replicant-frameworks_native/libs/utils/String8.cpp
Jeff Brown a3477c862a Added support for full PC-style keyboards.
BREAKING CHANGE: Redesigned the key character map format to
accomodate full keyboards with more comprehensive suite of modifiers.
Old key character maps will not work anymore and must be updated.
The new format is plain text only and it not compiled to a binary
file (so the "kcm" tool will be removed in a subsequent check-in).

Added FULL keyboard type to support full PC-style keyboards.

Added SPECIAL_FUNCTION keyboard type to support special function
keypads that do not have any printable keys suitable for typing
and only have keys like HOME and POWER

Added a special VIRTUAL_KEYBOARD device id convention that maps
to a virtual keyboard with a fixed known layout.  This is designed
to work around issues injecting input events on devices whose
built-in keyboard does not have a useful key character map (ie.
when the built-in keyboard is a special function keyboard only.)

Modified several places where events were being synthesized
to use the virtual keyboard.

Removed support for the "qwerty" default layout.
The new default layout is "Generic".  For the most part "qwerty"
was being used as a backstop in case the built-in keyboard did
not have a key character map (probably because it was a special
function keypad) and the framework needed to be able to inject
key events anyways.  The latter issue is resolved by using the
special VIRTUAL_KEYBOARD device instead of BUILT_IN_KEYBOARD.

Added the concept of a key modifier behavior so that
MetaKeyKeyListener can distinguish between keyboards that use
chorded vs. toggled modifiers.

Wrote more robust key layout and key character map parsers
to enable support for new keyboard features and user installable
key maps.

Fixed a bug in InputReader generating key ups when keys
are released out of sequence.

Updated tons of documentation.

Currently QwertyKeyListener is being used for full keyboards
with autotext and capitalization disabled.  This mostly works
but causes some problems with character pickers, etc.
These issues will be resolved in subsequent changes.

