On 10/5/2011 2:31 PM, Tim Chase wrote:
On 10/04/11 20:45, Terry Reedy wrote:
On 10/4/2011 9:50 AM, Valiev Sergey wrote:
- `[]` - used for list comprehension,
- `()` - used for generators,
- `[start:stop]` / `[start:stop:step]` - used for slices.
The idea is to use `(start:stop)` / `(start:stop:step)` as 'lazy
evaluated' slices (like itertools.islice).
What do you think about it?
a(b) is already used for function calls. Making a(b:c) be something
unreleated does not seem like a good idea to me. At present, a[b:c] ==
a[slice(b,c)]. However, a(slice(b,c)) is already a function call and
could not equal a(b:c).
I'm very -1 on the initial proposal with parens, but I wouldn't object
to generators growing a method (__getitem__?) to do slices via
itertools, something like
gen = (a for a in iterator if test(a))
for thing in gen[4::2]:
do_something(thing)
acting something like
gen = (a for a in iterator if test(a))
for thing in itertools.islice(gen, start=4, step=2):
'end' arg is required
do_something(thing)
islice(gen,4,None,2) is 11 more characters than
gen[4::2]
If you start with 'from itertools import islice as sl', then each use is
7 more chars. In the normal case with an 'end' arg, the difference would
be 4 chars less:
sl(gen,4,100,2) # 3 extra chars to type ;=)
gen[4:100:2]
This would complicate the language and make generators more different
from other iterators without adding new functionality
--
Terry Jan Reedy
--
http://mail.python.org/mailman/listinfo/python-list