scuffed-code/icu4c/source/i18n/repattrn.cpp

573 lines
16 KiB
C++

//
// file: repattrn.cpp
//
/*
**********************************************************************
* Copyright (C) 2002 International Business Machines Corporation *
* and others. All rights reserved. *
**********************************************************************
*/
#include "unicode/utypes.h"
#if !UCONFIG_NO_REGULAR_EXPRESSIONS
#include "unicode/regex.h"
#include "uassert.h"
#include "uvector.h"
#include "uvectr32.h"
#include "regexcmp.h"
#include "regeximp.h"
U_NAMESPACE_BEGIN
//--------------------------------------------------------------------------
//
// RegexPattern Default Constructor
//
//--------------------------------------------------------------------------
RegexPattern::RegexPattern() {
init();
};
//--------------------------------------------------------------------------
//
// Copy Constructor Note: This is a rather inefficient implementation,
// but it probably doesn't matter.
//
//--------------------------------------------------------------------------
RegexPattern::RegexPattern(const RegexPattern &other) : UObject(other) {
init();
*this = other;
}
//--------------------------------------------------------------------------
//
// Assignmenet Operator
//
//--------------------------------------------------------------------------
RegexPattern &RegexPattern::operator = (const RegexPattern &other) {
if (this == &other) {
// Source and destination are the same. Don't do anything.
return *this;
}
// Clean out any previous contents of object being assigned to.
zap();
// Give target object a default initialization
init();
// Copy simple fields
fPattern = other.fPattern;
fFlags = other.fFlags;
fLiteralText = other.fLiteralText;
fBadState = other.fBadState;
fMaxCaptureDigits = other.fMaxCaptureDigits;
fStaticSets = other.fStaticSets;
if (fBadState) {
return *this;
}
// Copy the pattern. It's just values, nothing deep to copy.
// TODO: something with status
UErrorCode status = U_ZERO_ERROR;
fCompiledPat->assign(*other.fCompiledPat, status);
fGroupMap->assign(*other.fGroupMap, status);
// Note: do not copy fMatcher. It'll be created on first use if the
// destination needs one.
// Copy the Unicode Sets.
// Could be made more efficient if the sets were reference counted and shared,
// but I doubt that pattern copying will be particularly common.
// Note: init() already added an empty element zero to fSets
int32_t i;
for (i=1; i<other.fSets->size(); i++) {
UnicodeSet *sourceSet = (UnicodeSet *)other.fSets->elementAt(i);
UnicodeSet *newSet = new UnicodeSet(*sourceSet);
if (newSet == NULL) {
fBadState = TRUE;
break;
}
fSets->addElement(newSet, status);
}
if (U_FAILURE(status)) {
fBadState = TRUE;
}
return *this;
}
//--------------------------------------------------------------------------
//
// init Shared initialization for use by constructors.
// Bring an uninitialized RegexPattern up to a default state.
//
//--------------------------------------------------------------------------
void RegexPattern::init() {
fFlags = 0;
fBadState = FALSE;
fMaxCaptureDigits = 1; // TODO: calculate for real.
fStaticSets = NULL;
fMatcher = NULL;
fFrameSize = 0;
fDataSize = 0;
UErrorCode status=U_ZERO_ERROR;
// Init of a completely new RegexPattern.
fCompiledPat = new UVector32(status);
fGroupMap = new UVector32(status);
fSets = new UVector(status);
if (U_FAILURE(status) || fCompiledPat == NULL || fSets == NULL) {
fBadState = TRUE;
return;
}
// Slot zero of the vector of sets is reserved. Fill it here.
fSets->addElement((int32_t)0, status);
}
//--------------------------------------------------------------------------
//
// zap Delete everything owned by this RegexPattern.
//
//--------------------------------------------------------------------------
void RegexPattern::zap() {
delete fMatcher;
fMatcher = NULL;
delete fCompiledPat;
fCompiledPat = NULL;
int i;
for (i=1; i<fSets->size(); i++) {
UnicodeSet *s;
s = (UnicodeSet *)fSets->elementAt(i);
if (s != NULL) {
delete s;
}
}
delete fSets;
fSets = NULL;
delete fGroupMap;
fGroupMap = NULL;
}
//--------------------------------------------------------------------------
//
// Destructor
//
//--------------------------------------------------------------------------
RegexPattern::~RegexPattern() {
zap();
};
//--------------------------------------------------------------------------
//
// Clone
//
//--------------------------------------------------------------------------
RegexPattern *RegexPattern::clone() const {
RegexPattern *copy = new RegexPattern(*this);
return copy;
};
//--------------------------------------------------------------------------
//
// operator == (comparison) Consider to patterns to be == if the
// pattern strings and the flags are the same.
