why first_bytes plus one in path_trie.cpp

[zh794390558]

STT/native_client/ctcdecode/path_trie.cpp

Lines 129 to 141 in f2e9c85

	int PathTrie::distance_to_codepoint_boundary(unsigned char *first_byte,
	const Alphabet& alphabet)
	{
	if (byte_is_codepoint_boundary(alphabet.DecodeSingle(character)[0])) {
	*first_byte = (unsigned char)character + 1;
	return 1;
	}
	if (parent != nullptr && parent->character != ROOT_) {
	return 1 + parent->distance_to_codepoint_boundary(first_byte, alphabet);
	}
	assert(false); // unreachable
	return 0;
	}

[reuben]

In bytes output mode, label values 0-254 correspond to byte values 1-255. Label value 255 is the CTC blank symbol. The +1 here is hardcoding that assumption into the decoder code. This should definitely have a comment explaining it, so thanks for raising.

See also https://stt.readthedocs.io/en/latest/Decoder.html#bytes-output-mode