diff options
author | Feng Xiao <xfxyjwf@gmail.com> | 2015-06-15 18:21:48 -0700 |
---|---|---|
committer | Feng Xiao <xfxyjwf@gmail.com> | 2015-06-15 18:24:01 -0700 |
commit | e96ff30120a3834f7d1e31e43e591bf7cfbd731f (patch) | |
tree | 7e4c317912ac3809fbff11bb692084debd4a27eb /src/google/protobuf/stubs/strutil.cc | |
parent | 4cbb612299eaf8259db25f5dcda46022fbd1db54 (diff) | |
download | protobuf-e96ff30120a3834f7d1e31e43e591bf7cfbd731f.tar.gz protobuf-e96ff30120a3834f7d1e31e43e591bf7cfbd731f.tar.bz2 protobuf-e96ff30120a3834f7d1e31e43e591bf7cfbd731f.zip |
Down-integrate from internal code base.
[ci skip]
Change-Id: I9391c09640e0b0b2b21c45a97a1fc91814d95c5d
Diffstat (limited to 'src/google/protobuf/stubs/strutil.cc')
-rw-r--r-- | src/google/protobuf/stubs/strutil.cc | 768 |
1 files changed, 746 insertions, 22 deletions
diff --git a/src/google/protobuf/stubs/strutil.cc b/src/google/protobuf/stubs/strutil.cc index 7ecc17ee..99e8bf1d 100644 --- a/src/google/protobuf/stubs/strutil.cc +++ b/src/google/protobuf/stubs/strutil.cc @@ -31,6 +31,7 @@ // from google3/strings/strutil.cc #include <google/protobuf/stubs/strutil.h> + #include <errno.h> #include <float.h> // FLT_DIG and DBL_DIG #include <limits> @@ -38,6 +39,8 @@ #include <stdio.h> #include <iterator> +#include <google/protobuf/stubs/stl_util.h> + #ifdef _WIN32 // MSVC has only _snprintf, not snprintf. // @@ -309,17 +312,6 @@ void JoinStrings(const vector<string>& components, #define IS_OCTAL_DIGIT(c) (((c) >= '0') && ((c) <= '7')) -inline int hex_digit_to_int(char c) { - /* Assume ASCII. */ - assert('0' == 0x30 && 'A' == 0x41 && 'a' == 0x61); - assert(isxdigit(c)); - int x = static_cast<unsigned char>(c); - if (x > '9') { - x += 9; - } - return x & 0xf; -} - // Protocol buffers doesn't ever care about errors, but I don't want to remove // the code. #define LOG_STRING(LEVEL, VECTOR) GOOGLE_LOG_IF(LEVEL, false) @@ -652,14 +644,15 @@ inline bool safe_parse_sign(string* text /*inout*/, return true; } -inline bool safe_parse_positive_int( - string text, int32* value_p) { +template<typename IntType> +bool safe_parse_positive_int( + string text, IntType* value_p) { int base = 10; - int32 value = 0; - const int32 vmax = std::numeric_limits<int32>::max(); + IntType value = 0; + const IntType vmax = std::numeric_limits<IntType>::max(); assert(vmax > 0); assert(vmax >= base); - const int32 vmax_over_base = vmax / base; + const IntType vmax_over_base = vmax / base; const char* start = text.data(); const char* end = start + text.size(); // loop over digits @@ -685,14 +678,15 @@ inline bool safe_parse_positive_int( return true; } -inline bool safe_parse_negative_int( - string text, int32* value_p) { +template<typename IntType> +bool safe_parse_negative_int( + const string& text, IntType* value_p) { int base = 10; - int32 value = 0; - const int32 vmin = std::numeric_limits<int32>::min(); + IntType value = 0; + const IntType vmin = std::numeric_limits<IntType>::min(); assert(vmin < 0); assert(vmin <= 0 - base); - int32 vmin_over_base = vmin / base; + IntType vmin_over_base = vmin / base; // 2003 c++ standard [expr.mul] // "... the sign of the remainder is implementation-defined." // Although (vmin/base)*base + vmin%base is always vmin. @@ -725,7 +719,8 @@ inline bool safe_parse_negative_int( return true; } -bool safe_int(string text, int32* value_p) { +template<typename IntType> +bool safe_int_internal(string text, IntType* value_p) { *value_p = 0; bool negative; if (!safe_parse_sign(&text, &negative)) { @@ -738,6 +733,16 @@ bool safe_int(string text, int32* value_p) { } } +template<typename IntType> +bool safe_uint_internal(string text, IntType* value_p) { + *value_p = 0; + bool negative; + if (!safe_parse_sign(&text, &negative) || negative) { + return