Unexpected reference count for unnamed integer literal - python

I understand that, in CPython 2.x and 3.x, some integers are singletons:
>>> a = 256; a is 256 # or any integer from -5 to 256
True
>>> a = 257; a is 257 # or any other integer outside the magic range
False
Accordingly, if I run sys.getrefcount on an integer in the range -5 to 256, I find that a lot of my imported packages have referenced that integer:
>>> sys.getrefcount(1)
1470
I also understand that sys.getrefcount returns 1 more than you might expect, because of its own reference to the argument:
>>> a = 257; sys.getrefcount(a)
2
What I don't get is this:
>>> sys.getrefcount(257)
3
Why 3, not 2? I could understand that I might have created a temporary variable in my own scope (count 1), and clearly sys.getrefcount would add another reference of its own to that (count 2) but where does the third one come from, and why didn't happen in the previous example? And more importantly: are there other contexts in which this might occur, leading to possible misinterpretation of sys.getrefcount outputs?
All of the above are replicable for me on 64-bit Python 2.7.12 and 3.5.1 by Anaconda, running on OSX, and also on a 32-bit Python 2.7.5 distribution running on Windows. However, on an older Python version (32-bit Python 2.5.4 on Windows), sys.getrefcount(257) returns 2 which is (to me) more expected.

You are running into an implementation detail here. The compiler can often cache immutable literal values:
>>> import dis
>>> compile("sys.getrefcount(257)", '', 'single').co_consts
(257, None)
>>> dis.dis(compile("sys.getrefcount(257)", '', 'single'))
1 0 LOAD_NAME 0 (sys)
2 LOAD_ATTR 1 (getrefcount)
4 LOAD_CONST 0 (257)
6 CALL_FUNCTION 1
8 PRINT_EXPR
10 LOAD_CONST 1 (None)
12 RETURN_VALUE
('single' is the mode used by the interactive interpreter).
We see 3 references here; one from the co_consts tuple on the code object, one on the stack (from the LOAD_CONST instruction), and one for the sys.getrefcount() method itself.

Related

'is' vs '==' in input and assignment [duplicate]

Something about the id of objects of type str (in python 2.7) puzzles me. The str type is immutable, so I would expect that once it is created, it will always have the same id. I believe I don't phrase myself so well, so instead I'll post an example of input and output sequence.
>>> id('so')
140614155123888
>>> id('so')
140614155123848
>>> id('so')
140614155123808
so in the meanwhile, it changes all the time. However, after having a variable pointing at that string, things change:
>>> so = 'so'
>>> id('so')
140614155123728
>>> so = 'so'
>>> id(so)
140614155123728
>>> not_so = 'so'
>>> id(not_so)
140614155123728
So it looks like it freezes the id, once a variable holds that value. Indeed, after del so and del not_so, the output of id('so') start changing again.
This is not the same behaviour as with (small) integers.
I know there is not real connection between immutability and having the same id; still, I am trying to figure out the source of this behaviour. I believe that someone whose familiar with python's internals would be less surprised than me, so I am trying to reach the same point...
Update
Trying the same with a different string gave different results...
>>> id('hello')
139978087896384
>>> id('hello')
139978087896384
>>> id('hello')
139978087896384
Now it is equal...
CPython does not promise to intern all strings by default, but in practice, a lot of places in the Python codebase do reuse already-created string objects. A lot of Python internals use (the C-equivalent of) the sys.intern() function call to explicitly intern Python strings, but unless you hit one of those special cases, two identical Python string literals will produce different strings.
Python is also free to reuse memory locations, and Python will also optimize immutable literals by storing them once, at compile time, with the bytecode in code objects. The Python REPL (interactive interpreter) also stores the most recent expression result in the _ name, which muddles up things some more.
As such, you will see the same id crop up from time to time.
