Kana mnemonic hashes
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Avril 70abb4303a
better docs
3 months ago
cli update makefile 5 months ago
contrib/kana-hash Common Lisp CFFI interface 12 months ago
include khash_max_length; describe algorithm in README 12 months ago
node move to ffi-napi 12 months ago
src better docs 3 months ago
.gitignore rework makefile 7 months ago
Cargo.toml better docs 3 months ago
LICENSE Create LICENSE 12 months ago
Makefile improve 16 iter 4 months ago
README.org small readme update 9 months ago
build.rs rework build system 7 months ago
package-binaries package binary script 12 months ago


libkhash - kana-hash

Kana mnemonic hashes

Example output

Input is uguu~ using the default salt.

Algorithm Output
SHA256 おシソまツアでぅせヅモァだゅノぴヲろヂォセづマふげぁユねハァがゅ
CRC32 わほヂァ
CRC64 づやワえほぢレご
SHA256 (truncated) おシソまツアでぅ


The dynamic library is built with Cargo and Rust, and the CLI example program is built with gcc.

Build and install

The default build configuration builds both the dynamic library, static library, and the CLI example program and copies them to /usr/local/lib/libkhash.so, /usr/local/lib/libkhash.a and /usr/local/bin/kana-hash respectively. Also installed is the C header to /usr/local/include/khash.h.

$ make && sudo make install

The install path can be changed by editing the INSTALL, INSTALL-BIN and INSTALL-INCLUDE paths in the Makefile.


To remove installed binaries, run:

$ sudo make uninstall

Other build configurations

The Makefile contains some other build directives.

Native code optimisations

By default libkhash builds the shared and static library with native architecture optimisations enabled. If you are intending to move the libraries to another architecture, this might not be desireable. To build without such optimisations, run:

$ make khash-nonative

To build and run all tests, run:

$ make test
Building the CLI

The default make directive builds both the library and the CLI example program. To build just the CLI example program, run:

$ cd cli && make

TODO Rust crate

This library is written in Rust and has a Rust library target. See Rustdocs for details

C header

A header file is provided for C programs wanting to use the khash interface. Documented more fully in ./include/khash.h. All symbols defined here begin with either KHASH_ (for macros) or khash_.


To create a context

#include <khash.h>

const char* input_salt = "salt!";
const char* input_data = "some data to hash".
khash_context ctx;
assert(khash_new_context(KHASH_ALGO_SHA256, KHASH_SALT_TYPE_SPECIFIC, input_salt, strlen(input_salt), &ctx) == KHASH_SUCCESS, "khash_new_context() failed.");

Find the buffer length we need and allocate a buffer.

size_t length;
assert(khash_length(&ctx, input_data, strlen(input_data), &length) == KHASH_SUCCESS, "khash_length() failed.");

Create the buffer and hash, then print the result to stdout.

  char* buffer = alloca(length+1);
  assert(khash_do(&ctx, input_data, strlen(input_data), buffer, length) == KHASH_SUCCESS, "khash_do() failed.");
  buffer[length+1] = 0; // Ensure we have a NUL terminator.

  setlocale(LC_ALL, ""); //Ensure we can print UTF-8.
  printf("Kana hash: %s\n", buffer);

Alternatively, we can allocate the max length needed instead of calculating it.

size_t length;
assert(khash_max_length(KHASH_ALGO_SHA256, strlen(input_data), &length) == KHASH_SUCCESS, "khash_max_length() failed.");
char* buffer = alloca(length+1);
memset(buffer, 0, length+1); //Ensure NUL terminators.



All macros defined are for options. They cannot be combied as flags. The KHASH_ALGO_ prefixed ones are for use as the algo parameter in the khash_new_context() function. The KHASH_SALT_TYPE_ prefixed ones are for use as the salt_type parameter. The KHASH_ERROR_ prefixed ones each indicate an error code returned by all of the functions.

