NDArray

NDArray is used for representing data in multiple dimensions. The NDArray structure allows you to set the rank, dimensions and values stored in memory. The values are stored as a contigious array. Strides are used to perform indexing in 2 or more dimensions.

Array Utilities

Method	Description
ndarray(int rank, int *shape)	Creates instance of ndarray data structure
reshape(ndarray_t *t, int rank, int *shape)	Reshape ndarray with new rank and shape dimensions
int nindex(ndarray_t *t, int *indices)	Gets one dimensional index value from array of indices
int *indices(ndarray_t *t, int index)	Get's indices associated with contigous index value
double nidx(ndarray_t *t, int *indices)	Get's value associated with indices
void nset(ndarray_t *t, int *indices, double value)	Set value at specific permutation of indices
void save_ndarray(ndarray_t *t, char *filename)	Save NDArray values to file in memory
ndarray_t *load_ndarray(char *filename)	Load NDArray values from saved file

ndarray(rank, shape)

---

• Parameters
  
  • rank: Integer value that represents amount of dimensions in value. Eg. 2d, 3d Arrays
  • shape : Array of integers that correspond to the dimensions of the value. Example values: (2, 3), (6, 6).

Notes: This method creates the NDArray using the ndarray_t structure from the core library. This does not work from C -> python (backwards compatible)

Example Code:

/* create 3 X 3 2d array */ 
ndarray_t *t = ndarray(2, (int[]){3, 3});
print_ndarray(t);

/* create 2 X 2 X 2 3d array */ 
ndarray_t *t1 = ndarray(3, (int[]){2, 2, 2});
print_ndarray(t1);

---

reshape(t,rank,shape)

---

• Parameters
  
  • t : Reference to instance of ndarray that will be reshaped
  • rank: Integer value that represents amount of dimensions in value. Eg. 2d, 3d Arrays
  • shape : Array of integers that correspond to the dimensions of the value. Example values: (2, 3), (6, 6).

Notes: This method requires you to provide a new rank and shape, even if the prior rank was the same.

Example Code:

/* create 3 X 3 2d array */ 
ndarray_t *t = ndarray(2, (int[]){3, 3});
print_ndarray(t);

/* reshape ndarray */
reshape(t, 3, (int[]){2, 2, 2})

---

nindex(t, indices)

---

• Parameters
  
  • t : Reference to instance of ndarray that will be reshaped
  • indices: Set of indices to map to one dimensional index. Length of indices is the rank of the value

Example Code:

/* create 3 X 3 2d array */ 
ndarray_t *t = ndarray(2, (int[]){3, 3});

/* example nindex usage for Unit tests */
int index = 0; 
for(int i = 0; i < 3; i++){
    for(int j = 0; j < 3; j++){
        int test_index =  nindex(t, (int[]){i, j});
        assert(test_index == index);    
        index += 1;     
    }
}

---

indices(t,index)

---

• Parameters
  
  • t : Reference to instance of ndarray
  • index: Integer value that maps to index of contiguous array

Notes: This method is a helper for flattening and performing transposes/contractions

Example Code:

/* create 3 X 3 2d array */ 
ndarray_t *t = ndarray(2, (int[]){3, 3});
print_ndarray(t);


/*  set values of ndarray   */ 
int index = 0; 
for(int i = 0; i < 3; i++){
    for(int j = 0; j < 3; j++){
        nset(t, (int[]){i, j}, index+1);
        index += 1;     
    }
}

/* validate indexs match */
index = 0;
for(int i = 0; i < 3; i++){
    for(int j = 0; j < 3; j++){
        int *idxs = indices(t, index); 
        assert(idxs[0] == i);
        assert(idxs[1] == j);  
        index += 1; 
    }
}

---

nidx(t,indices)

---

• Parameters
  
  • t : Reference to instance of ndarray
  • indices: Array of indices to map to value in ndarray

Example Code:

/* create 3 X 3 2d array */ 
ndarray_t *t = ndarray(2, (int[]){3, 3});

/*  set values of ndarray   */ 
int index = 0; 
for(int i = 0; i < 3; i++){
    for(int j = 0; j < 3; j++){
        nset(t, (int[]){i, j}, index+1);
        index += 1;     
    }
}

double value = nidx(t, (int[]){0, 0}) /* grab value at  (0, 0) */ 
double value1 = nidx(t, (int[]){0, 1}) /* grab value at  (0, 1) */ 

---

nset(t,indices, value)

---

• Parameters
  
  • t : Reference to instance of ndarray
  • indices: Array of indices to map to value in ndarray
  • value: Value we want to set in ndarray