Change-Id: Ica48f6097a551141c215bc0d2c6f7b3fb634d354
2010-11-18 09:49:03 -08:00

617 lines
14 KiB
C++

/*
* Copyright (C) 2005 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <utils/String8.h>
#include <utils/Log.h>
#include <utils/Unicode.h>
#include <utils/SharedBuffer.h>
#include <utils/String16.h>
#include <utils/TextOutput.h>
#include <utils/threads.h>
#include <private/utils/Static.h>
#include <ctype.h>
/*
* Functions outside android is below the namespace android, since they use
* functions and constants in android namespace.
*/
// ---------------------------------------------------------------------------
namespace android {
// Separator used by resource paths. This is not platform dependent contrary
// to OS_PATH_SEPARATOR.
#define RES_PATH_SEPARATOR '/'
static SharedBuffer* gEmptyStringBuf = NULL;
static char* gEmptyString = NULL;
extern int gDarwinCantLoadAllObjects;
int gDarwinIsReallyAnnoying;
static inline char* getEmptyString()
{
gEmptyStringBuf->acquire();
return gEmptyString;
}
void initialize_string8()
{
// HACK: This dummy dependency forces linking libutils Static.cpp,
// which is needed to initialize String8/String16 classes.
// These variables are named for Darwin, but are needed elsewhere too,
// including static linking on any platform.
gDarwinIsReallyAnnoying = gDarwinCantLoadAllObjects;
SharedBuffer* buf = SharedBuffer::alloc(1);
char* str = (char*)buf->data();
*str = 0;
gEmptyStringBuf = buf;
gEmptyString = str;
}
void terminate_string8()
{
SharedBuffer::bufferFromData(gEmptyString)->release();
gEmptyStringBuf = NULL;
gEmptyString = NULL;
}
// ---------------------------------------------------------------------------
static char* allocFromUTF8(const char* in, size_t len)
{
if (len > 0) {
SharedBuffer* buf = SharedBuffer::alloc(len+1);
LOG_ASSERT(buf, "Unable to allocate shared buffer");
if (buf) {
char* str = (char*)buf->data();
memcpy(str, in, len);
str[len] = 0;
return str;
}
return NULL;
}
return getEmptyString();
}
static char* allocFromUTF16(const char16_t* in, size_t len)
{
if (len == 0) return getEmptyString();
const ssize_t bytes = utf16_to_utf8_length(in, len);
if (bytes < 0) {
return getEmptyString();
}
SharedBuffer* buf = SharedBuffer::alloc(bytes+1);
LOG_ASSERT(buf, "Unable to allocate shared buffer");
if (!buf) {
return getEmptyString();
}
char* str = (char*)buf->data();
utf16_to_utf8(in, len, str);
return str;
}
static char* allocFromUTF32(const char32_t* in, size_t len)
{
if (len == 0) {
return getEmptyString();
}
const ssize_t bytes = utf32_to_utf8_length(in, len);
if (bytes < 0) {
return getEmptyString();
}
SharedBuffer* buf = SharedBuffer::alloc(bytes+1);
LOG_ASSERT(buf, "Unable to allocate shared buffer");
if (!buf) {
return getEmptyString();
}
char* str = (char*) buf->data();
utf32_to_utf8(in, len, str);
return str;
}
// ---------------------------------------------------------------------------
String8::String8()
: mString(getEmptyString())
{
}
String8::String8(const String8& o)
: mString(o.mString)
{
SharedBuffer::bufferFromData(mString)->acquire();
}
String8::String8(const char* o)
: mString(allocFromUTF8(o, strlen(o)))
{
if (mString == NULL) {
mString = getEmptyString();
}
}
String8::String8(const char* o, size_t len)
: mString(allocFromUTF8(o, len))
{
if (mString == NULL) {
mString = getEmptyString();
}
}
String8::String8(const String16& o)
: mString(allocFromUTF16(o.string(), o.size()))
{
}
String8::String8(const char16_t* o)
: mString(allocFromUTF16(o, strlen16(o)))
{
}
String8::String8(const char16_t* o, size_t len)
: mString(allocFromUTF16(o, len))
{
}
String8::String8(const char32_t* o)
: mString(allocFromUTF32(o, strlen32(o)))
{
}
String8::String8(const char32_t* o, size_t len)
: mString(allocFromUTF32(o, len))
{
}
String8::~String8()
{
SharedBuffer::bufferFromData(mString)->release();
}
void String8::clear() {
SharedBuffer::bufferFromData(mString)->release();
mString = getEmptyString();
}
void String8::setTo(const String8& other)
{
SharedBuffer::bufferFromData(other.mString)->acquire();
SharedBuffer::bufferFromData(mString)->release();
mString = other.mString;
}
status_t String8::setTo(const char* other)
{
const char *newString = allocFromUTF8(other, strlen(other));
SharedBuffer::bufferFromData(mString)->release();
mString = newString;
if (mString) return NO_ERROR;
mString = getEmptyString();
return NO_MEMORY;
}
status_t String8::setTo(const char* other, size_t len)
{
const char *newString = allocFromUTF8(other, len);
SharedBuffer::bufferFromData(mString)->release();
mString = newString;
if (mString) return NO_ERROR;
mString = getEmptyString();
return NO_MEMORY;
}
status_t String8::setTo(const char16_t* other, size_t len)
{
const char *newString = allocFromUTF16(other, len);
SharedBuffer::bufferFromData(mString)->release();
mString = newString;
if (mString) return NO_ERROR;
mString = getEmptyString();
return NO_MEMORY;
}
status_t String8::setTo(const char32_t* other, size_t len)
{
const char *newString = allocFromUTF32(other, len);
SharedBuffer::bufferFromData(mString)->release();
mString = newString;
if (mString) return NO_ERROR;
mString = getEmptyString();
return NO_MEMORY;
}
status_t String8::append(const String8& other)
{
const size_t otherLen = other.bytes();
if (bytes() == 0) {
setTo(other);
return NO_ERROR;
} else if (otherLen == 0) {
return NO_ERROR;
}
return real_append(other.string(), otherLen);
}
status_t String8::append(const char* other)
{
return append(other, strlen(other));
}
status_t String8::append(const char* other, size_t otherLen)
{
if (bytes() == 0) {
return setTo(other, otherLen);
} else if (otherLen == 0) {
return NO_ERROR;
}
return real_append(other, otherLen);
}
status_t String8::appendFormat(const char* fmt, ...)
{
va_list args;
va_start(args, fmt);