Name Description
KHASH_ALGO_DEFAULT The default algorithm used by the library (truncated SHA256)
KHASH_ALGO_CRC32 CRC32 checksum algorithm
KHASH_ALGO_CRC64 CRC64 checksum algorithm
KHASH_ALGO_SHA256 SHA256 hash algorithm
KHSAH_ALGO_SHA256_TRUNCATED SHA256 truncated to 64-bits
KHASH_SALT_TYPE_DEFAULT The default static salt used by the library
KHASH_SALT_TYPE_SPECIFIC A provided salt, as the data and of the size parameter passed to khash_new_context()
KHASH_SALT_TYPE_RANDOM A randomly generated salt
KHASH_SUCCESS The code returned by all of the functions when the operation was successful
KHASH_ERROR_IO There was an IO error
KHASH_ERROR_FORMAT The was a text formatting related error
KHASH_ERROR_LENGTH There was a hash length mismatch
KHASH_ERROR_RNG The random number generator failed
KHASH_ERROR_UNKNOWN There was an unknown error or the stack attempted to unwind past the FFI boundary.

There are 2 exported structs, although you will rarely need to access their members directly.

Name Field Description
khash_salt A salt allocated into a context by khash_new_context() and released by khash_free_context(). You shouldn't mess with its field directly.
salt_type The type of the salt.
size The size of the salt.
body A pointer to the body of the salt. (The memory allocated here is not guaranteed to be of the provided size.)
khash_context A context for the khash_ functions. Allocated by khash_new_context(). You can modify its fields if you want.
algo The algorithm for this context.
flags Placeholder for potential flags added in the future. Currently unused.
salt The allocated salt. You shouldn't directly mess with this field.

All defined functions return either KHASH_SUCCESS or one of the KHASH_ERROR_ values above.

Name Parameters Description
khash_new_context algo, salt_type, data, size, output Creates a new context for use with other libkhash functions. algo is expected to be one of the KHASH_ALGO_ macros listed above. Likewise salt_type is expected to be one of the KHASH_SALT_TYPE_ macros. data can be NULL unless salt_type is set to KHASH_SALT_TYPE_SPECIFIC, in which exactly size bytes are read from data. output is expected to be a valid pointer to a currently unused `khash_context` structure.
khash_free_context ctx Free a context allocated with khash_new_context(). ctx is expected to be a valid pointer to a currently allocated context.
khash_clone_context src, dst Clone a context allocated with khash_new_context() into another. The newly allocated dst must be properly released (with khash_free_context() or khash_do()) as well as the source. src is expected to be a valid pointer to an allocated context, and dst is expected to be a valid pointer to an unallocated context.
khash_length ctx, data, size, length Compute the length required to hold the output string for khash_do() for a given input. Will read exactly size bytes from data and compute the value into what is pointed to by length (which is expected to be a valid pointer to a type of size_t.) The resulting length does not include a NUL terminator for the string.
khash_do ctx, data, size, output, output_size Compute the kana-hash of size bytes from data and store no more than output_size of the the result into the string pointed to by output. Each pointer is expected to be valid. This function frees the supplied ctx after the hash has been computed, and thus ctx is no longer valid afterwards.
khash_max_length algo, input_len, output_len Calculate the max possible size for the given algorithm (expected to be one of the KHASH_ALGO_ macros) and input length, and store this result in output_len (expected to be a valid non-NULL pointer.) input_len is not required unless the algorithm is dynamically sized (all currently implemented ones are not.)

Node FFI bindings

NPM package in ./node

Installation (npm)

Follow the /flanchan/libkhash/src/branch/master/installation section first.

$ npm install --save /path/to/repo/node


Import the package
const hash = require('kana-hash');
Create a context

Create the context by specifying an algorithm identifier, and an optional salt. If provided, the salt must be of type Salt.

const ctx = new hash.Kana(hash.Kana.ALGO_DEFAULT, new hash.Salt("optional salt~"));
Create a hash

The once() function consumes the context and outputs a hash string.

const output = ctx.once("input string");
Creating a hash without consuming

If you want to reuse the context, use the hash() function.

const output = ctx.hash("input string");
Freeing the context

The context must be release after use if you have not called once().