Running just the line id(<string literal>) in the REPL goes through several steps:
The line is compiled, which includes creating a constant for the string object:
>>> compile("id('foo')", '<stdin>', 'single').co_consts
('foo', None)
This shows the stored constants with the compiled bytecode; in this case a string 'foo' and the None singleton. Simple expressions consisting of that produce an immutable value may be optimised at this stage, see the note on optimizers, below.
On execution, the string is loaded from the code constants, and id() returns the memory location. The resulting int value is bound to _, as well as printed:
>>> import dis
>>> dis.dis(compile("id('foo')", '<stdin>', 'single'))
1 0 LOAD_NAME 0 (id)
3 LOAD_CONST 0 ('foo')
6 CALL_FUNCTION 1
9 PRINT_EXPR
10 LOAD_CONST 1 (None)
13 RETURN_VALUE
The code object is not referenced by anything, reference count drops to 0 and the code object is deleted. As a consequence, so is the string object.
Python can then perhaps reuse the same memory location for a new string object, if you re-run the same code. This usually leads to the same memory address being printed if you repeat this code. This does depend on what else you do with your Python memory.
ID reuse is not predictable; if in the meantime the garbage collector runs to clear circular references, other memory could be freed and you'll get new memory addresses.
Next, the Python compiler will also intern any Python string stored as a constant, provided it looks enough like a valid identifier. The Python code object factory function PyCode_New will intern any string object that contains only ASCII letters, digits or underscores, by calling intern_string_constants(). This function recurses through the constants structures and for any string object v found there executes:
if (all_name_chars(v)) {
PyObject *w = v;
PyUnicode_InternInPlace(&v);
if (w != v) {
PyTuple_SET_ITEM(tuple, i, v);
modified = 1;
}
}
where all_name_chars() is documented as
/* all_name_chars(s): true iff s matches [a-zA-Z0-9_]* */
Since you created strings that fit that criterion, they are interned, which is why you see the same ID being used for the 'so' string in your second test: as long as a reference to the interned version survives, interning will cause future 'so' literals to reuse the interned string object, even in new code blocks and bound to different identifiers. In your first test, you don't save a reference to the string, so the interned strings are discarded before they can be reused.
Incidentally, your new name so = 'so' binds a string to a name that contains the same characters. In other words, you are creating a global whose name and value are equal. As Python interns both identifiers and qualifying constants, you end up using the same string object for both the identifier and its value:
>>> compile("so = 'so'", '<stdin>', 'single').co_names[0] is compile("so = 'so'", '<stdin>', 'single').co_consts[0]
True
If you create strings that are either not code object constants, or contain characters outside of the letters + numbers + underscore range, you'll see the id() value not being reused:
>>> some_var = 'Look ma, spaces and punctuation!'
>>> some_other_var = 'Look ma, spaces and punctuation!'
>>> id(some_var)
4493058384
>>> id(some_other_var)
4493058456
>>> foo = 'Concatenating_' + 'also_helps_if_long_enough'
>>> bar = 'Concatenating_' + 'also_helps_if_long_enough'
>>> foo is bar
False
>>> foo == bar
True
The Python compiler either uses the peephole optimizer (Python versions < 3.7) or the more capable AST optimizer (3.7 and newer) to pre-calculate (fold) the results of simple expressions involving constants. The peepholder limits it's output to a sequence of length 20 or less (to prevent bloating code objects and memory use), while the AST optimizer uses a separate limit for strings of 4096 characters. This means that concatenating shorter strings consisting only of name characters can still lead to interned strings if the resulting string fits within the optimizer limits of your current Python version.
