corsix / ffi-reflect Goto Github PK

Hi,
Let me first say that I'm quite excited about your reflection library and currently testing it out.
Concretely, I'm trying to iterate over the fields of substructs and it doesn't seem to work in the cases where I'd need it to. Maybe you have some advice. Here's what I am doing.

The following slightly shortened example from the docs works:

for refct in reflect.typeof[[
  struct {
    int a;
    union { int b; int c; };
    struct { int e; int f; };
  }
]]:members() do print(refct.what) end --> field, union, struct

Naming the outer struct has the fields of the inner struct inlined(?) but they can still be iterated.

-- from https://github.com/corsix/ffi-reflect/blob/master/test.lua
local function rec_members(refct, f)
  if refct.members then
    for refct in refct:members() do
      rec_members(refct, f)
    end
  else
    f(refct)
  end
end

ffi.cdef [[
  struct Test1 {
    int a;
    union { int b; int c; };
    struct { int e; int f; };
  };
]]
rec_members(reflect.typeof("struct Test1"), function(rc) print(rc.what) print(rc.name) end)
--> field a field b field c field e field f

If I name the inner struct, I cannot access its fields at all.

ffi.cdef [[
  struct Inner { int e; int f; };
  struct Test2 {
    int a;
    union { int b; int c; };
    struct Inner inner;
  };
]]
rec_members(reflect.typeof("struct Test2"), function(rc) print(rc.what) print(rc.name) end)
--> field a field b field c field inner

Is there another way to make this work? Any chance that getting this only requires a small change?

I see there are alas differences in LuaJIT 2.1. It is probably a moving target, so may be too early to look at it...

Running test.lua in windows7 LuaJIT 2.1 Mingw-W64 compiled for 32 bits gives a crash. (no info on gdb)
Inserting do return end before line 142 crash is gone.

Hello, I'm encountering the following

ffi.cdef[[ struct ImVec4 { float x,y,z,w; }; ]]

reflect.typeof(ffi.typeof("      struct ImVec4")) -- okay
reflect.typeof(ffi.typeof("const struct ImVec4")) -- complains due to some TODO

error :

./reflect.lua:208: TODO: CTA_QUAL
stack traceback:
	[C]: in function 'error'
	./reflect.lua:208: in function 'CTA'
	./reflect.lua:281: in function 'typeof'

which is into those lines belonging to reflect.lua :

-- Logic for merging an attribute CType onto the annotated CType.
local CTAs = {[0] =
  function(a, refct) error("TODO: CTA_NONE") end,
  function(a, refct) error("TODO: CTA_QUAL") end,          <<< ERROR THERE
  function(a, refct)
    a = 2^a.value
    refct.alignment = a
    refct.attributes.align = a
  end,
  function(a, refct)
    refct.transparent = true
    refct.attributes.subtype = refct.typeid
  end,
  function(a, refct) refct.sym_name = a.name end,
  function(a, refct) error("TODO: CTA_BAD") end,
}

The "types" that I get in input are not under my control, I get those const ImVec4 from a 3rd party : https://github.com/cimgui/cimgui/blob/docking_inter/generator/output/definitions.lua

Of course I could also workaround my problem by tricking my code into stripping the const problematics to reflect.typeof()

Would you have time... to do... this TODO: CTA_QUAL ? :)

Thanks,

Kind Regards

The assertion on line 105 of the bundled test.lua fails.
Here, element_type.convention is 'cdecl' rather than 'stdcall'.
Is this to be expected, or is there something weird happening?

Hello - I guess I would like to iterate over tab table:
(CTState or init_CTState()).tab

Could you suggest how could I do this?

Regards
Pawel

Hello,

Recently LuaJIT 2.1 branch enabled LJ_GC64 mode by default. It would be great to have support for it, I find this library very useful.

Currently on x86_64 reflect.getmetatable() is giving me Segmentation fault.

Hello, I have noticed that name field in CType is uint32_t:

uint32_t name;

In GC64 mode this field (GCref) is uint64_t - is it intentionally?

Regards
Pawel