status_t result = appendFormatV(fmt, args);
va_end(args);
return result;
}
status_t String8::appendFormatV(const char* fmt, va_list args)
{
int result = NO_ERROR;
int n = vsnprintf(NULL, 0, fmt, args);
if (n != 0) {
size_t oldLength = length();
char* buf = lockBuffer(oldLength + n);
if (buf) {
vsnprintf(buf + oldLength, n + 1, fmt, args);
} else {
result = NO_MEMORY;
}
}
return result;
}
status_t String8::real_append(const char* other, size_t otherLen)
{
const size_t myLen = bytes();
SharedBuffer* buf = SharedBuffer::bufferFromData(mString)
->editResize(myLen+otherLen+1);
if (buf) {
char* str = (char*)buf->data();
mString = str;
str += myLen;
memcpy(str, other, otherLen);
str[otherLen] = '\0';
return NO_ERROR;
}
return NO_MEMORY;
}
char* String8::lockBuffer(size_t size)
{
SharedBuffer* buf = SharedBuffer::bufferFromData(mString)
->editResize(size+1);
if (buf) {
char* str = (char*)buf->data();
mString = str;
return str;
}
return NULL;
}
void String8::unlockBuffer()
{
unlockBuffer(strlen(mString));
}
status_t String8::unlockBuffer(size_t size)
{
if (size != this->size()) {
SharedBuffer* buf = SharedBuffer::bufferFromData(mString)
->editResize(size+1);
if (! buf) {
return NO_MEMORY;
}
char* str = (char*)buf->data();
str[size] = 0;
mString = str;
}
return NO_ERROR;
}
ssize_t String8::find(const char* other, size_t start) const
{
size_t len = size();
if (start >= len) {
return -1;
}
const char* s = mString+start;
const char* p = strstr(s, other);
return p ? p-mString : -1;
}
void String8::toLower()
{
toLower(0, size());
}
void String8::toLower(size_t start, size_t length)
{
const size_t len = size();
if (start >= len) {
return;
}
if (start+length > len) {
length = len-start;
}
char* buf = lockBuffer(len);
buf += start;
while (length > 0) {
*buf = tolower(*buf);
buf++;
length--;
}
unlockBuffer(len);
}
void String8::toUpper()
{
toUpper(0, size());
}
void String8::toUpper(size_t start, size_t length)
{
const size_t len = size();
if (start >= len) {
return;
}
if (start+length > len) {
length = len-start;
}
char* buf = lockBuffer(len);
buf += start;
while (length > 0) {
*buf = toupper(*buf);
buf++;
length--;
}
unlockBuffer(len);
}
size_t String8::getUtf32Length() const
{
return utf8_to_utf32_length(mString, length());
}
int32_t String8::getUtf32At(size_t index, size_t *next_index) const
{
return utf32_from_utf8_at(mString, length(), index, next_index);
}
void String8::getUtf32(char32_t* dst) const
{
utf8_to_utf32(mString, length(), dst);
}
TextOutput& operator<<(TextOutput& to, const String8& val)
{
to << val.string();
return to;
}
// ---------------------------------------------------------------------------
// Path functions
void String8::setPathName(const char* name)
{
setPathName(name, strlen(name));
}
void String8::setPathName(const char* name, size_t len)
{
char* buf = lockBuffer(len);
memcpy(buf, name, len);
// remove trailing path separator, if present
if (len > 0 && buf[len-1] == OS_PATH_SEPARATOR)
len--;
buf[len] = '\0';
unlockBuffer(len);
}
String8 String8::getPathLeaf(void) const
{
const char* cp;
const char*const buf = mString;
cp = strrchr(buf, OS_PATH_SEPARATOR);
if (cp == NULL)
return String8(*this);
else
return String8(cp+1);
}
String8 String8::getPathDir(void) const
{
const char* cp;
const char*const str = mString;
cp = strrchr(str, OS_PATH_SEPARATOR);
if (cp == NULL)
return String8("");
else
return String8(str, cp - str);
}
String8 String8::walkPath(String8* outRemains) const
{
const char* cp;
const char*const str = mString;
const char* buf = str;
cp = strchr(buf, OS_PATH_SEPARATOR);
if (cp == buf) {
// don't include a leading '/'.
buf = buf+1;
cp = strchr(buf, OS_PATH_SEPARATOR);
}
if (cp == NULL) {
String8 res = buf != str ? String8(buf) : *this;
if (outRemains) *outRemains = String8("");
return res;
}
String8 res(buf, cp-buf);
if (outRemains) *outRemains = String8(cp+1);
return res;
}
/*
* Helper function for finding the start of an extension in a pathname.
*
* Returns a pointer inside mString, or NULL if no extension was found.
*/
char* String8::find_extension(void) const
{
const char* lastSlash;
const char* lastDot;
int extLen;
const char* const str = mString;
// only look at the filename
lastSlash = strrchr(str, OS_PATH_SEPARATOR);
if (lastSlash == NULL)
lastSlash = str;
else
lastSlash++;
// find the last dot
lastDot = strrchr(lastSlash, '.');
if (lastDot == NULL)
return NULL;
// looks good, ship it
return const_cast<char*>(lastDot);
}
String8 String8::getPathExtension(void) const
{
char* ext;
ext = find_extension();
if (ext != NULL)
return String8(ext);
else
return String8("");
}
String8 String8::getBasePath(void) const
{
char* ext;
const char* const str = mString;
ext = find_extension();
if (ext == NULL)
return String8(*this);
else
return String8(str, ext - str);
}
String8& String8::appendPath(const char* name)
{
// TODO: The test below will fail for Win32 paths. Fix later or ignore.
if (name[0] != OS_PATH_SEPARATOR) {
if (*name == '\0') {
// nothing to do
return *this;
}
size_t len = length();
if (len == 0) {
// no existing filename, just use the new one
setPathName(name);
return *this;
}
// make room for oldPath + '/' + newPath
int newlen = strlen(name);
char* buf = lockBuffer(len+1+newlen);
// insert a '/' if needed
if (buf[len-1] != OS_PATH_SEPARATOR)
buf[len++] = OS_PATH_SEPARATOR;
memcpy(buf+len, name, newlen+1);
len += newlen;
unlockBuffer(len);
return *this;
} else {
setPathName(name);
return *this;
}
}
String8& String8::convertToResPath()
{
#if OS_PATH_SEPARATOR != RES_PATH_SEPARATOR
size_t len = length();
if (len > 0) {
char * buf = lockBuffer(len);
for (char * end = buf + len; buf < end; ++buf) {
if (*buf == OS_PATH_SEPARATOR)
*buf = RES_PATH_SEPARATOR;
}
unlockBuffer(len);
}
#endif
return *this;
}
}; // namespace android