false; + } + return safe_parse_positive_int(text, value_p); +} + // ---------------------------------------------------------------------- // FastIntToBuffer() // FastInt64ToBuffer() @@ -1236,6 +1241,41 @@ char* DoubleToBuffer(double value, char* buffer) { return buffer; } +static int memcasecmp(const char *s1, const char *s2, size_t len) { + const unsigned char *us1 = reinterpret_cast<const unsigned char *>(s1); + const unsigned char *us2 = reinterpret_cast<const unsigned char *>(s2); + + for ( int i = 0; i < len; i++ ) { + const int diff = + static_cast<int>(static_cast<unsigned char>(ascii_tolower(us1[i]))) - + static_cast<int>(static_cast<unsigned char>(ascii_tolower(us2[i]))); + if (diff != 0) return diff; + } + return 0; +} + +inline bool CaseEqual(StringPiece s1, StringPiece s2) { + if (s1.size() != s2.size()) return false; + return memcasecmp(s1.data(), s2.data(), s1.size()) == 0; +} + +bool safe_strtob(StringPiece str, bool* value) { + GOOGLE_CHECK(value != NULL) << "NULL output boolean given."; + if (CaseEqual(str, "true") || CaseEqual(str, "t") || + CaseEqual(str, "yes") || CaseEqual(str, "y") || + CaseEqual(str, "1")) { + *value = true; + return true; + } + if (CaseEqual(str, "false") || CaseEqual(str, "f") || + CaseEqual(str, "no") || CaseEqual(str, "n") || + CaseEqual(str, "0")) { + *value = false; + return true; + } + return false; +} + bool safe_strtof(const char* str, float* value) { char* endptr; errno = 0; // errno only gets set on errors @@ -1247,6 +1287,34 @@ bool safe_strtof(const char* str, float* value) { return *str != 0 && *endptr == 0 && errno == 0; } +bool safe_strtod(const char* str, double* value) { + char* endptr; + *value = strtod(str, &endptr); + if (endptr != str) { + while (ascii_isspace(*endptr)) ++endptr; + } + // Ignore range errors from strtod. The values it + // returns on underflow and overflow are the right + // fallback in a robust setting. + return *str != '\0' && *endptr == '\0'; +} + +bool safe_strto32(const string& str, int32* value) { + return safe_int_internal(str, value); +} + +bool safe_strtou32(const string& str, uint32* value) { + return safe_uint_internal(str, value); +} + +bool safe_strto64(const string& str, int64* value) { + return safe_int_internal(str, value); +} + +bool safe_strtou64(const string& str, uint64* value) { + return safe_uint_internal(str, value); +} + char* FloatToBuffer(float value, char* buffer) { // FLT_DIG is 6 for IEEE-754 floats, which are used on almost all // platforms these days. Just in case some system exists where FLT_DIG @@ -1518,5 +1586,661 @@ int GlobalReplaceSubstring(const string& substring, return num_replacements; } +int CalculateBase64EscapedLen(int input_len, bool do_padding) { + // Base64 encodes three bytes of input at a time. If the input is not + // divisible by three, we pad as appropriate. + // + // (from http://tools.ietf.org/html/rfc3548) + // Special processing is performed if fewer than 24 bits are available + // at the end of the data being encoded. A full encoding quantum is + // always completed at the end of a quantity. When fewer than 24 input + // bits are available in an input group, zero bits are added (on the + // right) to form an integral number of 6-bit groups. Padding at the + // end of the data is performed using the '=' character. Since all base + // 64 input is an integral number of octets, only the following cases + // can arise: + + + // Base64 encodes each three bytes of input into four bytes of output. + int len = (input_len / 3) * 4; + + if (input_len % 3 == 0) { + // (from http://tools.ietf.org/html/rfc3548) + // (1) the final quantum of encoding input is an integral multiple of 24 + // bits; here, the final unit of encoded output will be an integral + // multiple