//
//--------------------------------------------------------------------------
UBool RegexPattern::operator ==(const RegexPattern &other) const {
UBool r = this->fFlags == other.fFlags &&
this->fPattern == other.fPattern &&
this->fBadState == FALSE &&
other.fBadState == FALSE;
return r;
}
//---------------------------------------------------------------------
//
// compile
//
//---------------------------------------------------------------------
RegexPattern *RegexPattern::compile(
const UnicodeString &regex,
uint32_t flags,
UParseError &pe,
UErrorCode &status) {
if (U_FAILURE(status)) {
return NULL;
}
if (flags != 0) {
status = U_REGEX_UNIMPLEMENTED;
return NULL;
}
RegexPattern *This = new RegexPattern;
if (This == NULL) {
status = U_MEMORY_ALLOCATION_ERROR;
return NULL;
}
if (This->fBadState) {
status = U_REGEX_INVALID_STATE;
return NULL;
}
This->fFlags = flags;
RegexCompile compiler(status);
compiler.compile(*This, regex, pe, status);
return This;
};
//
// compile with default flags.
//
RegexPattern *RegexPattern::compile( const UnicodeString &regex,
UParseError &pe,
UErrorCode &err)
{
return compile(regex, 0, pe, err);
}
//---------------------------------------------------------------------
//
// flags
//
//---------------------------------------------------------------------
uint32_t RegexPattern::flags() const {
return fFlags;
}
//---------------------------------------------------------------------
//
// matcher(UnicodeString, err)
//
//---------------------------------------------------------------------
RegexMatcher *RegexPattern::matcher(const UnicodeString &input,
UErrorCode &err) const {
RegexMatcher *retMatcher = NULL;
if (U_FAILURE(err)) {
return NULL;
}
if (fBadState) {
U_FAILURE(U_REGEX_INVALID_STATE);
return NULL;
}
retMatcher = new RegexMatcher(this);
if (retMatcher == NULL) {
err = U_MEMORY_ALLOCATION_ERROR;
return NULL;
}
retMatcher->reset(input);
return retMatcher;
};
//---------------------------------------------------------------------
//
// matches Convenience function to test for a match, starting
// with a pattern string and a data string.
//
//---------------------------------------------------------------------
UBool RegexPattern::matches(const UnicodeString &regex,
const UnicodeString &input,
UParseError &pe,
UErrorCode &status) {
if (U_FAILURE(status)) {return FALSE;}
UBool retVal;
RegexPattern *pat = NULL;
RegexMatcher *matcher = NULL;
pat = RegexPattern::compile(regex, 0, pe, status);
matcher = pat->matcher(input, status);
retVal = matcher->matches(status);
delete matcher;
delete pat;
return retVal;
}
//---------------------------------------------------------------------
//
// pattern
//
//---------------------------------------------------------------------
UnicodeString RegexPattern::pattern() const {
return fPattern;
}
//---------------------------------------------------------------------
//
// split
//
//---------------------------------------------------------------------
int32_t RegexPattern::split(const UnicodeString &input,
UnicodeString dest[],
int32_t destCapacity,
UErrorCode &status) const
{
//
// Check arguements for validity
//
if (U_FAILURE(status)) {
return 0;
};
if (destCapacity < 1) {
status = U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
//
// If we don't already have a cached matcher object from a previous call
// to split(), create one now.
//
if (fMatcher == NULL) {
RegexMatcher *m = matcher(input, status);
if (U_FAILURE(status)) {
return 0;
}
// Need to cast off const to cache the matcher
RegexPattern *nonConstThis = (RegexPattern *)this;
nonConstThis->fMatcher = m;
}
//
// Set our input text into the matcher
//
fMatcher->reset(input);
int32_t inputLen = input.length();
int32_t nextOutputStringStart = 0;
if (inputLen == 0) {
return 0;
}
//
// Loop through the input text, searching for the delimiter pattern
//
int i;
int32_t numCaptureGroups = fGroupMap->size();
for (i=0; ; i++) {
if (i==destCapacity-1) {
// There is only one output string left.
// Fill it with whatever is left from the input, then exit the loop.
dest[i].setTo(input, nextOutputStringStart, inputLen-nextOutputStringStart);
break;
}
if (fMatcher->find()) {
// We found another delimiter. Move everything from where we started looking
// up until the start of the delimiter into the next output string.
int32_t fieldLen = fMatcher->fMatchStart - nextOutputStringStart;
dest[i].setTo(input, nextOutputStringStart, fieldLen);
nextOutputStringStart = fMatcher->fMatchEnd;
// If the delimiter pattern has capturing parentheses, the captured
// text goes out into the next n destination strings.
int32_t groupNum;
for (groupNum=1; groupNum<=numCaptureGroups; groupNum++) {
if (i==destCapacity-1) {
break;
}
i++;
dest[i] = fMatcher->group(groupNum, status);
}
if (nextOutputStringStart == inputLen) {
// The delimiter was at the end of the string. We're done.
break;
}
if (i==destCapacity-1) {
// We've filled up the last output string with capture group data.