E.g. on Python 3.7, 'foo' * 20 will result in a single interned string, because constant folding turns this into a single value, while on Python 3.6 or older only 'foo' * 6 would be folded:
>>> import dis, sys
>>> sys.version_info
sys.version_info(major=3, minor=7, micro=4, releaselevel='final', serial=0)
>>> dis.dis("'foo' * 20")
1 0 LOAD_CONST 0 ('foofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoo')
2 RETURN_VALUE
and
>>> dis.dis("'foo' * 6")
1 0 LOAD_CONST 2 ('foofoofoofoofoofoo')
2 RETURN_VALUE
>>> dis.dis("'foo' * 7")
1 0 LOAD_CONST 0 ('foo')
2 LOAD_CONST 1 (7)
4 BINARY_MULTIPLY
6 RETURN_VALUE
This behavior is specific to the Python interactive shell. If I put the following in a .py file:
print id('so')
print id('so')
print id('so')
and execute it, I receive the following output:
2888960
2888960
2888960
In CPython, a string literal is treated as a constant, which we can see in the bytecode of the snippet above:
2 0 LOAD_GLOBAL 0 (id)
3 LOAD_CONST 1 ('so')
6 CALL_FUNCTION 1
9 PRINT_ITEM
10 PRINT_NEWLINE
3 11 LOAD_GLOBAL 0 (id)
14 LOAD_CONST 1 ('so')
17 CALL_FUNCTION 1
20 PRINT_ITEM
21 PRINT_NEWLINE
4 22 LOAD_GLOBAL 0 (id)
25 LOAD_CONST 1 ('so')
28 CALL_FUNCTION 1
31 PRINT_ITEM
32 PRINT_NEWLINE
33 LOAD_CONST 0 (None)
36 RETURN_VALUE
The same constant (i.e. the same string object) is loaded 3 times, so the IDs are the same.
In your first example a new instance of the string 'so' is created each time, hence different id.
In the second example you are binding the string to a variable and Python can then maintain a shared copy of the string.
A more simplified way to understand the behaviour is to check the following Data Types and Variables.
Section "A String Pecularity" illustrates your question using special characters as example.
So while Python is not guaranteed to intern strings, it will frequently reuse the same string, and is may mislead. It's important to know that you shouldn't check id or is for equality of strings.
To demonstrate this, one way I've discovered to force a new string in Python 2.6 at least:
>>> so = 'so'
>>> new_so = '{0}'.format(so)
>>> so is new_so
False
and here's a bit more Python exploration:
>>> id(so)
102596064
>>> id(new_so)
259679968
>>> so == new_so
True

Does python create an object for string constants in equality comparisons?

In python for comparisons like this, does python create a temporary object for the string constant "help" and then continue with the equality comparison ? The object would be GCed after some point.
s1 = "nohelp"
if s1 == "help":
# Blah Blah
String literals, like all Python constants, are created during compile time, when the source code is translated to byte code. And because all Python strings are immutable the interpreter can re-use the same string object if it encounters the same string literal in multiple places. It can even do that if the literal string is created via concatenation of literals, but not if the string is built by concatenating a string literal to an existing string object.
Here's a short demo that creates a few identical strings inside and outside of functions. It also dumps the disassembled byte code of one of the functions.
from __future__ import print_function
from dis import dis
def f1(s):
a = "help"
print('f1', id(s), id(a))
return s > a
def f2(s):
a = "help"
print('f2', id(s), id(a))
return s > a
a = "help"
print(id(a))
print(f1("he" + "lp"))
b = "h"
print(f2(b + "elp"))
print("\nf1")
dis(f1)
typical output on a 32 bit machine running Python 2.6.6
3073880672
f1 3073880672 3073880672
False
f2 3073636576 3073880672
False
f1
26 0 LOAD_CONST 1 ('help')
3 STORE_FAST 1 (a)
27 6 LOAD_GLOBAL 0 (print)
9 LOAD_CONST 2 ('f1')
12 LOAD_GLOBAL 1 (id)
15 LOAD_FAST 0 (s)
18 CALL_FUNCTION 1
21 LOAD_GLOBAL 1 (id)
24 LOAD_FAST 1 (a)
27 CALL_FUNCTION 1
30 CALL_FUNCTION 3
33 POP_TOP
28 34 LOAD_FAST 0 (s)
37 LOAD_FAST 1 (a)
40 COMPARE_OP 4 (>)
43 RETURN_VALUE
Note that the ids of all the "help" strings are identical, apart from the one constructed with b + "elp".