of 4 characters with no "=" padding, + } else if (input_len % 3 == 1) { + // (from http://tools.ietf.org/html/rfc3548) + // (2) the final quantum of encoding input is exactly 8 bits; here, the + // final unit of encoded output will be two characters followed by two + // "=" padding characters, or + len += 2; + if (do_padding) { + len += 2; + } + } else { // (input_len % 3 == 2) + // (from http://tools.ietf.org/html/rfc3548) + // (3) the final quantum of encoding input is exactly 16 bits; here, the + // final unit of encoded output will be three characters followed by one + // "=" padding character. + len += 3; + if (do_padding) { + len += 1; + } + } + + assert(len >= input_len); // make sure we didn't overflow + return len; +} + +// Base64Escape does padding, so this calculation includes padding. +int CalculateBase64EscapedLen(int input_len) { + return CalculateBase64EscapedLen(input_len, true); +} + +// ---------------------------------------------------------------------- +// int Base64Unescape() - base64 decoder +// int Base64Escape() - base64 encoder +// int WebSafeBase64Unescape() - Google's variation of base64 decoder +// int WebSafeBase64Escape() - Google's variation of base64 encoder +// +// Check out +// http://tools.ietf.org/html/rfc2045 for formal description, but what we +// care about is that... +// Take the encoded stuff in groups of 4 characters and turn each +// character into a code 0 to 63 thus: +// A-Z map to 0 to 25 +// a-z map to 26 to 51 +// 0-9 map to 52 to 61 +// +(- for WebSafe) maps to 62 +// /(_ for WebSafe) maps to 63 +// There will be four numbers, all less than 64 which can be represented +// by a 6 digit binary number (aaaaaa, bbbbbb, cccccc, dddddd respectively). +// Arrange the 6 digit binary numbers into three bytes as such: +// aaaaaabb bbbbcccc ccdddddd +// Equals signs (one or two) are used at the end of the encoded block to +// indicate that the text was not an integer multiple of three bytes long. +// ---------------------------------------------------------------------- + +int Base64UnescapeInternal(const char *src_param, int szsrc, + char *dest, int szdest, + const signed char* unbase64) { + static const char kPad64Equals = '='; + static const char kPad64Dot = '.'; + + int decode = 0; + int destidx = 0; + int state = 0; + unsigned int ch = 0; + unsigned int temp = 0; + + // If "char" is signed by default, using *src as an array index results in + // accessing negative array elements. Treat the input as a pointer to + // unsigned char to avoid this. + const unsigned char *src = reinterpret_cast<const unsigned char*>(src_param); + + // The GET_INPUT macro gets the next input character, skipping + // over any whitespace, and stopping when we reach the end of the + // string or when we read any non-data character. The arguments are + // an arbitrary identifier (used as a label for goto) and the number + // of data bytes that must remain in the input to avoid aborting the + // loop. +#define GET_INPUT(label, remain) \ + label: \ + --szsrc; \ + ch = *src++; \ + decode = unbase64[ch]; \ + if (decode < 0) { \ + if (ascii_isspace(ch) && szsrc >= remain) \ + goto label; \ + state = 4 - remain; \ + break; \ + } + + // if dest is null, we're just checking to see if it's legal input + // rather than producing output. (I suspect this could just be done + // with a regexp...). We duplicate the loop so this test can be + // outside it instead of in every iteration. + + if (dest) { + // This loop consumes 4 input bytes and produces 3 output bytes + // per iteration. We can't know at the start that there is enough + // data left in the string for a full iteration, so the loop may + // break out in the middle; if so 'state' will be set to the + // number of input