Hi Alex, Thanks for the explanation. I think I’ll have to adjust EmuLisp then, so it will follow Ersatz more closely.
/Jon On 23. Jun, 2014, at 16:36, Alexander Burger <[email protected]> wrote: > Hi Jon, > >> In the docs at <http://www.software-lab.de/doc/ref.html#symbol> I read >> that "The CDR of a symbol cell is also called VAL, and the CAR points to >> the symbol's tail”, but in the Ersatz code (built from fun.src) I see >> this ... > > Right. > >> final static class Symbol extends Any { >> Object Obj; >> Any Prop[]; >> String Name; >> >> Symbol(Any val, String nm) { >> Car = val == null? this : val; >> Name = nm; >> } >> >> … and here it seems to me that you store the Symbol’s value in the ‘Car’. > > Correct. ErsatzLisp cannot use that very efficient two-pointer cell > structure, because Java doesn't have pointers! :( > > Instead, the cell structure must be emulated in an object. This is > unfortuate, because instead of 8 bytes (on a 32-byte system), a symbol > in ErsatzLisp needs 44 bytes (iirc). > > >> Can you explain why it is CDR==VAL in the docs, but "Car = val” >> (instead of "Cdr = val”) in Ersatz? > > This is not so wrong as it might look ;-) > > The description above is about the _cell_ structure of a symbol > > Symbol > | > V > +-----+-----+ > | | | VAL | > +--+--+-----+ > | tail > | > V > > BUT the behavior of the function 'car' is different: It is just a > pointer-dereference! It takes the value where a pointer points to. > > In case of a (list) cell > > Pair > | > V > +-----+-----+ > | CAR | CDR | > +-----+-----+ > > the pointer points to the first _half_ of the cell, and dereferencing it > gives the 'CAR' value. > > But if you look at the diagram of the symbol above, you see that the pointer > points to the _second_ half of the cell, yielding the 'VAL'. > > > If you look at the doc of 'car', you see > > (car 'var) -> any > > telling us that 'car' accepts a 'var' (which is either a symbol or a cell), > while > the function 'cdr' > > (cdr 'lst) -> any > > only accepts a 'lst' (either NIL or a cell). > > So the function 'car' can be applied to both a symbol or a list cell, > and will return a meaningful value, while 'cdr' can't do that (as also > follows from the 'Symbol' diagram above: Applying 'cdr' to a symbol > would return the CAR of the _next_ cell in memory, giving meaningless > results). > > (car 'Sym) is equivalent to (val 'Sym). > > All that behavior is a result of the pointer tag systematics of > PicoLisp. > > ♪♫ Alex PԔ � &j)m����X�����zV�u�.n7�
