Re: can a C extension function determine its name?

On 9/30/07, Gerald Franz <gerald.franz <at> googlemail.com> wrote:
> Dear Lua users,
>
> I was just wondering whether it is possible for a C extension function to to
> determine the name it was called by Lua.
> My idea would be to register a single C extension function multiple times to
> a Lua state, but behave (slightly) differently depending on the name it is
> called.
> For example, this would allow to unify all glue code between a C++ class and
> a Lua state within a single static method and facilitate a sharing of common
> code parts (e.g., converting the first argument to the pointer to the object
> instance).

One idea is to set up a pointer to a C++ object (some kind of Command
object) as an upvalue when defining the closure; the single C callback
function can then retrieve it, and call through it to tell the C++
Command object to execute.

-- James

Capturing Lua variable as a string

Re: can a C extension function determine its name?

Re: Capturing Lua variable as a string

Lythoner LY wrote:
> Hi All,
> 
> I am working on an interesting industrial automation application where I
> am proposing Lua to our customer as a configuration and scripting language.
> I found Lua in search for better language with small runtime, easy to
> extend feature. I like Lua because it so simple to use, understand, very
> handy and compliments python on many places.
> I need the community help to standardize my configuration stuffs which I
> am writing now.
> 
> We needed to provide a simple shell scripting capability to our user
> where they can read and write data to the device(industrial automation).
> We have the core layers developed in C# and I was able to call C# from
> Lua and Lua from C#. And also I can write & read the data with the
> device. Perfect. I also modified one window based IronPython User
> control to support Lua on .NET. For us, Lua would have a window based
> User control to allow user to read and write data. I did replaces
> "print" function Lua with my C# callable function, hence I could capture
> the "print" function output. For me, it is OK so for.
> 
> Now I needed a help on configuration.
> 
> function DEF_PARAM(v)
> return v
> end
> 
> DeviceParam1= DEF_PARAM
> {
> NAME     = "DeviceParam1",
> TYPE     = INT
> }
> 
> 
> DeviceParam2= DEF_PARAM
> {
> NAME     = "DeviceParam2",
> TYPE     = INT
> }
> 
> 
> DeviceParam3= DEF_PARAM
> {
> NAME     = "DeviceParam3",
> TYPE     = INT
> }
> 
> The above is the configuration I use. Like "DeviceParam1", I have tens
> of thousand configurations parameters. DEF_PARAM is an example function
> which just return the variable as it is. But we do some extra business
> logics inside. I have taken them out in this posting.
> 
> DeviceParam1 is used as a variable(a table instance). I need it like
> that. But I have defined one more property called "NAME" which is
> assigned to "DeviceParam1" string. The same is repeated for all
> variables. Instead I want the string equivalent of variables
> DeviceParam1, DeviceParam2, DeviceParam3 handled internally by Lua and
> assigned to  NAME property.
> 
> Here is example.
> 
> function DEF_PARAM(v)
> -- do something here to set NAME property from v
> -- v["NAME"] = ???? How do I get string equivalent of variable v
> return v
> end
> 
> DeviceParam1= DEF_PARAM
> {
> -- NAME     = "DeviceParam1", -- I have taken off the NAME assignment here
> TYPE     = INT
> }
> 
> DeviceParam2= DEF_PARAM
> {
> TYPE     = INT
> }
> 
> 
> DeviceParam3= DEF_PARAM
> {
> TYPE     = INT
> }
> 
> 
> In shell,
> 
>> print(DeviceParam3.NAME)
> DeviceParam3
> 
>>print (type(DeviceParam3.NAME)
> string
> 
> Regards,
> 
> Krish

if you have a limited number of type, you can define them as global
variable :

INT = "INT"
FLOAT = "FLOAT"

function DEF_PARAM(v)
   -- do something here to set NAME property from v
   -- v["NAME"] = ???? How do I get string equivalent of variable v
   return v
end

Just load this script before your definition file, and TYPE will
automatically contain the text.

--

-- 
--
Julien Hamaide
Engineering Coach
10Tacle Studios Belgium / Elsewhere Entertainment

Re: can a C extension function determine its name?

Re: can a C extension function determine its name?

