aboutsummaryrefslogtreecommitdiff
path: root/src/mnemonics/electrum-words.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/mnemonics/electrum-words.cpp')
-rw-r--r--src/mnemonics/electrum-words.cpp82
1 files changed, 37 insertions, 45 deletions
diff --git a/src/mnemonics/electrum-words.cpp b/src/mnemonics/electrum-words.cpp
index b204e81cf..ffa82b21e 100644
--- a/src/mnemonics/electrum-words.cpp
+++ b/src/mnemonics/electrum-words.cpp
@@ -61,7 +61,6 @@
namespace
{
- const int seed_length = 24;
/*!
* \brief Finds the word list that contains the seed words and puts the indices
@@ -69,11 +68,11 @@ namespace
* \param seed List of words to match.
* \param has_checksum If word list passed checksum test, we need to only do a prefix check.
* \param matched_indices The indices where the seed words were found are added to this.
+ * \param language Language instance pointer to write to after it is found.
* \return true if all the words were present in some language false if not.
*/
bool find_seed_language(const std::vector<std::string> &seed,
- bool has_checksum, std::vector<uint32_t> &matched_indices, uint32_t &word_list_length,
- std::string &language_name)
+ bool has_checksum, std::vector<uint32_t> &matched_indices, Language::Base **language)
{
// If there's a new language added, add an instance of it here.
std::vector<Language::Base*> language_instances({
@@ -83,18 +82,6 @@ namespace
Language::Singleton<Language::Japanese>::instance(),
Language::Singleton<Language::OldEnglish>::instance()
});
- // To hold trimmed seed words in case of a checksum being present.
- std::vector<std::string> trimmed_seed;
- if (has_checksum)
- {
- // If it had a checksum, we'll just compare the unique prefix
- // So we create a list of trimmed seed words
- for (std::vector<std::string>::const_iterator it = seed.begin(); it != seed.end(); it++)
- {
- trimmed_seed.push_back(it->length() > Language::unique_prefix_length ?
- it->substr(0, Language::unique_prefix_length) : *it);
- }
- }
// Iterate through all the languages and find a match
for (std::vector<Language::Base*>::iterator it1 = language_instances.begin();
@@ -104,23 +91,22 @@ namespace
const std::unordered_map<std::string, uint32_t> &trimmed_word_map = (*it1)->get_trimmed_word_map();
// To iterate through seed words
std::vector<std::string>::const_iterator it2;
- // To iterate through trimmed seed words
- std::vector<std::string>::iterator it3;
bool full_match = true;
+ std::string trimmed_word;
// Iterate through all the words and see if they're all present
- for (it2 = seed.begin(), it3 = trimmed_seed.begin();
- it2 != seed.end(); it2++, it3++)
+ for (it2 = seed.begin(); it2 != seed.end(); it2++)
{
if (has_checksum)
{
+ trimmed_word = it2->substr(0, (*it1)->get_unique_prefix_length());
// Use the trimmed words and map
- if (trimmed_word_map.count(*it3) == 0)
+ if (trimmed_word_map.count(trimmed_word) == 0)
{
full_match = false;
break;
}
- matched_indices.push_back(trimmed_word_map.at(*it3));
+ matched_indices.push_back(trimmed_word_map.at(trimmed_word));
}
else
{
@@ -134,8 +120,7 @@ namespace
}
if (full_match)
{
- word_list_length = (*it1)->get_word_list().size();
- language_name = (*it1)->get_language_name();
+ *language = *it1;
return true;
}
// Some didn't match. Clear the index array.
@@ -146,18 +131,20 @@ namespace
/*!
* \brief Creates a checksum index in the word list array on the list of words.
- * \param word_list Vector of words
- * \return Checksum index
+ * \param word_list Vector of words
+ * \param unique_prefix_length the prefix length of each word to use for checksum
+ * \return Checksum index
*/
- uint32_t create_checksum_index(const std::vector<std::string> &word_list)
+ uint32_t create_checksum_index(const std::vector<std::string> &word_list,
+ uint32_t unique_prefix_length)
{
std::string trimmed_words = "";
for (std::vector<std::string>::const_iterator it = word_list.begin(); it != word_list.end(); it++)
{
- if (it->length() > 4)
+ if (it->length() > unique_prefix_length)
{
- trimmed_words += it->substr(0, Language::unique_prefix_length);
+ trimmed_words += it->substr(0, unique_prefix_length);
}
else
{
@@ -166,25 +153,26 @@ namespace
}
boost::crc_32_type result;
result.process_bytes(trimmed_words.data(), trimmed_words.length());
- return result.checksum() % seed_length;
+ return result.checksum() % crypto::ElectrumWords::seed_length;
}
/*!
* \brief Does the checksum test on the seed passed.
- * \param seed Vector of seed words
- * \return True if the test passed false if not.
+ * \param seed Vector of seed words
+ * \param unique_prefix_length the prefix length of each word to use for checksum
+ * \return True if the test passed false if not.
*/
- bool checksum_test(std::vector<std::string> seed)
+ bool checksum_test(std::vector<std::string> seed, uint32_t unique_prefix_length)
{
// The last word is the checksum.
std::string last_word = seed.back();
seed.pop_back();
- std::string checksum = seed[create_checksum_index(seed)];
+ std::string checksum = seed[create_checksum_index(seed, unique_prefix_length)];
- std::string trimmed_checksum = checksum.length() > 4 ? checksum.substr(0, Language::unique_prefix_length) :
+ std::string trimmed_checksum = checksum.length() > unique_prefix_length ? checksum.substr(0, unique_prefix_length) :
checksum;
- std::string trimmed_last_word = checksum.length() > 4 ? last_word.substr(0, Language::unique_prefix_length) :
+ std::string trimmed_last_word = last_word.length() > unique_prefix_length ? last_word.substr(0, unique_prefix_length) :
last_word;
return trimmed_checksum == trimmed_last_word;
}
@@ -211,11 +199,12 @@ namespace crypto
* \param language_name Language of the seed as found gets written here.
* \return false if not a multiple of 3 words, or if word is not in the words list
*/
- bool words_to_bytes(const std::string& words, crypto::secret_key& dst,
+ bool words_to_bytes(std::string words, crypto::secret_key& dst,
std::string &language_name)
{
std::vector<std::string> seed;
+ boost::algorithm::trim(words);
boost::split(seed, words, boost::is_any_of(" "));
// error on non-compliant word list
@@ -227,22 +216,25 @@ namespace crypto
// If it is seed with a checksum.
bool has_checksum = seed.size() == (seed_length + 1);
+
+ std::vector<uint32_t> matched_indices;
+ Language::Base *language;
+ if (!find_seed_language(seed, has_checksum, matched_indices, &language))
+ {
+ return false;
+ }
+ language_name = language->get_language_name();
+ uint32_t word_list_length = language->get_word_list().size();
+
if (has_checksum)
{
- if (!checksum_test(seed))
+ if (!checksum_test(seed, language->get_unique_prefix_length()))
{
// Checksum fail
return false;
}
seed.pop_back();
}
-
- std::vector<uint32_t> matched_indices;
- uint32_t word_list_length = 0;
- if (!find_seed_language(seed, has_checksum, matched_indices, word_list_length, language_name))
- {
- return false;
- }
for (unsigned int i=0; i < seed.size() / 3; i++)
{
@@ -335,7 +327,7 @@ namespace crypto
}
words.pop_back();
- words += (' ' + words_store[create_checksum_index(words_store)]);
+ words += (' ' + words_store[create_checksum_index(words_store, language->get_unique_prefix_length())]);
return false;
}