"Python" Memory analysis _list object Memory footprint Analysis

"Python" Memory Analysis _ _list and array in memory growth mode

After the list declaration structure is broadly divided into 3 parts, the variable name--list object (structural data + pointer array)--list content , where the ID represents the position of the list object,

V reference variable name, v[:] refers to the list object, which is also set up for the other Python sequence structure, which is supported by the following demonstration available IDs,

A=b, A and B point to the same list object

a=b[:], A's list object and B's List object point to the same list content

Q1: Whether the element storage address is contiguous

The memory address of the contents (structure 3) stored by the list object is tested first.

In [1]: a=[1,2,3, ' A ', ' B ', ' C ', ' de ', [4,5]]in [2]: ID (a) out[2]: 139717112576840In [3]: For I in a:   ...:     print (ID (i) )   ...:     1397172387699201397172387699521397172387699841397172398341921397172400774801397172405238881397171952811041 39717112078024In [4]: For I in a[6]:   ...:     print (ID (i))   ...:     139717240220952139717240202048In [5]: for I in a[7]:   ...:     print (ID (i))   ...:     139717238770016139717238770048

Then look at the relative address,

In [6]: For I in a:   ...:     print (ID (i) -139717238769920)   ...:     03264106427213075601753968-43488816-126691896in [7]: For I in a[6]:   ...:     print (ID (i)-139717238769920)   ...:     14510321432128In [8]: For I in a[7]: ...:     print (ID (i) -139717238769920)   ...:     96128

Visible, for the list object, its element content is not necessarily linear storage, but due to memory allocation problem, there will be a linear storage illusion, when the element appears in the container or relative to the previous element type changes, the memory space will be no longer contiguous .

Q2:list object address and element address are contiguous

In fact, Q1 has answered this question, after all, the element address itself is not continuous, but we still tested a bit,

In []: ID (a[0])-id (a) out[22]: 126193080

Very far, and we analyze the source, the list object body is an array of pointers, that is, the ID (a) refers to the location of the body is a pointer to the location of the array of pointers, of course, as well as the auxiliary object header information (python in a few common "black box" list).

Q3:list object (without element) memory consumption analysis

in [+]: sys.getsizeof ([120In, ' A ', ' B ', ' C ', ' de ']) out[16]: [+]: sys.getsizeof ([All-in-all, ' a ', ' B ', ' C ']) out[17]: 112In [+]: sys.getsizeof ([104, ' A ', ' B ')) out[18]:

Visible, list each object occupies 8 bytes of 32 bits of space, we look at the slices,

In []: Sys.getsizeof (A[:3]) out[20]: 88In [96In]: sys.getsizeof (A[:4]) out[21]: [max]: sys.getsizeof (A[3:4]) out[23]: 72In []: sys.getsizeof (A[3:5]) out[24]: 80

The Slice object is also 8 bytes per element, but the slice is also a list object, even if it is cut from the middle (without cutting the head), it also contains the storage footprint of the header information.

Note

1, [0] and [: 1] Different

in [+]: a[0]out[30]: 1In []: a[:1]out[31]: [1]

2. Empty list occupies space

in [+]: sys.getsizeof ([]) out[32]: 64

"Python" Memory analysis _list object Memory footprint Analysis