[Python-checkins] python/dist/src/Lib pickle.py,1.101,1.102

[Tim Peters]

> Modified Files:
> =09pickle.py
> Log Message:
> More protocol 2: TUPLE1, TUPLE2, TUPLE3.
>
> Also moved the special case for empty tuples from save() to save_tu=
ple().

<relief>


> +         if proto >=3D 1:
> +             n =3D len(object)
> +             if n <=3D 3:
> +                 if not object:
> +                     write(EMPTY_TUPLE)
> +                     return
> +                 if proto >=3D 2:
> +                     for element in object:
> +                         save(element)
> +                     # Subtle.  Same as in the big comment below
> +                     if id(object) in memo:
> +                         get =3D self.get(memo[id(object)][0])
> +                         write(POP_MARK + get)
> +                     else:
> +                         write(_tuplesize2code[n])
> +                     return

When proto >=3D 2 and n in (1, 2, 3), the resulting tuple never gets =
memoized,
so it doesn't seem possible for "id(object) in memo" to be true.  If =
it were
possible, then the POP_MARK opcode is going to look back for a MARK t=
hat was
never pushed ...

OK, if I try pickling the one-element tuple T containing a one-elemen=
t list
L containing T under proto 2, the opcodes are:

    0: =C7 PROTO      2
    2: ] EMPTY_LIST
    3: q BINPUT     0
    5: h BINGET     0
    7: =E0 TUPLE1
    8: a APPEND
    9: =E0 TUPLE1
   10: . STOP

That isn't right, but is what happens if "id(object) in memo" is neve=
r true.
I'll try to fix it.