bytes read. + + while (szsrc >= 4) { + // We'll start by optimistically assuming that the next four + // bytes of the string (src[0..3]) are four good data bytes + // (that is, no nulls, whitespace, padding chars, or illegal + // chars). We need to test src[0..2] for nulls individually + // before constructing temp to preserve the property that we + // never read past a null in the string (no matter how long + // szsrc claims the string is). + + if (!src[0] || !src[1] || !src[2] || + (temp = ((unsigned(unbase64[src[0]]) << 18) | + (unsigned(unbase64[src[1]]) << 12) | + (unsigned(unbase64[src[2]]) << 6) | + (unsigned(unbase64[src[3]])))) & 0x80000000) { + // Iff any of those four characters was bad (null, illegal, + // whitespace, padding), then temp's high bit will be set + // (because unbase64[] is -1 for all bad characters). + // + // We'll back up and resort to the slower decoder, which knows + // how to handle those cases. + + GET_INPUT(first, 4); + temp = decode; + GET_INPUT(second, 3); + temp = (temp << 6) | decode; + GET_INPUT(third, 2); + temp = (temp << 6) | decode; + GET_INPUT(fourth, 1); + temp = (temp << 6) | decode; + } else { + // We really did have four good data bytes, so advance four + // characters in the string. + + szsrc -= 4; + src += 4; + decode = -1; + ch = '\0'; + } + + // temp has 24 bits of input, so write that out as three bytes. + + if (destidx+3 > szdest) return -1; + dest[destidx+2] = temp; + temp >>= 8; + dest[destidx+1] = temp; + temp >>= 8; + dest[destidx] = temp; + destidx += 3; + } + } else { + while (szsrc >= 4) { + if (!src[0] || !src[1] || !src[2] || + (temp = ((unsigned(unbase64[src[0]]) << 18) | + (unsigned(unbase64[src[1]]) << 12) | + (unsigned(unbase64[src[2]]) << 6) | + (unsigned(unbase64[src[3]])))) & 0x80000000) { + GET_INPUT(first_no_dest, 4); + GET_INPUT(second_no_dest, 3); + GET_INPUT(third_no_dest, 2); + GET_INPUT(fourth_no_dest, 1); + } else { + szsrc -= 4; + src += 4; + decode = -1; + ch = '\0'; + } + destidx += 3; + } + } + +#undef GET_INPUT + + // if the loop terminated because we read a bad character, return + // now. + if (decode < 0 && ch != '\0' && + ch != kPad64Equals && ch != kPad64Dot && !ascii_isspace(ch)) + return -1; + + if (ch == kPad64Equals || ch == kPad64Dot) { + // if we stopped by hitting an '=' or '.', un-read that character -- we'll + // look at it again when we count to check for the proper number of + // equals signs at the end. + ++szsrc; + --src; + } else { + // This loop consumes 1 input byte per iteration. It's used to + // clean up the 0-3 input bytes remaining when the first, faster + // loop finishes. 'temp' contains the data from 'state' input + // characters read by the first loop. + while (szsrc > 0) { + --szsrc; + ch = *src++; + decode = unbase64[ch]; + if (decode < 0) { + if (ascii_isspace(ch)) { + continue; + } else if (ch == '\0') { + break; + } else if (ch == kPad64Equals || ch == kPad64Dot) { + // back up one character; we'll read it again when we check + // for the correct number of pad characters at the end. + ++szsrc; + --src; + break; + } else { + return -1; + } + } + + // Each input character gives us six bits of output. + temp = (temp << 6) | decode; + ++state; + if (state == 4) { + // If we've accumulated 24 bits of output, write that out as + // three bytes. + if (dest) { + if (destidx+3 > szdest) return -1; + dest[destidx+2] = temp; + temp >>= 8; + dest[destidx+1] = temp; + temp >>= 8; + dest[destidx] = temp; + } + destidx += 3; + state = 0; + temp = 0; + } + } + } + + // Process the leftover data contained in 'temp' at the end of the input. + int expected_equals = 0; + switch (state) { + case 0: + // Nothing left over; output is a multiple of 3 bytes. + break; + + case 1: + // Bad input; we