import numpy as np
v1 = np.random.randint(1, 7, 7)
a1 = np.random.randint(1, 10, (3, 3))
print(v1, "\n\n", a1)

for i in range(0, 7, 3):
  print(f"[{i}]: {v1[i]}", end='; ')
# negative indexes work the same way as well
print("\n\n")
for i in range(1, 6, 2):
  print(f"[{-i}]: {v1[-i]}", end='; ')

for i in range(0, 3):
    print(f"({i},{i}): {a1[i, i]}", end="; ")
# and with negative indices
print("\n\n")
for i in range(0, 3):
    print(f"({i},{-(i+1)}): {a1[i, -(i + 1)]}", end="; ")

for i in range(0, 3):
    a1[i, -(i + 1)] = 42.6
print(a1)

# To make things clearer let's start with a new vector
v2 = np.arange(10)
print(v2)

# first 3 elements
print(v2[:3])
# elements with index 6 or greater
print(v2[6:])
# values at indices 4 to 6 inclusive, recall 'stop' is not included
print(v2[4:7])
# how about a negative step value
print(v2[6:3:-1])
# every other element starting at index 1
print(v2[1::2])
# as with Python lists that negative step value can be used to reverser the vector
print(v2[::-1])

a2 = np.random.randint(1, 10, (4, 5))
print(a2)

# let's get the 2x3 array out of the middle
a_mid = a2[1:3, 1:4]
print(a_mid)

# how about all rows, every other column
a_23 = a2[:4, ::2]
print(a_23)

# reversing can also work, here we reverse the rows and the columns
a_23r = a_23[::-1, :3]
print(a_23r)
# but we could just as easily reversed the rows and the columns together
print()
a_23r2 = a_23[::-1, ::-1]
print(a_23r2)

print(f"first column of a2: {a2[:, 0]}")
# or the first row
print(f"\nfirst row of a2: {a2[0, :]}")
# for rows, we can actually drop the empty slice syntax
print(f"\nfirst row of a2: {a2[0]}")

a_2_2 = a2[1:3, 2:4]
print(a_2_2)
# let's change one of the elements
a_2_2[1,1] = 99
print(f"\na_2_2: {a_2_2}")
print(f"\na2: {a2}")

# getting a copy rather than a view
a_2_2_cp = a2[1:3, 2:4].copy()
print(a_2_2_cp)
# let's change one of the elements
a_2_2_cp[1,1] = 8
print(f"\na_2_2_cp: {a_2_2_cp}")
print(f"\na2: {a2}")

# let's reshape a new vector into a 3x3 array
v3 = np.arange(1, 19, 2)
print(f"v3: {v3}")
v3_3 = v3.reshape((3, 3))
print(f"\nv3_3:\n{v3_3}")
print(f"\nv3.size: {v3.size}; v3_3.size: {v3_3.size}\n")
# do note, v3.size must equal v3_3.size
v3_4 = v3.reshape((3, 4))

# how about v1 as the first column of an array (column vector)
v1_rv = v1.reshape(1, v1.size)
print(f"\nv1: {v1}")
print(f"\nv1_rv: {v1_rv}")
# how about the first column of an array (column vector)
v1_cv = v1.reshape(v1.size, 1)
print(f"\nv1_cv:\n{v1_cv}")
# there is also a keyword, newaxis, that works for the above two cases
v1_rv_2 = v1[np.newaxis, :]
print(f"\nv1_rv_2: {v1_rv_2}")
v1_cv_2 = v1[:, np.newaxis]
print(f"\nv1_cv_2:\n{v1_cv_2}")

# let's look at combining arrays, concatenation
# three methods available: np.concatenate, np.vstack, np.hstack
v7 = np.array([1, 2, 3])
v8 = np.array([4, 5, 6])
v9 = np.array([42, 42, 42])
# first argument is a tuple or list of arrays
print(np.concatenate([v7, v8]))
print(np.concatenate([v7, v8, v9]))

# works with two-dimensional arrays
a7 = np.array([v7, v8])
print("\n", a7)
# concatenate along the row axis
print("\n", np.concatenate([a7, a7]))
# or along the column axis
print("\n", np.concatenate([a7, a7], axis=1))

# when working with arrays of varying dimensions
# you may prefer to use one of the other methods
# N.B. shape and size are relevant
a8 = np.array([7, 8, 9])
# verically, i.e. row order
print(np.vstack([a8, a7]))
# horizontally, i.e. column order,
a9 = a8[:2].reshape((2, 1))
print("\n", np.hstack([a7, a9]))

# now let's look at splitting arrays
# again three methods: np.split, np.hsplit, np.vsplit
# we pass a list of split points to each,
# nbr of resulting arrays is nbr of split points + 1
v10 = np.concatenate([v7, v8, v9])
v11, v12, v13 = np.split(v10, [4, 7])
print("\n", v11,"\n", v12, "\n", v13)

# now the other two functions
matrix = np.arange(16).reshape((4, 4))
print(matrix)
m_up, m_dn = np.vsplit(matrix,[2])
m_lf, m_rt = np.hsplit(matrix,[2])
print("\n\nm_up:\n", m_up,"\n\nm_dn:\n", m_dn, "\n\nm_lf:\n", m_lf, "\n\nm_rt:\n", m_rt)