(BTW, Python will concatenate adjacent string literals, so instead of writing "he" + "lp" I could've written "he" "lp", or even "he""lp").
The string literals themselves are not freed until the process is cleaning itself up at termination, however a string like b would be GC'ed if it went out of scope.
Note that in CPython (standard Python) when objects are GC'ed their memory is returned to Python's allocation system for recycling, not to the OS. Python does return unneeded memory to the OS, but only in special circumstances. See Releasing memory in Python and Why doesn't memory get released to system after large queries (or series of queries) in django?
Another question that discusses this topic: Why strings object are cached in python

Python: Why do certain strings match using "is" and others do not? [duplicate]

Something about the id of objects of type str (in python 2.7) puzzles me. The str type is immutable, so I would expect that once it is created, it will always have the same id. I believe I don't phrase myself so well, so instead I'll post an example of input and output sequence.
>>> id('so')
140614155123888
>>> id('so')
140614155123848
>>> id('so')
140614155123808
so in the meanwhile, it changes all the time. However, after having a variable pointing at that string, things change:
>>> so = 'so'
>>> id('so')
140614155123728
>>> so = 'so'
>>> id(so)
140614155123728
>>> not_so = 'so'
>>> id(not_so)
140614155123728
So it looks like it freezes the id, once a variable holds that value. Indeed, after del so and del not_so, the output of id('so') start changing again.
This is not the same behaviour as with (small) integers.
I know there is not real connection between immutability and having the same id; still, I am trying to figure out the source of this behaviour. I believe that someone whose familiar with python's internals would be less surprised than me, so I am trying to reach the same point...
Update
Trying the same with a different string gave different results...
>>> id('hello')
139978087896384
>>> id('hello')
139978087896384
>>> id('hello')
139978087896384
Now it is equal...
CPython does not promise to intern all strings by default, but in practice, a lot of places in the Python codebase do reuse already-created string objects. A lot of Python internals use (the C-equivalent of) the sys.intern() function call to explicitly intern Python strings, but unless you hit one of those special cases, two identical Python string literals will produce different strings.
Python is also free to reuse memory locations, and Python will also optimize immutable literals by storing them once, at compile time, with the bytecode in code objects. The Python REPL (interactive interpreter) also stores the most recent expression result in the _ name, which muddles up things some more.
As such, you will see the same id crop up from time to time.
Running just the line id(<string literal>) in the REPL goes through several steps:
The line is compiled, which includes creating a constant for the string object:
>>> compile("id('foo')", '<stdin>', 'single').co_consts
('foo', None)
This shows the stored constants with the compiled bytecode; in this case a string 'foo' and the None singleton. Simple expressions consisting of that produce an immutable value may be optimised at this stage, see the note on optimizers, below.
On execution, the string is loaded from the code constants, and id() returns the memory location. The resulting int value is bound to _, as well as printed:
>>> import dis
>>> dis.dis(compile("id('foo')", '<stdin>', 'single'))
1 0 LOAD_NAME 0 (id)
3 LOAD_CONST 0 ('foo')
6 CALL_FUNCTION 1
9 PRINT_EXPR
10 LOAD_CONST 1 (None)
13 RETURN_VALUE
The code object is not referenced by anything, reference count drops to 0 and the code object is deleted. As a consequence, so is the string object.
Python can then perhaps reuse the same memory location for a new string object, if you re-run the same code. This usually leads to the same memory address being printed if you repeat this code. This does depend on what else you do with your Python memory.
ID reuse is not predictable; if in the meantime the garbage collector runs to clear circular references, other memory could be freed and you'll get new memory addresses.
