[Python-Dev] performance of {} versus dict()

[Terry Reedy]

On 11/14/2012 4:12 AM, Chris Withers wrote:
To somewhat paraphrase: '''
I prefer 'dict(a=1,b=2,c=3,d=4,e=5,f=6,g=7)' to
"{'a':1,'b':2,'c':3,'d':4,'e':5,'f':6,'g':7}".
I am sad that the former takes +-2 times as long to run (in 2.7).
Is the difference about the same in 3.x?
What can we do to speed up the former case?
'''

My responses, trying not to duplicate others.
1. Visual preference depends on the viewer. I prefer the dict display, 
perhaps because I am more accustomed to it.

2. The two types of expressions have overlapping but distinct use cases. 
This differences include that dict can be wrapped or replaced, while 
displays cannot.

3. a) 3.x has dict comprehensions. How do they stack up? b) If one were 
really initializing multiple dicts with the same starting items, and one 
were really concerned about speed, should one calculate the common base 
dict just once and then copy? Win7 64 with 3.3.0:

 >>> repeat("dict(a=1, b=2, c=3, d=4, e=5)")
[0.6200045004915467, 0.6212762582470646, 0.6114683222573376]
 >>> repeat("{'a': 1, 'b': 2, 'c': 3, 'd': 4, 'e': 5}")
[0.27170026972208233, 0.2594874604131968, 0.25977058559879584]
 >>> repeat("d.copy()", "d={'a': 1, 'b': 2, 'c': 3, 'd': 4, 'e': 5}")
[0.25768296004457625, 0.243041299387869, 0.2421860830290825]

 >>> repeat("{str(i):i for i in range(10)}")
[4.914327732926495, 4.874041570524014, 4.871596119002334]
 >>> repeat("{'0':0, '1':1, '2':2, '3':3, '4':4, '5':5, '6':6, '7':7, 
'8':8, '9':9}")
[0.5207065648769458, 0.5000415004344632, 0.49980294978922757]
 >>> repeat("d.copy()", "d={'0':0, '1':1, '2':2, '3':3, '4':4, '5':5, 
'6':6, '7':7, '8':8, '9':9}")
[0.571671864980317, 0.5516194699132484, 0.5514937389677925]


Assuming no overlooked errors in the above...
a) Dict comprehensions are much slower than calls, which makes the calls 
look good by comparison. b) Copying is not worthwhile.

4. There are about 3000 issues on the tracker. Nearly all are worth more 
attention than this ;-).

-- 
Terry Jan Reedy