have 6 bits left over. + return -1; + + case 2: + // Produce one more output byte from the 12 input bits we have left. + if (dest) { + if (destidx+1 > szdest) return -1; + temp >>= 4; + dest[destidx] = temp; + } + ++destidx; + expected_equals = 2; + break; + + case 3: + // Produce two more output bytes from the 18 input bits we have left. + if (dest) { + if (destidx+2 > szdest) return -1; + temp >>= 2; + dest[destidx+1] = temp; + temp >>= 8; + dest[destidx] = temp; + } + destidx += 2; + expected_equals = 1; + break; + + default: + // state should have no other values at this point. + GOOGLE_LOG(FATAL) << "This can't happen; base64 decoder state = " << state; + } + + // The remainder of the string should be all whitespace, mixed with + // exactly 0 equals signs, or exactly 'expected_equals' equals + // signs. (Always accepting 0 equals signs is a google extension + // not covered in the RFC, as is accepting dot as the pad character.) + + int equals = 0; + while (szsrc > 0 && *src) { + if (*src == kPad64Equals || *src == kPad64Dot) + ++equals; + else if (!ascii_isspace(*src)) + return -1; + --szsrc; + ++src; + } + + return (equals == 0 || equals == expected_equals) ? destidx : -1; +} + +// The arrays below were generated by the following code +// #include <sys/time.h> +// #include <stdlib.h> +// #include <string.h> +// main() +// { +// static const char Base64[] = +// "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"; +// char *pos; +// int idx, i, j; +// printf(" "); +// for (i = 0; i < 255; i += 8) { +// for (j = i; j < i + 8; j++) { +// pos = strchr(Base64, j); +// if ((pos == NULL) || (j == 0)) +// idx = -1; +// else +// idx = pos - Base64; +// if (idx == -1) +// printf(" %2d, ", idx); +// else +// printf(" %2d/*%c*/,", idx, j); +// } +// printf("\n "); +// } +// } +// +// where the value of "Base64[]" was replaced by one of the base-64 conversion +// tables from the functions below. +static const signed char kUnBase64[] = { + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, 62/*+*/, -1, -1, -1, 63/*/ */, + 52/*0*/, 53/*1*/, 54/*2*/, 55/*3*/, 56/*4*/, 57/*5*/, 58/*6*/, 59/*7*/, + 60/*8*/, 61/*9*/, -1, -1, -1, -1, -1, -1, + -1, 0/*A*/, 1/*B*/, 2/*C*/, 3/*D*/, 4/*E*/, 5/*F*/, 6/*G*/, + 07/*H*/, 8/*I*/, 9/*J*/, 10/*K*/, 11/*L*/, 12/*M*/, 13/*N*/, 14/*O*/, + 15/*P*/, 16/*Q*/, 17/*R*/, 18/*S*/, 19/*T*/, 20/*U*/, 21/*V*/, 22/*W*/, + 23/*X*/, 24/*Y*/, 25/*Z*/, -1, -1, -1, -1, -1, + -1, 26/*a*/, 27/*b*/, 28/*c*/, 29/*d*/, 30/*e*/, 31/*f*/, 32/*g*/, + 33/*h*/, 34/*i*/, 35/*j*/, 36/*k*/, 37/*l*/, 38/*m*/, 39/*n*/, 40/*o*/, + 41/*p*/, 42/*q*/, 43/*r*/, 44/*s*/, 45/*t*/, 46/*u*/, 47/*v*/, 48/*w*/, + 49/*x*/, 50/*y*/, 51/*z*/, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1 +}; +static const signed char kUnWebSafeBase64[] = { + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, 62/*-*/, -1, -1, + 52/*0*/, 53/*1*/, 54/*2*/, 55/*3*/, 56/*4*/, 57/*5*/, 58/*6*/, 59/*7*/, + 60/*8*/, 61/*9*/, -1, -1, -1, -1, -1, -1, + -1, 0/*A*/, 1/*B*/, 2/*C*/, 3/*D*/, 4/*E*/, 5/*F*/, 6/*G*/, + 07/*H*/, 8/*I*/, 9/*J*/, 10/*K*/, 11/*L*/, 12/*M*/, 13/*N*/, 14/*O*/, + 15/*P*/, 16/*Q*/, 17/*R*/, 18/*S*/, 19/*T*/, 20/*U*/, 21/*V*/, 22/*W*/, + 23/*X*/, 24/*Y*/, 25/*Z*/, -1, -1, -1, -1, 63/*_*/, + -1, 26/*a*/, 27/*b*/, 28/*c*/, 29/*d*/, 30/*e*/, 31/*f*/, 32/*g*/, + 33/*h*/, 34/*i*/, 35/*j*/, 36/*k*/, 37/*l*/, 38/*m*/, 39/*n*/, 40/*o*/, + 41/*p*/, 42/*q*/, 43/*r*/, 44/*s*/, 45/*t*/, 46/*u*/, 47/*v*/, 48/*w*/, + 49/*x*/, 50/*y*/, 51/*z*/, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1 +}; + +int WebSafeBase64Unescape(const char *src, int szsrc, char *dest, int szdest) { + return Base64UnescapeInternal(src, szsrc, dest, szdest, kUnWebSafeBase64); +} + +static bool Base64UnescapeInternal(const