To facilitate the development of contracts on ckb platform, we recently ported Lua, a scripting language famous for its simplicity and ease of use, to the ckb-vm. Ckb developers can now both run the standalone Lua interpreter in the ckb platform and embed Lua code into their main ckb contracts. This greatly lower the barrier to entry for ckb contracts development.
dylib.c a sample program that evaluates lua code with ckb-lua shared library. We showcase how to load and unpack script args by a few lines of Lua code, i.e.
local _code_hash, _hash_type, args, err = ckb.load_and_unpack_script()
print(err)
if err == nil then
ckb.dump(args)
end
Of course, this example is not enough for real world usage, but this article should be enough to show how to use ckb-lua shared library from the main contract program. You can unleash the potential of Lua with your creativity afterward. For the curious, see here for an implementation of sudt in lua. Also note that, the sample code is written in c. As an alternative, you can write test code in rust. Both loading shared library and mocking transaction to do unit tests are easy in rust.
I will now explain what this sample code does and how to run it.
The sample code depends on libckblua.so
shared library to run Lua code. We need to obtain the source of ckb-lua from the github repository, and build it ourselves. make all-via-docker
is enough to build all the binaries. ./build/libckblua.so
may be used afterward. The sample program ./build/dylibexample
built from dylib.c is also available to use.
To use the ckb-lua shared library, we need to store the shared library in some cell data and specify the cell with shared library data as a dependent cell. We have assembled a mock transaction file here. You can also mock this with the ckb rust SDK (like this). You can use it along with ckb-standalone-debugger to run the sample code. For example, you can run ./build/dylibexample
with ckb-debugger --tx-file ./tests/test_cases/dylib.json --script-group-type=type --cell-index=0 --cell-type=output --bin build/dylibexample
.
We need to pass the cell information to the args of the main c contract.
We performed some standard operations to obtain the args of the current script, i.e. loading the script with ckb_load_script
and then de-serialize the args with molecule.
After we have obtained the args, we can unpack the cell information contained within the args.
In our example, our args has the following format.
// <reserved args, 2 bytes> <code hash of the shared library, 32 bytes>
// <hash type of shared library, 1 byte>
Given the size and offset of fields, we can easily obtain the code hash and hash type of the dependent cell. Now, we can open the shared library by ckb_dlopen2
.
err = ckb_dlopen2(code_hash, hash_type, code_buff, code_buff_size, handle, &consumed_size);
is a ckb-c-stdlib function to load the shared library from dependent cell(s) with specific code_hash
and hash_type
. If the shared library is found, ckb_dlopen2
will then write the shared library handle to the handle
parameter. The handle
parameter can then be used to obtain pointers to our Lua functions.
We can use the following exported functions to call Lua scripts.
Three functions have been exported in the shared library.
void *lua_create_instance(uintptr_t min, uintptr_t max)
int lua_run_code(void *l, const char *code, size_t code_size, char *name)
void close_lua_instance(void *L)
void lua_toggle_exit(void *L, int enabled)
To create new Lua instance, we may use lua_create_instance
. In order to avoid the collision of lua’s malloc and the host program’s malloc function. We need to specify the upper and lower bound of the memory that can be used exclusively by Lua (this is a contract that should be enforced by the host program). The host program can allocate this space in the stack or heap. The return value of this function is an opaque pointer that can be used to evaluate Lua code with lua_run_code
. We can reclaim the resources by calling lua_close_instance
.
The actual evaluation of Lua code can be initiated by calling lua_run_code
. Both Lua source code and binary code are acceptable to lua_run_code
. Evaluating binary code can be less resource-hungry both in terms of the cycles needed to run the code and the storage space required to store the code. It takes four arguments. The first one l
is the opaque pointer returned from lua_create_instance
, the second one code
is a pointer which points to the code buffer, which can be either Lua source code or Lua bytecode. The third one code_size
is the size of the code
buffer. The last one name
is the name of the Lua program. This may be helpful in debugging. The lua script may early return control to the host program by calling ckb.exit_script(code)
. In this case, the return code of this function is the parameter code
. If no ckb.exit_script
is called, 0 will be returned if lua script is executed successfully, otherwise a negative number will be returned. Lua script should pass non-negative number to ckb.exit_script
in order to distinguish lua interpreter error and delibrately set return code.
