glue code generator to bind Nim and Lua together using Nim's powerful macro
Features:
- bind free proc
- bind proc as Lua method
- bind const
- bind enum
- bind object
- generic proc binding
- closure binding
- properties getter/setter
- automatic resolve overloaded proc
- easy namespace creation
- easy debugging
- consistent simple API
- can rename exported symbol
- support automatic type conversion
- can change binding dynamically at runtime too
- generate clean and optimized glue code that you can inspect at compile time
- it's free
planned features:
- complex data types conversion, at least standard container
- access Lua code/data from Nim
Current version API: no need to remember complicated API, the API is simple but powerful
- newNimLua
- bindEnum
- bindConst
- bindFunction/bindProc
- bindObject
Nim | Lua |
---|---|
char,int,uint,int8-64,uint8-64 | integer/number |
float, float32, float64 | number |
array[0..n, T], [n, T] | array[1..n] |
enum | integer/number |
string, cstring | string |
ref/object | userdata |
bool | boolean |
seq[T] | array[1..n] |
set[T] | table with unique element |
pointer | light user data |
ptr T | light user data |
range/subrange | integer |
openArray[T] | table -> seq[T] |
tuple | assoc-table or array |
varargs[T] | not supported |
import nimLUA, os
type
FRUIT = enum
APPLE, BANANA, PEACH, PLUM
SUBATOM = enum
ELECTRON, PROTON, NEUTRON
GENE = enum
ADENINE, CYTOSINE, GUANINE, THYMINE
proc test(L: PState, fileName: string) =
if L.doFile("test" & DirSep & fileName) != 0.cint:
echo L.toString(-1)
L.pop(1)
else:
echo fileName & " .. OK"
proc main() =
var L = newNimLua()
L.bindEnum(FRUIT, SUBATOM, GENE)
L.test("test.lua")
L.close()
main()
and you can access them at Lua side like this:
assert(FRUIT.APPLE == 0)
assert(FRUIT.BANANA == 1)
assert(FRUIT.PEACH == 2)
assert(FRUIT.PLUM == 3)
assert(GENE.ADENINE == 0)
assert(GENE.CYTOSINE == 1)
assert(GENE.GUANINE == 2)
assert(GENE.THYMINE == 3)
assert(SUBATOM.ELECTRON == 0)
assert(SUBATOM.PROTON == 1)
assert(SUBATOM.NEUTRON == 2)
another style:
L.bindEnum:
FRUIT
SUBATOM
GENE
if you want to rename the namespace, you can do this:
L.bindEnum:
GENE -> "DNA"
SUBATOM -> GLOBAL
or
L.bindEnum(GENE -> "DNA", SUBATOM -> GLOBAL)
a note on GLOBAL and "GLOBAL":
- GLOBAL without quote will not create namespace on Lua side but will bind the symbol in Lua globalspace
- "GLOBAL" with quote, will create "GLOBAL" namespace on Lua side
now Lua side will become:
assert(DNA.ADENINE == 0)
assert(DNA.CYTOSINE == 1)
assert(DNA.GUANINE == 2)
assert(DNA.THYMINE == 3)
assert(ELECTRON == 0)
assert(PROTON == 1)
assert(NEUTRON == 2)
import nimLUA
const
MANGOES = 10.0
PAPAYA = 11.0'f64
LEMON = 12.0'f32
GREET = "hello world"
connected = true
proc main() =
var L = newNimLua()
L.bindConst(MANGOES, PAPAYA, LEMON)
L.bindConst:
GREET
connected
L.close()
main()
by default, bindConst will not generate namespace, so how do you create namespace for const? easy:
L.bindConst("fruites", MANGOES, PAPAYA, LEMON)
L.bindConst("status"):
GREET
connected
first argument(actually second) to bindConst will become the namespace. Without namespace, symbol will be put into global namespace
if you use GLOBAL without quote as namespace, it will have no effect
operator ->
have same meaning with bindEnum, to rename exported symbol on Lua side
bindFunction is an alias to bindProc, they behave identically
import nimLUA
proc abc(a, b: int): int =
result = a + b
var L = newNimLua()
L.bindFunction(abc)
L.bindFunction:
abc -> "cba"
L.bindFunction("alphabet", abc)
bindFunction more or less behave like bindConst, without namespace, it will bind symbol to global namespace.
overloaded procs will be automatically resolved by their params count and types
operator ->
have same meaning with bindEnum, to rename exported symbol on Lua side
import nimLUA
type
Foo = ref object
name: string
proc newFoo(name: string): Foo =
new(result)
result.name = name
proc addv(f: Foo, a, b: int): int =
result = 2 * (a + b)
proc addv(f: Foo, a, b: string): string =
result = "hello: my name is $1, here is my message: $2, $3" % [f.name, a, b]
proc addk(f: Foo, a, b: int): string =
result = f.name & ": " & $a & " + " & $b & " = " & $(a+b)
proc main() =
var L = newNimLua()
L.bindObject(Foo):
newFoo -> constructor
addv
addk -> "add"
L.close()
main()
this time, Foo will become object name and also namespace name in Lua
"newFoo ->
constructor" have special meaning, it will create constructor on Lua side with special name: new
(this is an artefact)
but any other constructor like procs will be treated as constructor too:
L.bindObject(Foo):
newFoo #constructor #1 'newFoo'
newFoo -> constructor #constructor #2 'new'
newFoo -> "whatever" #constructor #3 'whatever'
makeFoo -> "constructor" #constructor #4 'constructor'
operator ->
on non constructor will behave the same as other binder.
