1. Re: Elegant structures (long)
First, I'll call them records (you'll see why later).
Assume the problem solved in this manner:
[Euphoria with records]
record point
integer x,y
end record
point a
a.x=1 a.y=12
That is equivalent to "Normal Euphoria":
constant point_x=1, point_y=2, point_length=2
type point(object x)
atom res
res=false
if sequence(x) then
if length(x)=point_length then
res=integer(x[point_x]) and integer(x[point_y])
end
end
return res
end type
point a
a=repeat(0,point_length)
a[point_x]=1 a[point_y]=12 -- was: a.x=1 a.y=12
So...
a.x=1 and a.y=12 are nothing but convenient mnemonics for
a[point_x]=1 and a[point_y]=12, which are themselves mnemonics
for:
a[1] = 1 -- [1] holds x-coordinate
a[2] = 12 -- [2] holds y-coordinate
So,
1. A record is a sequence to which a sequence of symbolic names, here,
{"x","y"}, has been associated to represent its indexing range, here
[1..2].
2. Those symbolic names represent integer constants *local* to the
record
type.
3. There is one set of such symbolic names (local constants) for
each dimension of the sequence.
Further:
-------
Nothing prevents a sequence from containing routine id's. For instance:
[Euphoria with records]
record point
integer x,y
procedure draw()
-- your code here
end procedure
end record
point a
a.x=1 a.y=1
a.draw()
[Normal Euphoria]:
procedure point_draw()
-- your code here
end procedure
constant point_x=1, point_y=2, point_draw_id=3, point_length=3
type point(object x)
atom res
res=false
if sequence(x) then
if length(x)=point_length then
res=integer(x[point_x]) and integer(x[point_y])
and x[point_draw_id]=routine_id("point_draw")
end
end
return res
end type
point a
a = repeat(0,point_length)
a[point_draw_id]=routine_id("point_draw")
call_proc(a[point_draw_id])
Note how records have now become objects (that is why I chose
the term "record" rather than "structure" -- in Oberon, "record"
covers both Pascal records and objects)
Note how the declaration "point a" in "Euphoria with records"
now clearly necessitates the implicit initialization of a. And
I now realize, belatedly, that the "record" declaration should
automatically generate the definition of a type() function, too.
Let me stop soon, before going into hairy questions such as:
should a record distinguish between variables and routine id's
(a.k.a. methods) by keeping all variables in one sequence, and
all methods in another? The answers seem quite beyond me.
Plus, if I keep indulging in these theoretical considerations,
I'll never get that racing database going.
Jacques Guy
