Is safe ( documented behaviour? ) to delete the domain of an iterator in execution

Question

I wanted to know if is safe ( documented behaviour? ) to delete the domain space of an iterator in execution in Python.

Consider the code:

import os
import sys

sampleSpace = [ x*x for x in range( 7 ) ]

print sampleSpace

for dx in sampleSpace:

    print str( dx )

    if dx == 1:

        del sampleSpace[ 1 ]
        del sampleSpace[ 3 ]

    elif dx == 25:

        del sampleSpace[ -1 ]

print sampleSpace

'sampleSpace' is what I call 'the domain space of an iterator' ( if there is a more appropriate word/phrase, lemme know ).

What I am doing is deleting values from it while the iterator 'dx' is running through it.

Here is what I expect from the code :

Iteration versus element being pointed to (*):

0: [*0, 1, 4, 9, 16, 25, 36]
1: [0, *1, 4, 9, 16, 25, 36] ( delete 2nd and 5th element after this iteration )
2: [0, 4, *9, 25, 36]
3: [0, 4, 9, *25, 36] ( delete -1th element after this iteration )
4: [0, 4, 9, 25*] ( as the iterator points to nothing/end of list, the loop terminates )

.. and here is what I get:

[0, 1, 4, 9, 16, 25, 36]
0
1
9
25
[0, 4, 9, 25]

As you can see - what I expect is what I get - which is contrary to the behaviour I have had from other languages in such a scenario.

Hence - I wanted to ask you if there is some rule like "the iterator becomes invalid if you mutate its space during iteration" in Python?

Is it safe ( documented behaviour? ) in Python to do stuff like this?

Answer 1

What do you mean by safe? Your code happens not to raise any errors, but it is a distinct possibility of course, consider this:

>>> a = range(3)
>>> for i in a:
    del a


Traceback (most recent call last):
  File "<pyshell#13>", line 2, in <module>
    del a
NameError: name 'a' is not defined
>>> a
[0, 1, 2]
>>> for i in a:
    del a[i+1]


Traceback (most recent call last):
  File "<pyshell#27>", line 2, in <module>
    del a[i+1]
IndexError: list assignment index out of range

It is not clear why would you want to do this, but there is no additional rules applicable to iterators. They're acting exactly as any other type would.

Answer 2

From the Python tutorial:

It is not safe to modify the sequence being iterated over in the loop (this can only happen for mutable sequence types, such as lists). If you need to modify the list you are iterating over (for example, to duplicate selected items) you must iterate over a copy. The slice notation makes this particularly convenient:

>>> for x in a[:]: # make a slice copy of the entire list
...    if len(x) > 6: a.insert(0, x)
...
>>> a
['defenestrate', 'cat', 'window', 'defenestrate']

Answer 3

Generally speaking no, it's not safe and you may get unpredictable behaviour. Iterators aren't required to behave in an specific way under these circumstances.

What's happening in your example is

# list is [0, 1, 4, 9, 16, 25, 36]

if dx == 1:
    # we're at index 1 when this is true
    del sampleSpace[ 1 ]
    # we've removed the item at index 1, and the iterator will move to the next valid position - still index 1, but in a mutated list. We got lucky in this case
    # the list now contains [0, 4, 9, 16, 25, 36]
    del sampleSpace[ 3 ]   
    # we remove the item at index 3 which is (now) value 16
    # the list now contains [0, 4, 9, 25, 36]
elif dx == 25:

    del sampleSpace[ -1 ]
    # we remove the final item, list now looks like
    # the list now contains [0, 4, 9, 25]