Next, the Python compiler will also intern any Python string stored as a constant, provided it looks enough like a valid identifier. The Python code object factory function PyCode_New will intern any string object that contains only ASCII letters, digits or underscores, by calling intern_string_constants(). This function recurses through the constants structures and for any string object v found there executes:
if (all_name_chars(v)) {
PyObject *w = v;
PyUnicode_InternInPlace(&v);
if (w != v) {
PyTuple_SET_ITEM(tuple, i, v);
modified = 1;
}
}
where all_name_chars() is documented as
/* all_name_chars(s): true iff s matches [a-zA-Z0-9_]* */
Since you created strings that fit that criterion, they are interned, which is why you see the same ID being used for the 'so' string in your second test: as long as a reference to the interned version survives, interning will cause future 'so' literals to reuse the interned string object, even in new code blocks and bound to different identifiers. In your first test, you don't save a reference to the string, so the interned strings are discarded before they can be reused.
Incidentally, your new name so = 'so' binds a string to a name that contains the same characters. In other words, you are creating a global whose name and value are equal. As Python interns both identifiers and qualifying constants, you end up using the same string object for both the identifier and its value:
>>> compile("so = 'so'", '<stdin>', 'single').co_names[0] is compile("so = 'so'", '<stdin>', 'single').co_consts[0]
True
If you create strings that are either not code object constants, or contain characters outside of the letters + numbers + underscore range, you'll see the id() value not being reused:
>>> some_var = 'Look ma, spaces and punctuation!'
>>> some_other_var = 'Look ma, spaces and punctuation!'
>>> id(some_var)
4493058384
>>> id(some_other_var)
4493058456
>>> foo = 'Concatenating_' + 'also_helps_if_long_enough'
>>> bar = 'Concatenating_' + 'also_helps_if_long_enough'
>>> foo is bar
False
>>> foo == bar
True
The Python compiler either uses the peephole optimizer (Python versions < 3.7) or the more capable AST optimizer (3.7 and newer) to pre-calculate (fold) the results of simple expressions involving constants. The peepholder limits it's output to a sequence of length 20 or less (to prevent bloating code objects and memory use), while the AST optimizer uses a separate limit for strings of 4096 characters. This means that concatenating shorter strings consisting only of name characters can still lead to interned strings if the resulting string fits within the optimizer limits of your current Python version.
E.g. on Python 3.7, 'foo' * 20 will result in a single interned string, because constant folding turns this into a single value, while on Python 3.6 or older only 'foo' * 6 would be folded:
>>> import dis, sys
>>> sys.version_info
sys.version_info(major=3, minor=7, micro=4, releaselevel='final', serial=0)
>>> dis.dis("'foo' * 20")
1 0 LOAD_CONST 0 ('foofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoo')
2 RETURN_VALUE
and
>>> dis.dis("'foo' * 6")
1 0 LOAD_CONST 2 ('foofoofoofoofoofoo')
2 RETURN_VALUE
>>> dis.dis("'foo' * 7")
1 0 LOAD_CONST 0 ('foo')
2 LOAD_CONST 1 (7)
4 BINARY_MULTIPLY
6 RETURN_VALUE
This behavior is specific to the Python interactive shell. If I put the following in a .py file:
print id('so')
print id('so')
print id('so')
and execute it, I receive the following output:
2888960
2888960
2888960
In CPython, a string literal is treated as a constant, which we can see in the bytecode of the snippet above:
2 0 LOAD_GLOBAL 0 (id)
3 LOAD_CONST 1 ('so')
6 CALL_FUNCTION 1
9 PRINT_ITEM
10 PRINT_NEWLINE
3 11 LOAD_GLOBAL 0 (id)
14 LOAD_CONST 1 ('so')
17 CALL_FUNCTION 1
20 PRINT_ITEM
21 PRINT_NEWLINE
4 22 LOAD_GLOBAL 0 (id)
25 LOAD_CONST 1 ('so')
28 CALL_FUNCTION 1
31 PRINT_ITEM
32 PRINT_NEWLINE
33 LOAD_CONST 0 (None)
36 RETURN_VALUE
The same constant (i.e. the same string object) is loaded 3 times, so the IDs are the same.
