*decimal_point)++; |
||||
} |
||||
*length = count; |
||||
} |
||||
|
||||
// Generates 'requested_digits' after the decimal point. It might omit
|
||||
// trailing '0's. If the input number is too small then no digits at all are
|
||||
// generated (ex.: 2 fixed digits for 0.00001).
|
||||
//
|
||||
// Input verifies: 1 <= (numerator + delta) / denominator < 10.
|
||||
static void BignumToFixed(int requested_digits, int* decimal_point, |
||||
Bignum* numerator, Bignum* denominator, |
||||
Vector<char> buffer, int* length) { |
||||
// Note that we have to look at more than just the requested_digits, since
|
||||
// a number could be rounded up. Example: v=0.5 with requested_digits=0.
|
||||
// Even though the power of v equals 0 we can't just stop here.
|
||||
if (-(*decimal_point) > requested_digits) { |
||||
// The number is definitively too small.
|
||||
// Ex: 0.001 with requested_digits == 1.
|
||||
// Set decimal-point to -requested_digits. This is what Gay does.
|
||||
// Note that it should not have any effect anyways since the string is
|
||||
// empty.
|
||||
*decimal_point = -requested_digits; |
||||
*length = 0; |
||||
return; |
||||
} else if (-(*decimal_point) == requested_digits) { |
||||
// We only need to verify if the number rounds down or up.
|
||||
// Ex: 0.04 and 0.06 with requested_digits == 1.
|
||||
DOUBLE_CONVERSION_ASSERT(*decimal_point == -requested_digits); |
||||
// Initially the fraction lies in range (1, 10]. Multiply the denominator
|
||||
// by 10 so that we can compare more easily.
|
||||
denominator->Times10(); |
||||
if (Bignum::PlusCompare(*numerator, *numerator, *denominator) >= 0) { |
||||
// If the fraction is >= 0.5 then we have to include the rounded
|
||||
// digit.
|
||||
buffer[0] = '1'; |
||||
*length = 1; |
||||
(*decimal_point)++; |
||||
} else { |
||||
// Note that we caught most of similar cases earlier.
|
||||
*length = 0; |
||||
} |
||||
return; |
||||
} else { |
||||
// The requested digits correspond to the digits after the point.
|
||||
// The variable 'needed_digits' includes the digits before the point.
|
||||
int needed_digits = (*decimal_point) + requested_digits; |
||||
GenerateCountedDigits(needed_digits, decimal_point, numerator, denominator, |
||||
buffer, length); |
||||
} |
||||
} |
||||
|
||||
// Returns an estimation of k such that 10^(k-1) <= v < 10^k where
|
||||
// v = f * 2^exponent and 2^52 <= f < 2^53.
|
||||
// v is hence a normalized double with the given exponent. The output is an
|
||||
// approximation for the exponent of the decimal approimation .digits * 10^k.
|
||||
//
|
||||
// The result might undershoot by 1 in which case 10^k <= v < 10^k+1.
|
||||
// Note: this property holds for v's upper boundary m+ too.
|
||||
// 10^k <= m+ < 10^k+1.
|
||||
// (see explanation below).
|
||||
//
|
||||
// Examples:
|
||||
// EstimatePower(0) => 16
|
||||
// EstimatePower(-52) => 0
|
||||
//
|
||||
// Note: e >= 0 => EstimatedPower(e) > 0. No similar claim can be made for e<0.
|
||||
static int EstimatePower(int exponent) { |
||||
// This function estimates log10 of v where v = f*2^e (with e == exponent).
|
||||
// Note that 10^floor(log10(v)) <= v, but v <= 10^ceil(log10(v)).
|
||||
// Note that f is bounded by its container size. Let p = 53 (the double's
|
||||
// significand size). Then 2^(p-1) <= f < 2^p.