// Give back the last string, to be used for the remainder of the input.
i--;
}
}
else
{
// We ran off the end of the input while looking for the next delimiter.
// All the remaining text goes into the current output string.
dest[i].setTo(input, nextOutputStringStart, inputLen-nextOutputStringStart);
break;
}
}
return i+1;
}
//---------------------------------------------------------------------
//
// dump Output the compiled form of the pattern.
// Debugging function only.
//
//---------------------------------------------------------------------
static const char * const opNames[] = {URX_OPCODE_NAMES};
void RegexPattern::dumpOp(int32_t index) const {
int32_t op = fCompiledPat->elementAti(index);
int32_t val = URX_VAL(op);
int32_t type = URX_TYPE(op);
int32_t pinnedType = type;
if (pinnedType >= sizeof(opNames)/sizeof(char *)) {
pinnedType = 0;
}
REGEX_DUMP_DEBUG_PRINTF("%4d %08x %-15s ", index, op, opNames[pinnedType]);
switch (type) {
case URX_NOP:
case URX_DOTANY:
case URX_FAIL:
case URX_BACKSLASH_A:
case URX_BACKSLASH_G:
case URX_BACKSLASH_X:
case URX_END:
// Types with no operand field of interest.
break;
case URX_RESERVED_OP:
case URX_START_CAPTURE:
case URX_END_CAPTURE:
case URX_STATE_SAVE:
case URX_JMP:
case URX_BACKSLASH_B:
case URX_BACKSLASH_D:
case URX_BACKSLASH_W:
case URX_BACKSLASH_Z:
case URX_CARET:
case URX_DOLLAR:
case URX_STRING_LEN:
case URX_CTR_INIT:
case URX_CTR_INIT_NG:
case URX_CTR_INIT_P:
case URX_CTR_LOOP:
case URX_CTR_LOOP_NG:
case URX_CTR_LOOP_P:
case URX_RELOC_OPRND:
case URX_STO_SP:
case URX_LD_SP:
case URX_BACKREF:
case URX_STO_INP_LOC:
case URX_JMPX:
// types with an integer operand field.
REGEX_DUMP_DEBUG_PRINTF("%d", val);
break;
case URX_ONECHAR:
REGEX_DUMP_DEBUG_PRINTF("%c", val<256?val:'?');
break;
case URX_STRING:
{
int32_t lengthOp = fCompiledPat->elementAti(index+1);
U_ASSERT(URX_TYPE(lengthOp) == URX_STRING_LEN);
int32_t length = URX_VAL(lengthOp);
int32_t i;
for (i=val; i<val+length; i++) {
UChar c = fLiteralText[i];
if (c < 32 || c >= 256) {c = '.';}
REGEX_DUMP_DEBUG_PRINTF("%c", c);
}
}
break;
case URX_SETREF:
{
UnicodeString s;
UnicodeSet *set = (UnicodeSet *)fSets->elementAt(val);
set->toPattern(s, TRUE);
for (int32_t i=0; i<s.length(); i++) {
REGEX_DUMP_DEBUG_PRINTF("%c", s.charAt(i));
}
}
break;
case URX_STATIC_SETREF:
{
UnicodeString s;
if (val & URX_NEG_SET) {
REGEX_DUMP_DEBUG_PRINTF("NOT ");
val &= ~URX_NEG_SET;
}
UnicodeSet *set = fStaticSets[val];
set->toPattern(s, TRUE);
for (int32_t i=0; i<s.length(); i++) {
REGEX_DUMP_DEBUG_PRINTF("%c", s.charAt(i));
}
}
break;
default:
REGEX_DUMP_DEBUG_PRINTF("??????");
break;
}
REGEX_DUMP_DEBUG_PRINTF("\n");
}
void RegexPattern::dump() const {
int index;
int i;
REGEX_DUMP_DEBUG_PRINTF("Original Pattern: ");
for (i=0; i<fPattern.length(); i++) {
REGEX_DUMP_DEBUG_PRINTF("%c", fPattern.charAt(i));
}
REGEX_DUMP_DEBUG_PRINTF("\n");
REGEX_DUMP_DEBUG_PRINTF("Pattern Valid?: %s\n", fBadState? "no" : "yes");
REGEX_DUMP_DEBUG_PRINTF("\nIndex Binary Type Operand\n"
"-------------------------------------------\n");
for (index = 0; index<fCompiledPat->size(); index++) {
dumpOp(index);
}
REGEX_DUMP_DEBUG_PRINTF("\n\n");
};
const char RegexPattern::fgClassID = 0;
//----------------------------------------------------------------------------------
//
// regex_cleanup Memory cleanup function, free/delete all
// cached memory. Called by ICU's u_cleanup() function.
//
//----------------------------------------------------------------------------------
U_CFUNC UBool
regex_cleanup(void) {
RegexCompile::cleanup();
return TRUE;
};
U_NAMESPACE_END
#endif // !UCONFIG_NO_REGULAR_EXPRESSIONS