diff options
author | lloyd <[email protected]> | 2011-05-12 14:47:53 +0000 |
---|---|---|
committer | lloyd <[email protected]> | 2011-05-12 14:47:53 +0000 |
commit | 44aa55d0aeb96d452450df5c024db7275a560925 (patch) | |
tree | aadfe42a7f9bd2c546a8d3c47fb5a1b639baf43d /src/codec | |
parent | 6b97d4b061b07f036dc3f823d6cd082fa67a71bb (diff) |
Add base64_decode. Seems to work OK, though incremental decoding is
not well tested.
Diffstat (limited to 'src/codec')
-rw-r--r-- | src/codec/base64/base64.cpp | 151 | ||||
-rw-r--r-- | src/codec/base64/base64.h | 70 |
2 files changed, 221 insertions, 0 deletions
diff --git a/src/codec/base64/base64.cpp b/src/codec/base64/base64.cpp index d6c5ec7c0..27e6853bd 100644 --- a/src/codec/base64/base64.cpp +++ b/src/codec/base64/base64.cpp @@ -97,5 +97,156 @@ std::string base64_encode(const MemoryRegion<byte>& input) return base64_encode(&input[0], input.size()); } +size_t base64_decode(byte output[], + const char input[], + size_t input_length, + size_t& input_consumed, + bool final_inputs, + bool ignore_ws) + { + /* + * Base64 Decoder Lookup Table + * Warning: assumes ASCII encodings + */ + static const byte BASE64_TO_BIN[256] = { + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x80, + 0x80, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0x80, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0x3E, 0xFF, 0xFF, 0xFF, 0x3F, 0x34, 0x35, + 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B, 0x3C, 0x3D, 0xFF, 0xFF, + 0xFF, 0x81, 0xFF, 0xFF, 0xFF, 0x00, 0x01, 0x02, 0x03, 0x04, + 0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, + 0x0F, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, + 0x19, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x1A, 0x1B, 0x1C, + 0x1D, 0x1E, 0x1F, 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, + 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F, 0x30, + 0x31, 0x32, 0x33, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, + 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }; + + byte* out_ptr = output; + byte decode_buf[4]; + size_t decode_buf_pos = 0; + size_t final_truncate = 0; + bool seen_pad = false; + + clear_mem(output, input_length * 3 / 4); + + for(size_t i = 0; i != input_length; ++i) + { + const byte bin = BASE64_TO_BIN[static_cast<byte>(input[i])]; + + if(seen_pad && bin != 0x81) + throw std::invalid_argument("base64_decode: invalid padding"); + + if(bin <= 0x3F) + { + decode_buf[decode_buf_pos] = bin; + decode_buf_pos += 1; + } + else if(!(final_inputs && bin == 0x81)) + { + if(bin == 0x80 && ignore_ws) + continue; + + std::string bad_char(1, input[i]); + if(bad_char == "\t") + bad_char = "\\t"; + else if(bad_char == "\n") + bad_char = "\\n"; + + throw std::invalid_argument( + std::string("base64_decode: invalid base64 character '") + + bad_char + "'"); + } + + /* + * If we either see a pad character, or we are the the end + * of the input + */ + if(final_inputs && (bin == 0x81 || i == input_length - 1)) + { + seen_pad = true; + if(decode_buf_pos) + { + for(size_t i = decode_buf_pos; i != 4; ++i) + decode_buf[i] = 0; + final_truncate = (4 - decode_buf_pos); + decode_buf_pos = 4; + } + } + + if(decode_buf_pos == 4) + { + out_ptr[0] = (decode_buf[0] << 2) | (decode_buf[1] >> 4); + out_ptr[1] = (decode_buf[1] << 4) | (decode_buf[2] >> 2); + out_ptr[2] = (decode_buf[2] << 6) | decode_buf[3]; + + out_ptr += 3; + decode_buf_pos = 0; + } + } + + input_consumed = input_length - decode_buf_pos; + size_t written = (out_ptr - output) - final_truncate; + + return written; + } + +size_t base64_decode(byte output[], + const char input[], + size_t input_length, + bool ignore_ws) + { + size_t consumed = 0; + size_t written = base64_decode(output, input, input_length, + consumed, true, ignore_ws); + + if(consumed != input_length) + throw std::invalid_argument("base64_decode: input did not have full bytes"); + + return written; + } + +size_t base64_decode(byte output[], + const std::string& input, + bool ignore_ws) + { + return base64_decode(output, &input[0], input.length(), ignore_ws); + } + +SecureVector<byte> base64_decode(const char input[], + size_t input_length, + bool ignore_ws) + { + SecureVector<byte> bin(1 + (input_length * 3) / 4); + + size_t written = base64_decode(&bin[0], + input, + input_length, + ignore_ws); + + bin.resize(written); + return bin; + } + +SecureVector<byte> base64_decode(const std::string& input, + bool ignore_ws) + { + return base64_decode(&input[0], input.size(), ignore_ws); + } + } diff --git a/src/codec/base64/base64.h b/src/codec/base64/base64.h index 6daac73d8..23a2558bd 100644 --- a/src/codec/base64/base64.h +++ b/src/codec/base64/base64.h @@ -48,6 +48,76 @@ std::string BOTAN_DLL base64_encode(const byte input[], */ std::string BOTAN_DLL base64_encode(const MemoryRegion<byte>& input); +/** +* Perform base64 decoding +* @param output an array of at least input_length*3/4 bytes +* @param input some base64 input +* @param input_length length of input in bytes +* @param input_consumed is an output parameter which says how many +* bytes of input were actually consumed. If less than +* input_length, then the range input[consumed:length] +* should be passed in later along with more input. +* @param final_inputs true iff this is the last input, in which case + padding is allowed +* @param ignore_ws ignore whitespace on input; if false, throw an + exception if whitespace is encountered +* @return number of bytes written to output +*/ +size_t BOTAN_DLL base64_decode(byte output[], + const char input[], + size_t input_length, + size_t& input_consumed, + bool final_inputs, + bool ignore_ws = true); + +/** +* Perform base64 decoding +* @param output an array of at least input_length*3/4 bytes +* @param input some base64 input +* @param input_length length of input in bytes +* @param ignore_ws ignore whitespace on input; if false, throw an + exception if whitespace is encountered +* @return number of bytes written to output +*/ +size_t BOTAN_DLL base64_decode(byte output[], + const char input[], + size_t input_length, + bool ignore_ws = true); + +/** +* Perform base64 decoding +* @param output an array of at least input_length/3*4 bytes +* @param input some base64 input +* @param ignore_ws ignore whitespace on input; if false, throw an + exception if whitespace is encountered +* @return number of bytes written to output +*/ +size_t BOTAN_DLL base64_decode(byte output[], + const std::string& input, + bool ignore_ws = true); + +/** +* Perform base64 decoding +* @param input some base64 input +* @param input_length the length of input in bytes +* @param ignore_ws ignore whitespace on input; if false, throw an + exception if whitespace is encountered +* @return decoded base64 output +*/ +SecureVector<byte> BOTAN_DLL base64_decode(const char input[], + size_t input_length, + bool ignore_ws = true); + +/** +* Perform base64 decoding +* @param input some base64 input +* @param ignore_ws ignore whitespace on input; if false, throw an + exception if whitespace is encountered +* @return decoded base64 output +*/ +SecureVector<byte> BOTAN_DLL base64_decode(const std::string& input, + bool ignore_ws = true); + } #endif |