|
||||
//
|
||||
// Given that log10(v) == log2(v)/log2(10) and e+(len(f)-1) is quite close
|
||||
// to log2(v) the function is simplified to (e+(len(f)-1)/log2(10)).
|
||||
// The computed number undershoots by less than 0.631 (when we compute log3
|
||||
// and not log10).
|
||||
//
|
||||
// Optimization: since we only need an approximated result this computation
|
||||
// can be performed on 64 bit integers. On x86/x64 architecture the speedup is
|
||||
// not really measurable, though.
|
||||
//
|
||||
// Since we want to avoid overshooting we decrement by 1e10 so that
|
||||
// floating-point imprecisions don't affect us.
|
||||
//
|
||||
// Explanation for v's boundary m+: the computation takes advantage of
|
||||
// the fact that 2^(p-1) <= f < 2^p. Boundaries still satisfy this requirement
|
||||
// (even for denormals where the delta can be much more important).
|
||||
|
||||
const double k1Log10 = 0.30102999566398114; // 1/lg(10)
|
||||
|
||||
// For doubles len(f) == 53 (don't forget the hidden bit).
|
||||
const int kSignificandSize = Double::kSignificandSize; |
||||
double estimate = ceil((exponent + kSignificandSize - 1) * k1Log10 - 1e-10); |
||||
return static_cast<int>(estimate); |
||||
} |
||||
|
||||
// See comments for InitialScaledStartValues.
|
||||
static void InitialScaledStartValuesPositiveExponent( |
||||
uint64_t significand, int exponent, int estimated_power, |
||||
bool need_boundary_deltas, Bignum* numerator, Bignum* denominator, |
||||
Bignum* delta_minus, Bignum* delta_plus) { |
||||
// A positive exponent implies a positive power.
|
||||
DOUBLE_CONVERSION_ASSERT(estimated_power >= 0); |
||||
// Since the estimated_power is positive we simply multiply the denominator
|
||||
// by 10^estimated_power.
|
||||
|
||||
// numerator = v.
|
||||
numerator->AssignUInt64(significand); |
||||
numerator->ShiftLeft(exponent); |
||||
// denominator = 10^estimated_power.
|
||||
denominator->AssignPowerUInt16(10, estimated_power); |
||||
|
||||
if (need_boundary_deltas) { |
||||
// Introduce a common denominator so that the deltas to the boundaries are
|
||||
// integers.
|
||||
denominator->ShiftLeft(1); |
||||
numerator->ShiftLeft(1); |
||||
// Let v = f * 2^e, then m+ - v = 1/2 * 2^e; With the common
|
||||
// denominator (of 2) delta_plus equals 2^e.
|
||||
delta_plus->AssignUInt16(1); |
||||
delta_plus->ShiftLeft(exponent); |
||||
// Same for delta_minus. The adjustments if f == 2^p-1 are done later.
|
||||
delta_minus->AssignUInt16(1); |
||||
delta_minus->ShiftLeft(exponent); |
||||
} |
||||
} |
||||
|
||||
// See comments for InitialScaledStartValues
|
||||
static void InitialScaledStartValuesNegativeExponentPositivePower( |
||||
uint64_t significand, int exponent, int estimated_power, |
||||
bool need_boundary_deltas, Bignum* numerator, Bignum* denominator, |
||||
Bignum* delta_minus, Bignum* delta_plus) { |
||||
// v = f * 2^e with e < 0, and with estimated_power >= 0.
|
||||
// This means that e is close to 0 (have a look at how estimated_power is
|
||||
// computed).
|
||||
|
||||
// numerator = significand
|
||||
// since v = significand * 2^exponent this is equivalent to
|
||||
// numerator = v * / 2^-exponent
|
||||
numerator->AssignUInt64(significand); |
||||
// denominator = 10^estimated_power * 2^-exponent (with exponent < 0)
|
||||
denominator->AssignPowerUInt16(10, estimated_power); |
||||
denominator->ShiftLeft(-exponent); |
||||
|
||||
if (need_boundary_deltas) { |
||||
// Introduce a common denominator so that the deltas to the boundaries are
|
||||
// integers.