Calling lua_close_instance
will release all the resources used by Lua. It takes the opaque pointer returned from lua_create_instance
as a parameter.
lua_toggle_exit
can be used to toggle whether ckb.exit
is enabled. Since ckb.exit(code)
can stop the execution of the whole VM and return code
to ckb-vm. It may be undesirable in some situation. By default, we have disabled it. Users may call lua_toggle_exit(l, 1)
to enable it (here parameter l
is the lua instance).
Most of the functions in Lua standard library have been ported to ckb-lua. You can expect all the platform-independent functions to work. But some platform-dependent functions (e.g. all functions in the module os
) are not provided in ckb-lua.
TODO: add list of functions that are not available in ckb-lua.
On the other hand, we have added a few ckb-specific helper functions to interact with ckb-vm. These include functions for issuing syscalls, debugging, and unpacking data structures. All of them are located in the ckb
Lua module.
See the appendix for a list of exported constants and functions.
Actually running the lua code is quite easy.
CreateLuaInstanceFuncType create_func =
must_load_function(handle, "lua_create_instance");
EvaluateLuaCodeFuncType evaluate_func =
must_load_function(handle, "lua_run_code");
CloseLuaInstanceFuncType close_func =
must_load_function(handle, "lua_close_instance");
const size_t mem_size = 1024 * 512;
uint8_t mem[mem_size];
void* l = create_func((uintptr_t)mem, (uintptr_t)(mem + mem_size));
if (l == NULL) {
printf("creating lua instance failed\n");
return;
}
int ret = evaluate_func(l, code, code_size, "test");
if (ret != 0) {
printf("evaluating lua code failed: %d\n", ret);
return;
}
close_func(l);
We first load the functions from the previous returned dynamic library handle, and then create a stack space for Lua to use. Afterwards, we can call the exported functions.
The Lua code
local _code_hash, _hash_type, args, err = ckb.load_and_unpack_script()
print(err)
if err == nil then
ckb.dump(args)
end
loads the ckb script and unpacks it into code_hash
, hash_type
and args
. If there is anything wrong when performing this operation, an error would return in err
. Since the script args can be arbitrary binary data, it may be hard to inspect. We have also added a dump
function to dump the hex and ASCII representations of the data.
The prototype of ckb-lua is now somewhat complete. We are now gathering feedback on the usage of ckb-lua. If you have any use cases that are not easily met by the current ckb-lua implementation in mind, please feel free to leave comments at the ckb-lua repository.
The functions here may take a variable length of arguments. There is a partial loading like mechanism for Lua functions. Functions below with partial loading supported may be additionally passed to the argument length and offset. The behavior of these Lua functions is classified as follows
- Both length and offset are omitted.
In this case, the whole data would be loaded, e.g. calling
ckb.load_witness(0, ckb.SOURCE_INPUT)
would load the whole witness of input cell of index 0. - The offset is omitted, and the length passed is zero.
In this case, the length of the whole data would be returned, e.g.
ckb.load_witness(0, ckb.SOURCE_INPUT, 0)
will return the length of the witness instead of the witness data. - The offset is omitted, and the length passed is non-zero.
In this case, the initial data of the argument length bytes would be returned, e.g.
ckb.load_witness(0, ckb.SOURCE_INPUT, 10)
will return the initial 10 bytes of witness. - Both length and offset are passed, and the length passed is zero.
In this case, the data length starting from offset would be returned, e.g. calling
ckb.load_witness(0, ckb.SOURCE_INPUT, 0, 10)
would the data length starting from offset 10. - Both length and offset are passed, and the length passed is non-zero.
In this case, the data starting from offset of length would be returned, e.g. calling
ckb.load_witness(0, ckb.SOURCE_INPUT, 2, 10)
would return 2 bytes of data starting from 10.
Most of they may return an error as the last argument. The error is a non-zero integer that represents the kind of error that happened during the execution of the function. Callers should check errors to be nil before using other arguments.
