advanced listcomprehenions?

[Terry Reedy]

Duncan Booth wrote:
> Mark Wooding <mdw at distorted.org.uk> wrote:
> 
>> This is still inelegant, though.  We can glue the results mod 3 and 5
>> together using the Chinese Remainder Theorem and working mod 15
>> instead.  For example,
>>
>>   [['Fizz', 'FizzBuzz', False, None, 'Buzz'][(pow(i, 4, 15) + 1)%7] or
>>    str(i) for i in xrange(1, 101)]

The lookup table is a constant.  If made a tuple, it will be compiled as 
a constant (as least in 2.6, maybe 2.5).  In any case, it could (and to 
me should) be lifted out of the string comp.

>>
>> (A less mathematical approach would just use i%15 to index a table.  But
>> that's not interesting. ;-) )
>>
> 
> Ooh. Doesn't having 5 elements make you shudder? (Even though you did 
> change one to avoid a repeated value.) You have 4 options for output, so 
> for elegance that list should also have 4 elements:
> 
> [[str(i), 'FizzBuzz', 'Fizz', 'Buzz'][25/(pow(i, 4, 15) + 1)%4] for i in 
> xrange(1, 101)]
> 
> I feel it is even more elegant with the lookup table in its natural order:
> 
> [['Fizz', 'Buzz', 'FizzBuzz', str(i)][62/(pow(i, 4, 15) + 1)%4] for i in 
> xrange(1, 101)]

These make the lookup table variable, so it has to be recalculated for 
each i.

tjr