Cloning an existing context

The new context must also be freed with either once() or finish().

const new_ctx = ctx.clone();

To create a hash in one line you can do one of the following.

  const hash1 = new Kana(Kana.ALGO_DEFAULT, new Salt("salt~")).once("input string~"); //Allocates the exact space required for the output string.
  const hash2 = Kana.single(Kana.ALGO_DEFAULT, new Salt("salt~"), "input string~"); //Allocates the max space required for the output string, instead of the exact. Might be faster.

Interface documentation

The 2 exported objects are Kana and Salt. Kana's constructor expects between 0 and 2 arguments.

  • The first is either an algorithm definition or empty, if empty Kana uses the default algorithm (truncated SHA256).

  • The second is either an instance of Salt or empty, if empty Kana uses the default library salt.

Salt's constructor expects 0 or 1 argument.

  • Either a string to use as the specific salt or empty, if empty there is no salt.

Kana also has a static function single(algo, salt, input) which automatically creates a context, computes the hash, frees the context, and then returns the computed hash as a JavaScript string.

Defined constants
Name Type Description
Kana.ALGO_DEFAULT Algorithm definition The default algorithm specified by the library (set to sha256 truncated)
Kana.ALGO_CRC32 Algorithm definition CRC32 checksum algorithm
Kana.ALGO_CRC64 Algorithm definition CRC64 checksum algorithm
Kana.ALGO_SHA256 Algorithm definition SHA256 hashing algorithm
Kana.ALGO_SHA256_TRUNCATED Algorithm definition Truncated SHA256 algorithm, to 64-bits
Salt.None Salt No salt
Salt.Random Salt A cryptographically secure random salt
Salt.Default Salt The library's default static salt


The strings generated by this library are meant to be pretty, not secure. It is not a secure way of representing a hash as many collisions are possible.

Digest algorithm

The kana algorithm is a 16-bit block digest that works as follows:

  • The most and least significant 8 bits are each seperated into Stage 0 and Stage 1 each operating on the first and second byte respectively.

  • Stage 0:

    1. The byte is sign tested (bitwise AND 0x80), store this as a boolean in sign0 (Negative becomes 1, positive becomes 0.)

    2. The valid first character range is looked up using the result of the sign test (either 0 or 1), store the range in range, and the slice KANA taken from the range in kana.

    3. The first index is calculated as the unsigned first byte modulo the size (exclusive) of range. Store this as index.

    4. Compute the value of the first byte bitwise XOR the second byte, store this as index1.

    5. The swap table is checked to see if index + start of range has an entry. Then each following step is checked in order:

      • If the swap entry exists and index1 bitwise AND 0x2 is 0, set the first character of the output to the value found in the swap table.

      • If the swap entry exists and index1 bitwise AND 0x8 is 0 and index + start of range has an entry in the 2nd swap table, set the first character of the output to the value found in the 2nd swap table.

      • In any other case, set the first character of the output to the value found in the kana slice at index.

  • Stage 1:

    1. Compute a sub table for index plus the start of range using the ranges defined in KANA_SUB_VALID_FOR and store it in sub. If there is no sub table possible, skip to step 3.

    2. If there is an entry in sub for the index of the 2nd byte modulo the size of KANA_SUB, set the second output character to be that character.

    3. If there was no value set from the sub table, the 2nd output character becomes the first output character from inputting the 2nd byte back through Stage 0 as the first byte.

  • Concatenate both characters and move to the next 16-bit block.


  • It is valid for a single iterator to produce between 0 and 2 characters but no more.

  • If an input given to the algorithm that cannot be divided exactly into 16-bit blocks (i.e. one byte is left over), a padding byte of 0 is added as the 2nd byte to make it fit.

For more information see mnemonic.rs and map.rs.


GPL'd with love <3