Note when partial loading support is enabled, the description for arguments length and offset is omitted. You may optionally pass length and offset in that case.
description: dump a returned buffer to stdout
calling example: ckb.dump(buf)
arguments: buf (a buffer returned from syscalls below)
return values: none
side effects: the buf is printed
description: exit the ckb-vm execution
calling example: ckb.exit(code)
arguments: code (exit code)
return values: none
side effects: exit the ckb-vm execution
see also: ckb_exit
syscall
description: stop lua script execution, return to the ckb-vm caller
calling example: ckb.exit_script(code)
arguments: code (this will be the return code of lua_run_code
, should be positive, see lua_run_code
above for details)
return values: none
side effects: stop lua script execution, return to the ckb-vm caller
description: load the cell data and mount the file system witin in
calling example: ckb.mount(source, index)
arguments: source (the source of the cell to load), index (the index of the cell to load within all cells with source source
)
return values: err (may be nil object to represent possible error)
side effects: the files within the file system will be available to use if no error happened
see also: file system documentation
description: load the transaction hash
calling example: buf, err = ckb.load_tx_hash()
arguments: none
partial loading supported: yes
return values: buf (the buffer that contains the transaction hash), err (may be nil object to represent possible error)
see also: ckb_load_tx_hash
syscall
description: load the hash of current script
calling example: buf, error = ckb.load_script_hash()
arguments: none
partial loading supported: yes
return values: buf (the buffer that contains the script hash), err (may be nil object to represent possible error)
see also: ckb_load_script_hash
syscall
description: load current script
calling example: buf, error = ckb.load_script()
arguments: none
partial loading supported: yes
return values: buf (the buffer that contains script), err (may be nil object to represent possible error)
see also: ckb_load_script
syscall
description: load and unpack current script
calling example: code_hash, hash_type, args, error = ckb.load_and_unpack_script()
arguments: none
return values: code_hash (the code hash of current script), hash_type (the hash type of current script), args (the args of current script), err (may be nil object to represent possible error)
description: load current transaction
calling example: buf, error = ckb.load_transaction()
arguments: none
partial loading supported: yes
return values: buf (the buffer that contains current transaction), err (may be nil object to represent possible error)
see also: ckb_load_transaction
syscall
description: load cell
calling example: buf, error = ckb.load_cell(index, source)
arguments: index (the index of the cell), source (the source of the cell)
partial loading supported: yes
return values: buf (the buffer that contains cell), err (may be nil object to represent possible error)
see also: ckb_load_cell
syscall
description: load input cell
calling example: buf, error = ckb.load_input(index, source)
arguments: index (the index of the cell), source (the source of the cell)
partial loading supported: yes
return values: buf (the buffer that contains the input cell), err (may be nil object to represent possible error)
see also: ckb_load_input
syscall
description: load cell header
calling example: buf, error = ckb.load_header(index, source)
arguments: index (the index of the cell), source (the source of the cell)
partial loading supported: yes
return values: buf (the buffer that contains the header), err (may be nil object to represent possible error)
see also: ckb_load_header
syscall
description: load the witness
calling example: buf, error = ckb.load_witness(index, source, length, offset)
arguments: index (the index of the cell), source (the source of the cell)
partial loading supported: yes
return values: buf (the buffer that contains the witness), err (may be nil object to represent possible error)
see also: ckb_load_witness
syscall
description: load cell data
calling example: buf, error = ckb.load_cell_data(index, source, length, offset)
arguments: index (the index of the cell), source (the source of the cell)
partial loading supported: yes