In your first example a new instance of the string 'so' is created each time, hence different id.
In the second example you are binding the string to a variable and Python can then maintain a shared copy of the string.
A more simplified way to understand the behaviour is to check the following Data Types and Variables.
Section "A String Pecularity" illustrates your question using special characters as example.
So while Python is not guaranteed to intern strings, it will frequently reuse the same string, and is may mislead. It's important to know that you shouldn't check id or is for equality of strings.
To demonstrate this, one way I've discovered to force a new string in Python 2.6 at least:
>>> so = 'so'
>>> new_so = '{0}'.format(so)
>>> so is new_so
False
and here's a bit more Python exploration:
>>> id(so)
102596064
>>> id(new_so)
259679968
>>> so == new_so
True

Why doesn't 1 + 1 use BINARY_ADD?

I do this:
>>> dis.dis(lambda: 1 + 1)
0 LOAD_CONST 2 (2)
3 RETURN_VALUE
I was expecting a BINARY_ADD opcode to perform the addition. How was the sum computed?
This is the work of Python's peephole optimizer. It evaluates simple operations with only constants during the compile time itself and stores the result as a constant in the generated bytecode.
Quoting from the Python 2.7.9 Source code,
/* Fold binary ops on constants.
LOAD_CONST c1 LOAD_CONST c2 BINOP --> LOAD_CONST binop(c1,c2) */
case BINARY_POWER:
case BINARY_MULTIPLY:
case BINARY_TRUE_DIVIDE:
case BINARY_FLOOR_DIVIDE:
case BINARY_MODULO:
case BINARY_ADD:
case BINARY_SUBTRACT:
case BINARY_SUBSCR:
case BINARY_LSHIFT:
case BINARY_RSHIFT:
case BINARY_AND:
case BINARY_XOR:
case BINARY_OR:
if (lastlc >= 2 &&
ISBASICBLOCK(blocks, i-6, 7) &&
fold_binops_on_constants(&codestr[i-6], consts)) {
i -= 2;
assert(codestr[i] == LOAD_CONST);
cumlc = 1;
}
break;
Basically, it looks for instructions like this
LOAD_CONST c1
LOAD_CONST c2
BINARY_OPERATION
and evaluates that and replaces those instructions with the result and a LOAD_CONST instruction. Quoting the comment in the fold_binops_on_constants function,
/* Replace LOAD_CONST c1. LOAD_CONST c2 BINOP
with LOAD_CONST binop(c1,c2)
The consts table must still be in list form so that the
new constant can be appended.
Called with codestr pointing to the first LOAD_CONST.
Abandons the transformation if the folding fails (i.e. 1+'a').
If the new constant is a sequence, only folds when the size
is below a threshold value. That keeps pyc files from
becoming large in the presence of code like: (None,)*1000.
*/
The actual evaluation of this particular code happens in this block,
case BINARY_ADD:
newconst = PyNumber_Add(v, w);
break;
The Python interpreter interprets from the inside out, that is, it reads the 1 + 1 evaluates it to 2, then creates a function object that returns the constant 2 (notice the order here!). Finally, the dis function evaluates. the newly created lambda function object, which simply returns a 2.
Thus, the 1+1 has already been computed when the lambda function object is created, and the dis.dis() function knows nothing about the addition that took place when the interpreter read 1+1 and evaluated it to 2.
If you do something like:
>>> dis.dis(lambda: x + 1)
1 0 LOAD_GLOBAL 0 (x)
3 LOAD_CONST 1 (1)
6 BINARY_ADD
7 RETURN_VALUE
You'll notice that a BINARY_ADD instruction is used, since x + 1 can't be further simplified by itself.

