Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #63888

Re: L[:]

Show key headers only | View raw

 On 1/13/2014 4:00 AM, Laszlo Nagy
 wrote:
 > 
 >> Unless L is aliased, this is silly code.
 > There is another use case. If you intend to modify a
 list within a for
 > loop that goes over the same list, then you need to
 iterate over a copy.
 > And this cannot be called an "alias" because it has no
 name:
 
 for i in somelist: creates a second reference to somelist
 that somewhere in the loop code has a name, so it is
 effectively an 'alias'. The essential point is that there
 are two access paths to the same object.
 
 > for idx,item in enumerate(L[:]):
 >     # do something with L here,
 including modification
 
 The copy is only needed in the above if one inserts or
 deletes. But if one inserts or deletes more than one item,
 one nearly always does better to iterate through the
 original and construct a new list with new items added and
 old items not copied.
 

====> Hi, first, thank you all for your replies -much appreciated!
Terry, this would be making a shallow copy, right? If so, then "list(L)" is slightly nicer IMHO, but that's probably a matter of taste (however, I don't like copy.copy, even though that's perhaps most clear --oh well nitpicking ;-)

I also found that item assignment ([1] below) is much faster than using the more standard (I think) .append ([2]).
# [1]
for idx,item in enumerate(L[:]):
   if some_condition: 
      L[idx] = foobarify(item)
# [2]
L2 = []     
for idx,item in enumerate(L):
   if some_condition: 
       L2.append(foobarify(item))
   else:
       L2.append(item)

Back to comp.lang.python | Previous | Next | Find similar | Unroll thread

Thread

Re: L[:] Albert-Jan Roskam <fomcl@yahoo.com> - 2014-01-14 02:42 -0800

csiph-web