Question about SETFIELD nil emission in table constructor codegen

Hi all,

I’m looking at table constructor codegen in Lua 5.5.

For code like:

local t = {
  a = nil,
  b = nil,
  c = nil,
}

luac -l -p outputs:

1       [1]     VARARGPREP      0
2       [1]     NEWTABLE        0 3 0   ; 0
3       [1]     EXTRAARG        0
4       [2]     SETFIELD        0 0 1k  ; "a" nil
5       [3]     SETFIELD        0 2 1k  ; "b" nil
6       [4]     SETFIELD        0 3 1k  ; "c" nil
7       [5]     RETURN          1 1 1   ; 0 out

From ltable.c, luaH_newkey does not insert a new entry when the value is
nil (if (!ttisnil(value)) ...), so these writes do not seem to result in
actual table entries.

My question is more about semantics:
for constructor fields that are statically unique and explicitly nil, is
there any specific semantic reason these SETFIELD instructions must still
be emitted? Or is this mainly a simplicity / uniformity choice in the code
generator?

Thanks.

-- 
You received this message because you are subscribed to the Google Groups "lua-l" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
To view this discussion visit https://groups.google.com/d/msgid/lua-l/CABugeg9rp54TBAfHspEYNzVp2GfXD%3DjN4%2B%3DA16%2Bevox1_hqQTg%40mail.gmail.com.