Are strings pooled in Python?

Question

Does Python have a pool of all strings and are they (strings) singletons there?

More precise, in the following code, are one or two strings created in memory?

a = str(num)
b = str(num)

Boice

Just for reference, strings can't be singletons. A singleton is a class for which there can only be one instance, and that instance must be accessible globally. There can (hopefully) be many instances of the str class; therefore it's not a singleton. – Huldahuldah

The concept you're looking for is string interning: en.wikipedia.org/wiki/String_interning – Thurmond

@Huldahuldah Thank you for comment. I meant something like value-singleton (pool or string interning is right word for it - en.wikipedia.org/wiki/String_interning). – Boice

Answer 1

Strings are immutable in Python, so the implementation can decide whether to intern (that's a term often associated with C#, meaning that some strings are stored in a pool) strings or not.

In your example, you're dynamically creating strings. CPython does not always look into the pool to detect whether the string is already there - it also doesn't make sense because you first have to reserve memory in order to create the string, and then compare it to the pool content (inefficient for long strings).

But for strings of length 1, CPython does look into the pool (cf. "stringobject.c"):

static PyStringObject *characters[UCHAR_MAX + 1];

...

PyObject *
PyString_FromStringAndSize(const char *str, Py_ssize_t size)
{

...

    if (size == 1 && str != NULL &&
    (op = characters[*str & UCHAR_MAX]) != NULL)
    {
        #ifdef COUNT_ALLOCS
            one_strings++;
        #endif

        Py_INCREF(op);
        return (PyObject *)op;
    }

...

So:

a = str(num)
b = str(num)
print a is b # <-- this will print False in most cases (but try str(1) is str(1))

But when using constant strings directly in your code, CPython uses the same string instance:

a = "text"
b = "text"
print a is b # <-- this will print True

Confirmand

@Andidog: If CPython does not look into the pool to check if the string is already there, then why does print a is b print true when num is equal to 5? – Selenite

@Brian: Sorry, that was a bit inaccurate. Edited my answer to explain the way CPython implements that. – Confirmand

Good answer. The only detail I'd add is to note that Python does have intern() – Piscary

@keturn: Thanks, didn't even know about intern() yet. – Confirmand

str(1) is str(1) returns true in python 2.7 but false in python 3 – Pelmas

Answer 2

In general, strings are not interned in Python, but they do sometimes seem to be:

>>> str(5) is str(5)
True
>>> str(50) is str(50)
False

This isn't uncommon in Python, where common objects might be optimized in ways that unusual ones are not:

>>> int(5+0) is int(5+0)
True
>>> int(50+0) is int(50+0)
True
>>> int(500+0) is int(500+0)
False

And keep in mind, all of these sorts of details will differ between implementations of Python, and even between versions of the same implementation.

Thurmond

Answer 3

Strings are not interned in general. In your example two strings will be created (with the exception of values between 0 and 9). To test this we can use the is operator to see if the two strings are the same object:

>>> str(1056) is str(1056)
False

Coplanar

What about this: In [1]: x = str(5) In [2]: y = str(5) In [3]: id(x) Out[3]: 3077925280L In [4]: id(y) Out[4]: 3077925280L ? – Hysteric

gruszczy: That's a good question. This is a special case that only applies to the numbers 0 to 9. In general though, the statement is not true. I've clarified my answer. – Coplanar

0 to 9 is a specific case on a specific compiler (though admittedly, it's the compiler most people use). Other compilers may choose a different number of pre-defined strings. – Selenite

Answer 4

the constant of pool in python distinguish small integer pool and large integer pool, and the small integer pool is range in [-5, 257); and other integers in the large integer pool. In Cython, if defines a linked-list to store this data, where fetch the data becomes very convenient and fast.

# ifndef NSMALLPOSINTS
    # define NSMALLPOSINTS 257
# endif

# ifndef NSMALLNEGINTS
    # define NSMALLNEGINTS 5
# endif

# if NSMALLPOSINTS + NSMALLNEGINTS > 0
    static PyIntObject * small_ints[NSMALLPOSINTS + NSMALLNEGINTS];
# endif

BTW: the integer 257 maybe stranger; if two objects that they have the same values are in the same field, their address may be or not be the same, it depends on the context of process; whereas, if they are in the different fields, their addresses must be difference

and by the way, according to string type, cython also provide a constant pool that the length of string should be one, whereas it may be not the same object

a = str(11)
b = str(11)
print a == b      # True 
print a is b      # False

c = str("A")
d = str("A")   
print c == d    # True
print c is d    # True

aa = 12
bb = 12
print aa == bb    # True
print aa is bb    # True

cc = 333
dd = 333
print cc == dd    # True
print cc is dd    # False

comparing their addresses, is transparently getting the ahead solutions

Sucy

Please don't insert screenshots of the code. Copy the text instead. – Serosa