python string with whitespace doesn't point to same memory address [duplicate]

Something about the id of objects of type str (in python 2.7) puzzles me. The str type is immutable, so I would expect that once it is created, it will always have the same id. I believe I don't phrase myself so well, so instead I'll post an example of input and output sequence.
>>> id('so')
140614155123888
>>> id('so')
140614155123848
>>> id('so')
140614155123808
so in the meanwhile, it changes all the time. However, after having a variable pointing at that string, things change:
>>> so = 'so'
>>> id('so')
140614155123728
>>> so = 'so'
>>> id(so)
140614155123728
>>> not_so = 'so'
>>> id(not_so)
140614155123728
So it looks like it freezes the id, once a variable holds that value. Indeed, after del so and del not_so, the output of id('so') start changing again.
This is not the same behaviour as with (small) integers.
I know there is not real connection between immutability and having the same id; still, I am trying to figure out the source of this behaviour. I believe that someone whose familiar with python's internals would be less surprised than me, so I am trying to reach the same point...
Update
Trying the same with a different string gave different results...
>>> id('hello')
139978087896384
>>> id('hello')
139978087896384
>>> id('hello')
139978087896384
Now it is equal...
CPython does not promise to intern all strings by default, but in practice, a lot of places in the Python codebase do reuse already-created string objects. A lot of Python internals use (the C-equivalent of) the sys.intern() function call to explicitly intern Python strings, but unless you hit one of those special cases, two identical Python string literals will produce different strings.
Python is also free to reuse memory locations, and Python will also optimize immutable literals by storing them once, at compile time, with the bytecode in code objects. The Python REPL (interactive interpreter) also stores the most recent expression result in the _ name, which muddles up things some more.
As such, you will see the same id crop up from time to time.
Running just the line id(<string literal>) in the REPL goes through several steps:
The line is compiled, which includes creating a constant for the string object:
>>> compile("id('foo')", '<stdin>', 'single').co_consts
('foo', None)
This shows the stored constants with the compiled bytecode; in this case a string 'foo' and the None singleton. Simple expressions consisting of that produce an immutable value may be optimised at this stage, see the note on optimizers, below.
On execution, the string is loaded from the code constants, and id() returns the memory location. The resulting int value is bound to _, as well as printed:
>>> import dis
>>> dis.dis(compile("id('foo')", '<stdin>', 'single'))
1 0 LOAD_NAME 0 (id)
3 LOAD_CONST 0 ('foo')
6 CALL_FUNCTION 1
9 PRINT_EXPR
10 LOAD_CONST 1 (None)
13 RETURN_VALUE
The code object is not referenced by anything, reference count drops to 0 and the code object is deleted. As a consequence, so is the string object.
Python can then perhaps reuse the same memory location for a new string object, if you re-run the same code. This usually leads to the same memory address being printed if you repeat this code. This does depend on what else you do with your Python memory.
ID reuse is not predictable; if in the meantime the garbage collector runs to clear circular references, other memory could be freed and you'll get new memory addresses.
Next, the Python compiler will also intern any Python string stored as a constant, provided it looks enough like a valid identifier. The Python code object factory function PyCode_New will intern any string object that contains only ASCII letters, digits or underscores, by calling intern_string_constants(). This function recurses through the constants structures and for any string object v found there executes:
if (all_name_chars(v)) {
PyObject *w = v;
PyUnicode_InternInPlace(&v);
if (w != v) {
PyTuple_SET_ITEM(tuple, i, v);
modified = 1;
}
}
where all_name_chars() is documented as
/* all_name_chars(s): true iff s matches [a-zA-Z0-9_]* */
Since you created strings that fit that criterion, they are interned, which is why you see the same ID being used for the 'so' string in your second test: as long as a reference to the interned version survives, interning will cause future 'so' literals to reuse the interned string object, even in new code blocks and bound to different identifiers. In your first test, you don't save a reference to the string, so the interned strings are discarded before they can be reused.
