# Chapter 5

## Linear Probing

When a collision occurs, we go through the hash table and find the next available slot.

## Quadratic Probing

Going through a hash table and finding the next available slot by adding a quadratic value to the current index.

$(H + c1*i + c2*i^2) \mod \textit{(table size)}$

Where $H$ is the hash value, $c1$ and $c2$ are constants, and $i$ is the number of times we've probed.

For instance:

| Index | Value |
|-------|-------|
| 0     | 20    |
| 1     | 41    |
| 2     | null  |
| 3     | null  |
| 4     | null  |

- $c1 = 2$
- $c2 = 4$

So running `insert(40)` would result in:

- $H = 40 \bmod 5 = 0$
- $i = 1$
- $(0 + (2 * 1) + (4 * 1^2)) \bmod 5$ -> $(0 + 2 + 4) \bmod 5 = 1$
- $i = 2$ - retrying because there's already a value at 1
- $(0 + (2 * 2) + (4 * 2^2)) \bmod 5$ -> $(0 + 4 + 16) \bmod 5 = 0$
- $i = 3$ - retrying because there's already a value at 0
- $(0 + (2 * 3) + (4 * 3^2)) \bmod 5$ -> $(0 + 6 + 36) \bmod 5 = 2$
- There is no value at 2, so we can insert 40 there.

| Index | Value |
|-------|-------|
| 0     | 20    |
| 1     | 41    |
| 2     | 40    |
| 3     | null  |
| 4     | null  |

And a script to do it automatically:

```python
import sys

size = int(sys.argv[-5])  # size of the table
num = int(sys.argv[-4])  # number to hash
c1 = int(sys.argv[-3])  # c1
c2 = int(sys.argv[-2])  # c2
i = int(sys.argv[-1])  # number of times we've probed

h = num % size  # hash value

print((h + (c1 * i) + (c2 * i * i)) % size)
```

## Double Hashing

Another collision resolution thing, where we use a second hash function to find the next available slot. The formula is:

$(h1(key) + i * h2(key)) \bmod (table size)$

Where $h1$ is the first hash function, $h2$ is the second hash function, and $i$ is the number of times we've probed.

For instance:

| Index | Value |
|-------|-------|
| 0     | 20    |
| 1     | 41    |
| 2     | null  |
| 3     | null  |
| 4     | null  |

Assuming:

- $h1() = h \bmod 5$
- $h2() = (i * h \bmod 3)$ - idk if there's any logic to this, but it's just an example

So running `insert(40)` would result in:

- $h1(40) = 40 \bmod 5 = 0$ - using the second function because there's already a value at 0
- $h2(40) = (1 * 40) \bmod 3 = 1$;  $i = 0 + (1 * 1) = 1$ - there's already a value at 1
- $h2(40) = (2 * 40) \bmod 3 = 2$;  $i = 0 + (2 * 2) = 4$ - there's no value at 4, so we can insert 40 there.

## Common Hash Functions

### Mid-Square hashing

1. Square the key
2. Take the middle $R$ digits - $R$ must be greater than or equal to $log_{10}(\text{table size})$
3. Use that as the hash value - if it's above the number of slots, we can use the modulo of the table size. e.g., if the table size is 100, and the hash value is 123, we can use 23.

#### Binary

Usually mid-square hashing is done using binary, since that's what computers work with and is faster. It works the same, but using base 2:

1. Square the key
2. Take the middle $R$ bits - $R$ must be greater than or equal to $log_2(\text{table size})$
3. Use that as the hash value - if it's above the number of slots, we can use the modulo of the table size.

### Multiplicative string hashing

Multiplicative string hashing in Python-like pseudocode:

```py
multiplicativeStringHash(string key) {
   string_hash = initial_value   # using Bernstein's hash, initial_value is 5381
   hash_multiplier = 33  # using Bernstein's hash, hash_multiplier is 33

   for character in key {
      string_hash *= hash_multiplier
      strChar += ascii(character)  # returns the ASCII number for the character
   }

   return string_hash % table_size
}
```

Bernstein's hash uses an initial value of 5381 and a hash multiplier of 33, and works well for hashing short English strings.