char* src, int slen, string* dest, + const signed char* unbase64) { + // Determine the size of the output string. Base64 encodes every 3 bytes into + // 4 characters. any leftover chars are added directly for good measure. + // This is documented in the base64 RFC: http://tools.ietf.org/html/rfc3548 + const int dest_len = 3 * (slen / 4) + (slen % 4); + + dest->resize(dest_len); + + // We are getting the destination buffer by getting the beginning of the + // string and converting it into a char *. + const int len = Base64UnescapeInternal(src, slen, string_as_array(dest), + dest_len, unbase64); + if (len < 0) { + dest->clear(); + return false; + } + + // could be shorter if there was padding + GOOGLE_DCHECK_LE(len, dest_len); + dest->erase(len); + + return true; +} + +bool Base64Unescape(StringPiece src, string* dest) { + return Base64UnescapeInternal(src.data(), src.size(), dest, kUnBase64); +} + +bool WebSafeBase64Unescape(StringPiece src, string* dest) { + return Base64UnescapeInternal(src.data(), src.size(), dest, kUnWebSafeBase64); +} + +int Base64EscapeInternal(const unsigned char *src, int szsrc, + char *dest, int szdest, const char *base64, + bool do_padding) { + static const char kPad64 = '='; + + if (szsrc <= 0) return 0; + + if (szsrc * 4 > szdest * 3) return 0; + + char *cur_dest = dest; + const unsigned char *cur_src = src; + + char *limit_dest = dest + szdest; + const unsigned char *limit_src = src + szsrc; + + // Three bytes of data encodes to four characters of cyphertext. + // So we can pump through three-byte chunks atomically. + while (cur_src < limit_src - 3) { // keep going as long as we have >= 32 bits + uint32 in = BigEndian::Load32(cur_src) >> 8; + + cur_dest[0] = base64[in >> 18]; + in &= 0x3FFFF; + cur_dest[1] = base64[in >> 12]; + in &= 0xFFF; + cur_dest[2] = base64[in >> 6]; + in &= 0x3F; + cur_dest[3] = base64[in]; + + cur_dest += 4; + cur_src += 3; + } + // To save time, we didn't update szdest or szsrc in the loop. So do it now. + szdest = limit_dest - cur_dest; + szsrc = limit_src - cur_src; + + /* now deal with the tail (<=3 bytes) */ + switch (szsrc) { + case 0: + // Nothing left; nothing more to do. + break; + case 1: { + // One byte left: this encodes to two characters, and (optionally) + // two pad characters to round out the four-character cypherblock. + if ((szdest -= 2) < 0) return 0; + uint32 in = cur_src[0]; + cur_dest[0] = base64[in >> 2]; + in &= 0x3; + cur_dest[1] = base64[in << 4]; + cur_dest += 2; + if (do_padding) { + if ((szdest -= 2) < 0) return 0; + cur_dest[0] = kPad64; + cur_dest[1] = kPad64; + cur_dest += 2; + } + break; + } + case 2: { + // Two bytes left: this encodes to three characters, and (optionally) + // one pad character to round out the four-character cypherblock. + if ((szdest -= 3) < 0) return 0; + uint32 in = BigEndian::Load16(cur_src); + cur_dest[0] = base64[in >> 10]; + in &= 0x3FF; + cur_dest[1] = base64[in >> 4]; + in &= 0x00F; + cur_dest[2] = base64[in << 2]; + cur_dest += 3; + if (do_padding) { + if ((szdest -= 1) < 0) return 0; + cur_dest[0] = kPad64; + cur_dest += 1; + } + break; + } + case 3: { + // Three bytes left: same as in the big loop above. We can't do this in + // the loop because the loop above always reads 4 bytes, and the fourth + // byte is past the end of the input. + if ((szdest -= 4) < 0) return 0; + uint32 in = (cur_src[0] << 16) + BigEndian::Load16(cur_src + 1); + cur_dest[0] = base64[in >> 18]; + in &= 0x3FFFF; + cur_dest[1] = base64[in >> 12]; + in &= 0xFFF; + cur_dest[2] = base64[in >> 6]; + in &= 0x3F; + cur_dest[3] = base64[in]; + cur_dest += 4; + break; + } + default: + // Should not be reached: blocks of 4 bytes are handled + // in the while loop before this switch statement. + GOOGLE_LOG(FATAL) << "Logic