I hope we all have some idea almost what binary search is and does. But I’thou not going to explain the algorithm stride by step, rather I’m going to give an insight into how binary search works and can be used.
Check out: geeksforgeeks.org/binarysearch if you lot’re unaware of Binary Search.
Given a sorted array, we find the centermost element and check the element with the key. If the middlemost element is equal to key, we’ve found the cardinal. If the middlemost element is greater than the key, we search on the left onehalf of the centremost element, else nosotros search on the right half.
Hither’south an iterative code for Binary Search in Coffee
Discover that in line 6, we use
int mid = (depression + loftier) / ii;
Only computing mid this style is ineffective. Why? Permit’s take an instance.Let us take integers from an integer depression to an integer high (both included).
Discover the mids computed by formulae in 3rd and fourth columns.

for depression = iii and high = 11, the number of elements (#elements) = 9So in that location is only 1 mid, i.e., 7* Both formulae have computed the mid correctly

for low = 3 and high = ten, #elements = 8So in that location are ii mids, halfdozen (lower mid) and 7 (higher mid)* Both formulae have computed the lower mid correctly

for low = eleven and high = 3, #elements = 9So there is but 1 mid, i.eastward., 7* Both formulae have computed the mid correctly

for low = 10 and high = 3, #elements = 8So there are 2 mids, 7 (lower mid) and 8 (college mid)* The formula
(low + loftier) / 2
has failed to compute the lower mid correctly but the other formula has computed information technology correctly.
So, nosotros should ever use the below formula to compute lower mid as it is much more reliable:
int mid = depression + ((high – depression)/2);
When #elements = odd, we have only 1 mid. So we can utilise the above formula to compute mid.When #elements = even, the in a higher place formula only gives the lower mid. The higher mid can be computed past the beneath formula:
int mid2 = low + ((high – low + 1) / 2);
Now, the interesting part…
Let’s become back to the iterative code for Binary Search. Discover three things.1. How we are moving low and high2. How we are computing mid3. The condition in while loop
The dazzler of Binary Search lies in these 3 things alone. Allow usa explore this further.
For a simple binary search where we just have to find the chemical element in the assortment,we employ the following:1.
depression = mid  i
for moving
low
two.
loftier = mid + 1
for moving
high
3.
mid = low + ((high  low) / 2)
(why? discussed above)4.
low <= high
in the while loop
Is it always the case that nosotros use the aforementioned conditions for binary search?Information technology depends…on our problem argument.
Example #1:
Given an integer x, find the maximum chemical element y in an array of size Due north that satisfies the status
y <= x
Solution:
We know that x might not exist in the array. So the elementary binary search for x in the given array won’t work. But for binary search, all we know is x, the key. We need to discover the chemical element in the array that’due south the closest to x and less than or equal to 10 (if it exists).
3 things should e’er come to your mind when using binary search:1. How should we move low and high?2. How should we compute mid?three. What would be the condition in while loop?
We always start with ‘How to movement depression and loftier?’ and and then find out how to compute mid and what might be the while condition. Not the other way round.

How to motility low and high? When
array[mid] < x
, there is a possibility that the current element might be the reply (since y can be less than 10).Then, we shouldn’t discard mid while moving low.Hence, low becomes
depression = mid
, not
low = mid + 1
High remains the same as information technology is irrelevant to our problem statement. 
_How to compute mid?_When we use
mid = low + ((high  low) / 2)
, we are calculating the lower mid. And so when the #elements in the assortment is even, theif(array[mid] > ten)
becomes false, so control will become to else clause where
depression = mid
. This results in an infinite loop. (why? take an example)Hence, we take the college mid, i.e.,
mid = low + ((high  low + 1) / 2)

_What would be the condition in while loop?_Since we are storing the maximum chemical element that is less than or equal to ten in low, nosotros should stop when low = high and return
array[low]
or
array[high]
. Hence, the status in while loop is
low < high
.
The Solution in Java looks similar this:
Example #2:
Given an integer 10, find the minimum element y in an assortment of size N that satisfies the condition
y >= ten
Solution:
Using a similar analytical approach we used in example ane (effort it out yourself), we can say that

low = mid + ane
for moving depression 
high = mid
for moving loftier 
mid = low + ((loftier  low) / 2)
for calculating mid 
low < loftier
for the status in while loop
Hence we tin tabulate dissimilar scenarios of using Binary Search as follows:
Feel free to experiment with these conditions and I hope you take gotten some insight into how to use binary search.
Namaste!
50 O A D I N 1000
. . . comments &
more!
Source: https://hackernoon.com/binarysearchindetail914944a1434a