overloaded proc will be automatically resolved by their params count and types, including overloaded constructor
destructor will be generated automatically for ref object, none for regular object. GC safety works as usual on both side of Nim and Lua, no need to worry, except when you manually allocated memory
local foo = Foo.new("fred")
local m = foo:add(3, 4)
-- "fred: 3 + 4 = 7"
print(m)
assert(foo:addv(4,5) == 2 * (4+5))
-- "hello: my name is fred, here is my message: abc, nop"
print(foo:addv("abc", "nop"))
operator ->
when applied to object, will rename exported symbol on Lua side:
L.bindObject(Foo -> "cat"):
newFoo -> constructor
on Lua side:
local c = cat.new("fred") --not 'Foo' anymore
both bindObject and bindFunction and bindConst can add member to existing namespace
if you want to turn off this functionality, call nimLuaOptions(nloAddMember, false)
L.bindObject(Foo): #namespace creation
newFoo -> constructor
L.bindObject(Foo): #add new member
addv
addk -> "add"
L.bindFunction("gem"): #namespace "gem" creation
mining
L.bindFunction("gem"): #add 'polish' member
polish
It's ok to call bindObject without any additional member/method if you want to register object type and use it later. For example if you want to create your own object constructor
Usually a C library have constructor(s) and destructor function. The constructor will return an opaque pointer.
On Nim side, we usually use something like:
type
CContext* = distinct pointer
proc createCContext*(): CContext {.cdecl, importc.}
proc deleteCContext*(ctx: CContext) {.cdecl, importc.}
Of course this is not an object or ref object, but we treat it as an object in this case.
Therefore bindObject will work like usual.
Only this time, we also need to specify the destructor function using ~
operator.
L.bindObject(CContext):
createCContext -> "create"
~deleteCContext
Lua basic data types cannot be passed by reference, but Nim does
if you have something like this in Nim:
proc abc(a, b: var int) =
a = a + 1
b = b + 5
then on Lua side:
a = 10
b = 20
a, b = abc(a, b)
assert(a == 11)
assert(b == 25)
basically, outval will become retval, FIFO ordered
proc mew[T, K](a: T, b: K): T =
discard
L.bindFunction:
mew[int, string]
mew[int, string] -> "mewt"
proc main() =
...
var test = 1237
proc cl() =
echo test
L.bindFunction:
[cl]
[cl] -> "clever"
type
Ship = object
speed*: int
power: int
L.bindObject(Ship):
speed(set)
speed(get) -> "currentSpeed"
speed(get, set) -> "velocity"
then you can access the object's properties on lua side using '.' (dot) and not ':' (colon)
local b = Ship.newShip()
b.speed = 19
assert(b.speed == nil) -- setter only
assert(b.currentSpeed == 19) -- getter only
b.velocity = 20
assert(b.velocity == 20) -- getter & setter
you can call nimLuaOptions(nloDebug, true/false)
nimLuaOptions(nloDebug, true) #turn on debug
L.bindEnum:
GENE
SUBATOM
nimLuaOptions(nloDebug, false) #turn off debug mode
L.bindFunction:
machine
engine
lua_error, lua_checkstring, lua_checkint, lua_checkudata and other lua C API that can throw error
are dangerous functions when called from Nim context. lua_error use longjmp
when compiled to C
or throw
when compiled to C++.
Although Nim compiled to C, Nim have it's own stack frame. Calling lua_error and other functions that can throw error will disrupt Nim stack frame, and application will crash.
nimLUA avoid using those dangerous functions and and use it's own set of functions that is considerably safe. those functions are:
lua | nimLUA |
---|---|
lua_error | N/A |
lua_checkstring | nimCheckString |
lua_checkinteger | nimCheckInteger |
lua_checkbool | nimCheckBool |
lua_checknumber | nimCheckNumber |
N/A | nimCheckCstring |
N/A | nimCheckChar |
lua_newmetatable | nimNewMetaTable |
lua_getmetatable | nimGetMetaTable |
lua_checkudata | nimCheckUData |
NLError* = object
source: string
currentLine: int
msg: string
NLErrorFunc* = proc(ctx: pointer, err: NLError) {.nimcall.}
proc NLSetErrorHandler*(L: PState, errFunc: NLErrorFunc)
proc NLSetErrorContext*(L: PState, errCtx: pointer)
This is actually not a real error handler, because you cannot use raise exception. The purpose of this function is to provide information to user about wrong argument type passed from Lua to Nim.
nimLUA already provide a default error handler in case you forget to provide one.
still under development, contributions are welcome
nimble install nimLUA
You can use compiler switch -d:SHARED_LIB_NAME="yourlibname"
$> nim c -r -d:SHARED_LIB_NAME="lua534.dll" test/test