On 01/19/2014 02:59 PM, Mark Lawrence wrote:
On 19/01/2014 13:23, spir wrote:
On 01/18/2014 07:20 PM, Pierre Dagenais wrote:
Hello,
I wish to fill a list called years[] with a hundred lists called
year1900[], year1901[], year1902[], ..., year1999[]. That is too much
typing of course. Any way of doing this in a loop? I've tried stuff like
("year" + str(1900)) = [0,0] but nothing works.
Any solution?
Thank you,
PierreD.
I think Danny & Wiktor's solutions are the right ones. Danny's is faster
a simpler (because of direct array indexing), but does not allow giving
the true year "names" (actually numbers). Wiktor more correctly matches
you expectation bt is a bit slower because we're searching individual
years in a dict.
I suspect that the speed difference between accessing a list by direct indexing
and looking up something in a dict is negligible. I'll happily leave Steven
D'Aprano to supply us with the actual figures from the timeit module :)
It's 2-3 times slower, I guess (measures in other langs, including lua &
hand-implemented in C, but not in python). The task is accessing an entry which
key is a plain natural number, meaning a sparse array. It is typically
implemented as a "mod table" (my term), meaning a hash table without hashing
since keys already are natural numbers: only remains the modulo (actually just a
mask since the base is a power of 2). The overhead is consistently 2-3 times,
but as you say is often neglectable in practice since time remains small
compared to whatever one does with data, once accessed.
Lua for this reason nevertheless optimises function local scopes by getting rid
of names (unless one explicitely uses them, sort of metaprogramming) and
replacing them with indexes in a plain array (actually kind of custom call
stack). This is as said 2-3 times faster than access of globals (which names
remain, but *interned* thus we have a sparse array, and the scope is the Lua
version of a dict).
I guess (not sure) python optimises access of dicts used as scopes (also of
object attributes) by interning id-strings and thus beeing able to replace them
by hash values already computed once for interning, or other numeric codes, as
keys in dicts. Thus, we fall back to the case of a sparse array as above (but in
this case names are still there and accessible in the string pool).
[This is my preferred method, despite the additional complication of a string
pool, and additional weight of the numeric key in strings. And once we have the
machinary in place, it can be used for other purposes, like interning small user
strings or, as in python, object attr names.]
Denis
_______________________________________________
Tutor maillist - Tutor@python.org
To unsubscribe or change subscription options:
https://mail.python.org/mailman/listinfo/tutor