Example Code:

/* create 3 X 3 2d array */ 
ndarray_t *t = ndarray(2, (int[]){3, 3});

/*  set values of ndarray   */ 
int index = 0; 
for(int i = 0; i < 3; i++){
    for(int j = 0; j < 3; j++){
        nset(t, (int[]){i, j}, index+1); /* set value */
        index += 1;     
    }
}

---

save_ndarray(t,filename)

---

• Parameters
  
  • t : Reference to instance of ndarray
  • filename: Path to save ndarray in memory

Example Code:

/* create 3 X 3 2d array */ 
ndarray_t *t = ndarray(2, (int[]){3, 3});

/*  set values of ndarray   */ 
int index = 0; 
for(int i = 0; i < 3; i++){
    for(int j = 0; j < 3; j++){
        nset(t, (int[]){i, j}, index+1); /* set value */
        index += 1;     
    }
}

/* save ndarray  to path */
save_ndarray(t, "path/to/file/values")

---

load_ndarray(filename)

---

• Parameters
  
  • filename: Path reference to saved ndarray

Note: The loaded ndarray will set your shape and dimensions for you, all you need is the filepath.

Example Code:

/* load saved ndarray */
ndarray_t *loaded = load_ndarray("path/to/file/values");

---

Operations

Method	Description
ndot(ndarray_t *a, ndarray_t *b)	Dot Product of two ndarray's
nadd(ndarray_t *t, ndarray_t *val)	Add two ndarray's
nsubtract(ndarray_t *t1, ndarray_t *t2)	Subtract two ndarray's
nscale_add(ndarray_t *a, ndarray_t *b)	Add A ndarray by vector
ntranspose(ndarray_t *t)	Tranpose of ndarray instance (rows as columns)
permute(ndarray_t *t, int *indice_order)	Contract NDArray and reorder indices

ndot(a, b)

---

• Parameters
  
  • a: First tensor to perform dot product
  • b: Second tensor to perform dot product

Note: Operation returns a new ndarray with the result of a and b.

Example Code:

/* add two 3 X 3 ndarrays */
ndarray_t *t1 = ndarray(2, (int[]){3, 3}); 
ndarray_t *t2 = ndarray(2, (int[]){3, 3}); 
ndarray_t *result = ndot(t1, t2);

---

nadd(a, b)

---

• Parameters
  
  • a: First tensor to add
  • b: Second tensor to add

Note: Operation returns a new ndarray with the result of a and b.

Example Code:

/* add two 3 X 3 ndarrays */
ndarray_t *t1 = ndarray(2, (int[]){3, 3}); 
ndarray_t *t2 = ndarray(2, (int[]){3, 3}); 
ndarray_t *result = nadd(t1, t2);

---

nsubtract(t1, t2)

---

• Parameters
  
  • t1: First tensor to subtract
  • t2: Second tensor to subtract

Note: Operation returns a new ndarray with the result of a and b.

Example Code:

/* add two 3 X 3 ndarrays */
ndarray_t *t1 = ndarray(2, (int[]){3, 3}); 
ndarray_t *t2 = ndarray(2, (int[]){3, 3}); 
ndarray_t *result = nsubtract(t1, t2);

---

nscale_add(a, b)

---

• Parameters
  
  • a: First tensor to scale add
  • b: Second tensor to scale add

Note: Operation returns a new ndarray with the result of a and b.

Example Code:

/* add two 3 X 3 ndarrays */
ndarray_t *t1 = ndarray(2, (int[]){3, 3}); 
ndarray_t *t2 = ndarray(2, (int[]){1, 3}); /* 1  X 3 */
ndarray_t *result = nscale_add(t1, t2);

---

ntranspose(t)

---

• Parameters
  
  • t: Instance of ndarray to perform transpose on

Note: Operation returns a new ndarray with the result of tranpose operation.

Example Code:

/* add two 3 X 3 ndarrays */
ndarray_t *t1 = ndarray(2, (int[]){3, 4}); 
ndarray_t *result = ntranspose(t1); /* 4 X 3 ndarray */

---

permute(t)

---

• Parameters
  
  • t: Instance of ndarray to perform permute/contraction on

Note: Operation returns a new ndarray with the result of contraction.

Example Code:

/* add two 3 X 3 ndarrays */
ndarray_t *t1 = ndarray(3, (int[]){3, 4, 5}); 
ndarray_t *result = permute(t1, (int[]){2, 1, 0}); /* 5 X 4 X 3  ndarray */

---