Re: Capturing Lua variable as a string

> The same is repeated for all variables. Instead I want the string
> equivalent of variables DeviceParam1, DeviceParam2, DeviceParam3
> handled internally by Lua and assigned to NAME property.

Set up a proxy for _G and set a __newindex for the proxy that does that.
Something  like this:

    local function set(t,k,v)
	    print("SET",k,v,type(v))
	    if type(v)=="table" then v.NAME=k end
	    rawset(t,k,v)
    end
    setmetatable(getfenv(),{__index=_G,__newindex=set})

Re: Capturing Lua variable as a string

Julien Hamaide wrote:
> Lythoner LY wrote:
>> Hi All,
>>
 >> (snip)
> 
> if you have a limited number of type, you can define them as global
> variable :
> 
> INT = "INT"
> FLOAT = "FLOAT"
> 
> function DEF_PARAM(v)
>    -- do something here to set NAME property from v
>    -- v["NAME"] = ???? How do I get string equivalent of variable v
>    return v
> end
> 
> Just load this script before your definition file, and TYPE will
> automatically contain the text.
> 
> 
> 
> 

I think he meant the name of the variable:

 >> In shell,
 >>
 >>> print(DeviceParam3.NAME)
 >> DeviceParam3
 >>
 >>> print (type(DeviceParam3.NAME)
 >> string

Unfortunately, it is not possible to know what is the name of the 
variable you are assigning to (inside the DEF_PARAM function, of 
course). What I suggest, is to __newindex the chunk's environment and 
post-assign the name. Something like:

myconfiguration.lua:
DeviceParam1 = DEF_PARAM{ TYPE = INT }
DeviceParam2 = DEF_PARAM{ TYPE = INT }
DeviceParam3 = DEF_PARAM{ TYPE = INT }

~~ end of myconfiguration.lua:

configloader.lua
function load( filename )
     local f = assert( loadfile( filename ) )
     local e = setmetatable( {}, {
         __index = _G;
         __newindex = function( t, k, v )
             if type( k ) == 'string' and type( v ) == 'table' then
                 v.NAME = k
             end
         end; } )
     setfenv( f, e )
     f()
end

~~ end of configloader.lua

Of course, this will cause all assignments of tables to global variables 
  be changed. You can make some tricks to only assign the NAME fields to 
a particular subset. First, you can check for the presence of a TYPE 
attribute as a pre-condition (trivial). Second, you can store the newly 
created table in another table while creating it with DEF_PARAM and then 
check if the table being assigned belongs to it:

configloader2.lua
function load( filename )
     local f = assert( loadfile( filename ) )

     local t = {}
     local old_DEF_PARAM = DEF_PARAM

     local e = setmetatable(
         {
             DEF_PARAM = function( info )
                 -- In case DEF_PARAM creates another table...
                 info = old_DEF_PARAM( info )
                 t[ info ] = true
                 return info
             end;
         }, {
             __index = _G;
             __newindex = function( t, k, v )
                 if t[ v ] then
                     v.NAME = k
                 end
             end;
         }
     )
     setfenv( f, e )
     f()
end

~~ end of configloader2.lua

Hope that helps,

--rb

Re: Capturing Lua variable as a string

Oops, one mistake:

configloader2.lua
function load( filename )
     local f = assert( loadfile( filename ) )

     local t = {}
     local old_DEF_PARAM = DEF_PARAM

     local e = setmetatable(
         {
             DEF_PARAM = function( info )
                 -- In case DEF_PARAM creates another table...
                 info = old_DEF_PARAM( info )
                 t[ info ] = true
                 return info
             end;
         }, {
             __index = _G;
             __newindex = function( self, k, v )
                 if t[ v ] then
                     v.NAME = k
                 end
             end;
         }
     )
     setfenv( f, e )
     f()
end

~~ end of configloader2.lua

I was also using a 't' as the name of 'self' on '__newindex'. Sorry.
--rb

Weird irreproducible error

When running exactly the same test with the same data over and over, I
keep getting different results, usually with a message that says
"attempt to index field '?' (a nil value)", in different places in my
code. If I run under the ldb debugger and inspect the data that is
supposedly nil, it actually isn't.

I'm using 5.1.2. Has anyone seen this behavior before?

Thanks.

Steve