Incidentally, your new name so = 'so' binds a string to a name that contains the same characters. In other words, you are creating a global whose name and value are equal. As Python interns both identifiers and qualifying constants, you end up using the same string object for both the identifier and its value:
>>> compile("so = 'so'", '<stdin>', 'single').co_names[0] is compile("so = 'so'", '<stdin>', 'single').co_consts[0]
True
If you create strings that are either not code object constants, or contain characters outside of the letters + numbers + underscore range, you'll see the id() value not being reused:
>>> some_var = 'Look ma, spaces and punctuation!'
>>> some_other_var = 'Look ma, spaces and punctuation!'
>>> id(some_var)
4493058384
>>> id(some_other_var)
4493058456
>>> foo = 'Concatenating_' + 'also_helps_if_long_enough'
>>> bar = 'Concatenating_' + 'also_helps_if_long_enough'
>>> foo is bar
False
>>> foo == bar
True
The Python compiler either uses the peephole optimizer (Python versions < 3.7) or the more capable AST optimizer (3.7 and newer) to pre-calculate (fold) the results of simple expressions involving constants. The peepholder limits it's output to a sequence of length 20 or less (to prevent bloating code objects and memory use), while the AST optimizer uses a separate limit for strings of 4096 characters. This means that concatenating shorter strings consisting only of name characters can still lead to interned strings if the resulting string fits within the optimizer limits of your current Python version.
E.g. on Python 3.7, 'foo' * 20 will result in a single interned string, because constant folding turns this into a single value, while on Python 3.6 or older only 'foo' * 6 would be folded:
>>> import dis, sys
>>> sys.version_info
sys.version_info(major=3, minor=7, micro=4, releaselevel='final', serial=0)
>>> dis.dis("'foo' * 20")
1 0 LOAD_CONST 0 ('foofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoofoo')
2 RETURN_VALUE
and
>>> dis.dis("'foo' * 6")
1 0 LOAD_CONST 2 ('foofoofoofoofoofoo')
2 RETURN_VALUE
>>> dis.dis("'foo' * 7")
1 0 LOAD_CONST 0 ('foo')
2 LOAD_CONST 1 (7)
4 BINARY_MULTIPLY
6 RETURN_VALUE
This behavior is specific to the Python interactive shell. If I put the following in a .py file:
print id('so')
print id('so')
print id('so')
and execute it, I receive the following output:
2888960
2888960
2888960
In CPython, a string literal is treated as a constant, which we can see in the bytecode of the snippet above:
2 0 LOAD_GLOBAL 0 (id)
3 LOAD_CONST 1 ('so')
6 CALL_FUNCTION 1
9 PRINT_ITEM
10 PRINT_NEWLINE
3 11 LOAD_GLOBAL 0 (id)
14 LOAD_CONST 1 ('so')
17 CALL_FUNCTION 1
20 PRINT_ITEM
21 PRINT_NEWLINE
4 22 LOAD_GLOBAL 0 (id)
25 LOAD_CONST 1 ('so')
28 CALL_FUNCTION 1
31 PRINT_ITEM
32 PRINT_NEWLINE
33 LOAD_CONST 0 (None)
36 RETURN_VALUE
The same constant (i.e. the same string object) is loaded 3 times, so the IDs are the same.
In your first example a new instance of the string 'so' is created each time, hence different id.
In the second example you are binding the string to a variable and Python can then maintain a shared copy of the string.
A more simplified way to understand the behaviour is to check the following Data Types and Variables.
Section "A String Pecularity" illustrates your question using special characters as example.
So while Python is not guaranteed to intern strings, it will frequently reuse the same string, and is may mislead. It's important to know that you shouldn't check id or is for equality of strings.
To demonstrate this, one way I've discovered to force a new string in Python 2.6 at least:
>>> so = 'so'
>>> new_so = '{0}'.format(so)
>>> so is new_so
False
and here's a bit more Python exploration:
>>> id(so)
102596064
>>> id(new_so)
259679968
>>> so == new_so
True

Categories