Sorted Returns List and Reversed Returns Iterator

[++imanshu]

Hi,

Is there a reason why two similarly named functions Sorted and
Reversed return different types of data or is it an accident.

Thanks,
++imanshu

 

[John Machin]

Hi,

     Is there a reason why two similarly named functions Sorted and
Reversed return different types of data or is it an accident.

You seem to have an interesting notion of "similarly named".
name0[-2:] == name1[-2:], perhaps? The two functions (eventually, in
the case of "reversed") return data in the order one would expect from
their names.

x = [1, 3, 5, 2, 4, 6]
sorted(x) [1, 2, 3, 4, 5, 6]
reversed(x)

list(reversed(x)) [6, 4, 2, 5, 3, 1]

 

[John Machin]

Hi,

     Is there a reason why two similarly named functions Sorted and
Reversed return different types of data or is it an accident.

You seem to have an interesting notion of "similarly named".
name0[-2:] == name1[-2:], perhaps? The two functions (eventually, in
the case of "reversed") return data in the order one would expect from
their names.

x = [1, 3, 5, 2, 4, 6]
sorted(x) [1, 2, 3, 4, 5, 6]
reversed(x)

<listreverseiterator object at 0x00AA5550>


[6, 4, 2, 5, 3, 1]-

Sorry; having re-read the message subject:

reversed came later; returning an iterator rather than a list provides
more flexibility.

Cheers,
John

 

[++imanshu]

You seem to have an interesting notion of "similarly named".
name0[-2:] == name1[-2:], perhaps? The two functions (eventually, in
the case of "reversed") return data in the order one would expect from
their names.

x = [1, 3, 5, 2, 4, 6]
sorted(x) [1, 2, 3, 4, 5, 6]
reversed(x)

<listreverseiterator object at 0x00AA5550>

list(reversed(x))

[6, 4, 2, 5, 3, 1]-

Sorry; having re-read the message subject:

reversed came later; returning an iterator rather than a list provides
more flexibility.

Cheers,
John

I agree. Iterator is more flexible. Together and both might have
returned the same types.

Thanks,
++imanshu

 

[Peter Otten]

Is there a reason why two similarly named functions Sorted and
Reversed return different types of data or is it an accident.

You seem to have an interesting notion of "similarly named".
name0[-2:] == name1[-2:], perhaps? The two functions (eventually, in
the case of "reversed") return data in the order one would expect from
their names.

x = [1, 3, 5, 2, 4, 6]
sorted(x)
[1, 2, 3, 4, 5, 6]
reversed(x)

<listreverseiterator object at 0x00AA5550>

list(reversed(x))
[6, 4, 2, 5, 3, 1]-

Sorry; having re-read the message subject:

reversed came later; returning an iterator rather than a list provides
more flexibility.

Cheers,
John

I agree. Iterator is more flexible. Together and both might have
returned the same types.

It's easy to generate a reversed sequence on the fly but impractical for a
sorted one. Python is taking the pragmatic approach here.

Peter

 

[Fredrik Lundh]

reversed came later; returning an iterator rather than a list provides
more flexibility.

As in flexibility for the implementer, the day someone invents a sort
algorithm that doesn't have to look at all source items before it starts
producing output?

Because I fail to see how returning an object that supports forward
iteration only is more flexible for the *user* than returning an object
that supports random access, mutation, restartable iteration, *and*
forward iteration.

</F>

 

[Terry Reedy]

++imanshu wrote:

I disagree. Neither is more flexible. You can iter the list returned
by sorted and list the iter returned by reversed. Both do the minimum
work necessary. See below.

> Together and both might have returned the same types.

True, but only by doing potentially unnecessary work and requiring the
caller to do potentially unnecessary work that might even prevent the
program from working. This is less flexible.

Suppose sorted now returns alist with 50 million items. Suppose it
instead returned iter(alist) but the caller wants to randomly index the
items. Since the caller could not access the existing 50 million item
list, the caller would have to make another 50 million item copy. This
is non-trivial and might not even work do to memory limitations.

It's easy to generate a reversed sequence on the fly but impractical for a
sorted one. Python is taking the pragmatic approach here.

To expand on this: sorting and reversing are algorithmically different
operations. Sorting requires that one have all items in hand in a
mutable sequence (list) for arbitrary re-ordering. Sorted works on any
iterable and starts by making a new list. There is no point to not
returning that list after it is sorted. It would be more work and less
useful to do more.

sorted(iterable, key=None, reverse=False):
newlist = list(iterable)
newlist.sort(key, reverse)
return newlist

Iterating over a concrete sequence in reverse order, on the other hand,
is trivial. It would be more work and less useful to do more.

def _reversed(seq): # 'hidden' generator function
n = len(seq)
while n:
n -= 1
yield seq[n]

def reversed(seq):
if hasattr(seq, '__reversed__'):
return seq.__reversed__() # I presume this is tried first
else:
return _reversed(seq) # generic fall-back

Terry Jan Reedy

 

[++imanshu]

I disagree.  Neither is more flexible.  You can iter the list returned
by sorted and list the iter returned by reversed.  Both do the minimum
work necessary.  See below.

 > Together and both might have returned the same types.

True, but only by doing potentially unnecessary work and requiring the
caller to do potentially unnecessary work that might even prevent the
program from working.  This is less flexible.

Suppose sorted now returns alist with 50 million items.  Suppose it
instead returned iter(alist) but the caller wants to randomly index the
items.  Since the caller could not access the existing 50 million item
list, the caller would have to make another 50 million item copy.  This
is non-trivial and might not even work do to memory limitations.

It's easy to generate a reversed sequence on the fly but impractical for a
sorted one. Python is taking the pragmatic approach here.

To expand on this: sorting and reversing are algorithmically different
operations.   Sorting requires that one have all items in hand in a
mutable sequence (list) for arbitrary re-ordering.  Sorted works on any
iterable and starts by making a new list.  There is no point to not
returning that list after it is sorted.  It would be more work and less
useful to do more.

sorted(iterable, key=None, reverse=False):
   newlist = list(iterable)
   newlist.sort(key, reverse)
   return newlist

Iterating over a concrete sequence in reverse order, on the other hand,
is trivial.  It would be more work and less useful to do more.

def _reversed(seq): # 'hidden' generator function
   n = len(seq)
   while n:
     n -= 1
     yield seq[n]

def reversed(seq):
   if hasattr(seq, '__reversed__'):
     return seq.__reversed__() # I presume this is tried first
   else:
     return _reversed(seq) # generic fall-back

Terry Jan Reedy

Thanks for giving the 'behind the scenes' reasons. It looks
reasonable now.

Thank You,
++imanshu