problem? szsrc = " << szsrc; + break; + } + return (cur_dest - dest); +} + +static const char kBase64Chars[] = +"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"; + +static const char kWebSafeBase64Chars[] = +"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_"; + +int Base64Escape(const unsigned char *src, int szsrc, char *dest, int szdest) { + return Base64EscapeInternal(src, szsrc, dest, szdest, kBase64Chars, true); +} +int WebSafeBase64Escape(const unsigned char *src, int szsrc, char *dest, + int szdest, bool do_padding) { + return Base64EscapeInternal(src, szsrc, dest, szdest, + kWebSafeBase64Chars, do_padding); +} + +void Base64EscapeInternal(const unsigned char* src, int szsrc, + string* dest, bool do_padding, + const char* base64_chars) { + const int calc_escaped_size = + CalculateBase64EscapedLen(szsrc, do_padding); + dest->resize(calc_escaped_size); + const int escaped_len = Base64EscapeInternal(src, szsrc, + string_as_array(dest), + dest->size(), + base64_chars, + do_padding); + GOOGLE_DCHECK_EQ(calc_escaped_size, escaped_len); + dest->erase(escaped_len); +} + +void Base64Escape(const unsigned char *src, int szsrc, + string* dest, bool do_padding) { + Base64EscapeInternal(src, szsrc, dest, do_padding, kBase64Chars); +} + +void WebSafeBase64Escape(const unsigned char *src, int szsrc, + string *dest, bool do_padding) { + Base64EscapeInternal(src, szsrc, dest, do_padding, kWebSafeBase64Chars); +} + +void Base64Escape(StringPiece src, string* dest) { + Base64Escape(reinterpret_cast<const unsigned char*>(src.data()), + src.size(), dest, true); +} + +void WebSafeBase64Escape(StringPiece src, string* dest) { + WebSafeBase64Escape(reinterpret_cast<const unsigned char*>(src.data()), + src.size(), dest, false); +} + +void WebSafeBase64EscapeWithPadding(StringPiece src, string* dest) { + WebSafeBase64Escape(reinterpret_cast<const unsigned char*>(src.data()), + src.size(), dest, true); +} + +// Helper to append a Unicode code point to a string as UTF8, without bringing +// in any external dependencies. +int EncodeAsUTF8Char(uint32 code_point, char* output) { + uint32 tmp = 0; + int len = 0; + if (code_point <= 0x7f) { + tmp = code_point; + len = 1; + } else if (code_point <= 0x07ff) { + tmp = 0x0000c080 | + ((code_point & 0x07c0) << 2) | + (code_point & 0x003f); + len = 2; + } else if (code_point <= 0xffff) { + tmp = 0x00e08080 | + ((code_point & 0xf000) << 4) | + ((code_point & 0x0fc0) << 2) | + (code_point & 0x003f); + len = 3; + } else { + // UTF-16 is only defined for code points up to 0x10FFFF, and UTF-8 is + // normally only defined up to there as well. + tmp = 0xf0808080 | + ((code_point & 0x1c0000) << 6) | + ((code_point & 0x03f000) << 4) | + ((code_point & 0x000fc0) << 2) | + (code_point & 0x003f); + len = 4; + } + tmp = ghtonl(tmp); + memcpy(output, reinterpret_cast<const char*>(&tmp) + sizeof(tmp) - len, len); + return len; +} + +// Table of UTF-8 character lengths, based on first byte +static const unsigned char kUTF8LenTbl[256] = { + 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, + 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, + 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, + 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, + + 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, + 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1, + 2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2, + 3,3,3,3,3,3,3,3, 3,3,3,3,3,3,3,3, 4,4,4,4,4,4,4,4, 4,4,4,4,4,4,4,4 +}; + +// Return length of a single UTF-8 source character +int UTF8FirstLetterNumBytes(const char* src, int len) { + if (len == 0) { + return 0; + } + return kUTF8LenTbl[*reinterpret_cast<const uint8*>(src)]; +} + } // namespace protobuf } // namespace google |