|
||||
denominator->ShiftLeft(1); |
||||
numerator->ShiftLeft(1); |
||||
// Let v = f * 2^e, then m+ - v = 1/2 * 2^e; With the common
|
||||
// denominator (of 2) delta_plus equals 2^e.
|
||||
// Given that the denominator already includes v's exponent the distance
|
||||
// to the boundaries is simply 1.
|
||||
delta_plus->AssignUInt16(1); |
||||
// Same for delta_minus. The adjustments if f == 2^p-1 are done later.
|
||||
delta_minus->AssignUInt16(1); |
||||
} |
||||
} |
||||
|
||||
// See comments for InitialScaledStartValues
|
||||
static void InitialScaledStartValuesNegativeExponentNegativePower( |
||||
uint64_t significand, int exponent, int estimated_power, |
||||
bool need_boundary_deltas, Bignum* numerator, Bignum* denominator, |
||||
Bignum* delta_minus, Bignum* delta_plus) { |
||||
// Instead of multiplying the denominator with 10^estimated_power we
|
||||
// multiply all values (numerator and deltas) by 10^-estimated_power.
|
||||
|
||||
// Use numerator as temporary container for power_ten.
|
||||
Bignum* power_ten = numerator; |
||||
power_ten->AssignPowerUInt16(10, -estimated_power); |
||||
|
||||
if (need_boundary_deltas) { |
||||
// Since power_ten == numerator we must make a copy of 10^estimated_power
|
||||
// before we complete the computation of the numerator.
|
||||
// delta_plus = delta_minus = 10^estimated_power
|
||||
delta_plus->AssignBignum(*power_ten); |
||||
delta_minus->AssignBignum(*power_ten); |
||||
} |
||||
|
||||
// numerator = significand * 2 * 10^-estimated_power
|
||||
// since v = significand * 2^exponent this is equivalent to
|
||||
// numerator = v * 10^-estimated_power * 2 * 2^-exponent.
|
||||
// Remember: numerator has been abused as power_ten. So no need to assign it
|
||||
// to itself.
|
||||
DOUBLE_CONVERSION_ASSERT(numerator == power_ten); |
||||
numerator->MultiplyByUInt64(significand); |
||||
|
||||
// denominator = 2 * 2^-exponent with exponent < 0.
|
||||
denominator->AssignUInt16(1); |
||||
denominator->ShiftLeft(-exponent); |
||||
|
||||
if (need_boundary_deltas) { |
||||
// Introduce a common denominator so that the deltas to the boundaries are
|
||||
// integers.
|
||||
numerator->ShiftLeft(1); |
||||
denominator->ShiftLeft(1); |
||||
// With this shift the boundaries have their correct value, since
|
||||
// delta_plus = 10^-estimated_power, and
|
||||
// delta_minus = 10^-estimated_power.
|
||||
// These assignments have been done earlier.
|
||||
// The adjustments if f == 2^p-1 (lower boundary is closer) are done later.
|
||||
} |
||||
} |
||||
|
||||
// Let v = significand * 2^exponent.
|
||||
// Computes v / 10^estimated_power exactly, as a ratio of two bignums, numerator
|
||||
// and denominator. The functions GenerateShortestDigits and
|
||||
// GenerateCountedDigits will then convert this ratio to its decimal
|
||||
// representation d, with the required accuracy.
|
||||
// Then d * 10^estimated_power is the representation of v.
|
||||
// (Note: the fraction and the estimated_power might get adjusted before
|
||||
// generating the decimal representation.)
|
||||
//
|
||||
// The initial start values consist of:
|
||||
// - a scaled numerator: s.t. numerator/denominator == v / 10^estimated_power.
|
||||
// - a scaled (common) denominator.
|
||||
// optionally (used by GenerateShortes | ||||