1. Conventions

These are general conventions that hold for all parts of Cytnx.

1.1. Function naming conventions

Generally, the function naming scheme in Cytnx follows the rules:

  1. If the function is acting on objects (taking object as arguments), they will start with the first letter being capical. Examples are the linalg functions, Contract etc…

    cytnx.linalg.Svd(A)
cytnx.linalg.Qr(A)
cytnx.linalg.Sum(A)

cytnx.Contract(A,B)

  2. If a function is a member function, or a generating function (such as zeros(), ones() …), then they usually start with a lower case letter, for example:

    A = cytnx.UniTensor.zeros([2,3,4])

A.permute(0,2,1)
B = A.contiguous()

  3. Objects in Cytnx always start with capical letters, for example:

    A = cytnx.UniTensor()
B = cytnx.Bond()
C = cytnx.Network()
D = cytnx.Tensor()

  4. Functions end with underscore indicate that the input will be changed. For member functions, this is an inplace operation

    A = cytnx.zeros([2,3,4])

A.contiguous_() # A gets changed
B = A.contiguous() # A is not changed, but return a copy B (see Tensor for further info)

A.permute_(0,2,1) # A gets changed
C = A.permute(0,2,1) # A is not changed but return a new B as A's permute

1.2. Everything is a reference

To provide a direct translation between the C++ API and Python, as well as to reduce the redundant memory allocation, all the objects (except Accessor and LinOp) in Cytnx are references, also in C++, just like in Python.

Let’s look at the following example. Consider the Tensor object in Cytnx

  • In Python:

1A = cytnx.Tensor([2,3])
2B = A
3
4print(B is A) # true

  • In C++:

1auto A = cytnx::Tensor({2, 3});
2auto B = A;
3
4cout << cytnx::is(B, A) << endl;

Output >>

True

Note

In Python, this is very straight forward. We implement the is clause in the Cytnx C++ API, such that C++ and Python can have exactly the same behavior. You can use is() to check if two objects are the same.

1.3. clone

In the case where a copy of an object is needed, you can use clone().

  • In Python:

1A = cytnx.Tensor([2,3]);
2B = A;
3C = A.clone();
4
5print(B is A)
6print(C is A)

  • In C++:

1auto A = cytnx::Tensor({2, 3});
2auto B = A;
3auto C = A.clone();
4
5cout << cytnx::is(B, A) << endl;
6cout << cytnx::is(C, A) << endl;

Output >>

True
False

Note

Here, we use Tensor as an example, but in general all the objects in Cytnx (except Accessor and LinOp) also follow the same behavior, and you can use is and clone as well.