return values: buf (the buffer that contains the cell data), err (may be nil object to represent possible error)
see also: ckb_load_cell_data
syscall
description: load cell data field
calling example: buf, error = ckb.load_cell_by_field(index, source, field)
arguments: index (the index of the cell), source (the source of the cell), field (the field to load)
partial loading supported: yes
return values: buf (the buffer that contains the cell data field), err (may be nil object to represent possible error)
see also: ckb_load_cell_by_field
syscall
description: load input field
calling example: buf, error = ckb.load_input_by_field(index, source, field)
arguments: index (the index of the cell), source (the source of the cell), field (the field to load)
partial loading supported: yes
return values: buf (the buffer that contains the input field), err (may be nil object to represent possible error)
see also: ckb_load_input_by_field
syscall
description: load header by field
calling example: ckb.load_header_by_field(index, source, field)
arguments: index (the index of the cell), source (the source of the cell), field (the field to load)
partial loading supported: yes
return values: buf (the buffer that contains the header field), err (may be nil object to represent possible error)
see also: ckb_load_header_by_field
syscall
description: unpack the buffer that contains the molecule structure Script
calling example: ckb.unpack_script(buf)
arguments: buf (the buffer that contains the molecule structure Script
)
return values: table (a table contains keys code_hash
, hash_type
and args
), err (may be nil object to represent possible error)
see also: The molecule definition of Script
description: unpack the buffer that contains the molecule structure WitnessArgs
calling example: ckb.unpack_witnessargs(buf)
arguments: buf (the buffer that contains the molecule structure WitnessArgs
)
return values: table (a table contains keys lock
, input_type
and output_type
), err (may be nil object to represent possible error)
see also: The molecule definition of WitnessArgs
description: unpack the buffer that contains the molecule structure OutPoint
calling example: ckb.unpack_outpoint(buf)
arguments: buf (the buffer that contains the molecule structure OutPoint
)
return values: table (a table contains keys tx_hash
and index
), err (may be nil object to represent possible error)
see also: The molecule definition of OutPoint
description: unpack the buffer that contains the molecule structure CellInput
calling example: ckb.unpack_cellinput(buf)
arguments: buf (the buffer that contains the molecule structure CellInput
)
return values: table (a table contains keys since
and previous_output
), err (may be nil object to represent possible error)
see also: The molecule definition of CellInput
description: unpack the buffer that contains the molecule structure CellOutput
calling example: ckb.unpack_celloutput(buf)
arguments: buf (the buffer that contains the molecule structure CellOutput
)
return values: table (a table contains keys capacity
, lock
and type
), err (may be nil object to represent possible error)
see also: The molecule definition of CellOutput
description: unpack the buffer that contains the molecule structure CellDep
calling example: ckb.unpack_celldep(buf)
arguments: buf (the buffer that contains the molecule structure CellDep
)
return values: table (a table contains keys out_point
and dep_type
), err (may be nil object to represent possible error)
see also: The molecule definition of CellDep
While most constants here are directly taken from ckb_consts.h, we also defined some are ckb-lua specific constants like ckb.LUA_ERROR_INTERNAL
.
ckb.SUCCESS
ckb.INDEX_OUT_OF_BOUND
ckb.ITEM_MISSING
ckb.LENGTH_NOT_ENOUGH
ckb.INVALID_DATA
ckb.LUA_ERROR_INTERNAL
ckb.LUA_ERROR_OUT_OF_MEMORY
ckb.LUA_ERROR_ENCODING
ckb.LUA_ERROR_SCRIPT_TOO_LONG
ckb.LUA_ERROR_INVALID_ARGUMENT
ckb.SOURCE_INPUT
ckb.SOURCE_OUTPUT
ckb.SOURCE_CELL_DEP
ckb.SOURCE_HEADER_DEP
ckb.SOURCE_GROUP_INPUT
ckb.SOURCE_GROUP_OUTPUT
ckb.CELL_FIELD_CAPACITY
ckb.CELL_FIELD_DATA_HASH
ckb.CELL_FIELD_LOCK
ckb.CELL_FIELD_LOCK_HASH
ckb.CELL_FIELD_TYPE
ckb.CELL_FIELD_TYPE_HASH
ckb.CELL_FIELD_OCCUPIED_CAPACITY
ckb.INPUT_FIELD_OUT_POINT
ckb.INPUT_FIELD_SINCE
ckb.HEADER_FIELD_EPOCH_NUMBER
ckb.HEADER_FIELD_EPOCH_START_BLOCK_NUMBER
ckb.HEADER_FIELD_EPOCH_LENGTH