Data Structures
Why Data Structures
Mostafa S. Ibrahim
Teaching, Training and Coaching since more than a decade!
Artificial Intelligence & Computer Vision Researcher
PhD from Simon Fraser University - Canada
Bachelor / Msc from Cairo University - Egypt
Ex-(Software Engineer / ICPC World Finalist)
2 Common Questions
● Why do we need data structures?
● Why do we study data structures if we have built-in ones?
The need
● Last lecture should implicitly be an answer that
● Primitive arrays are not enough
● We face many cases where we need our own user-defined data structures
○ What if want to make correspondence between a string and some object
■ E.g. data[“mostafa”] = someobject
■ But array is only based on integers? We need something that has greater flexibility!
● This is where map or hash-table can help us!
■ E.g. we have 1 Billion word text from articles and want to search for a specific word
● Array of words is very slow.
● But Trie or Suffix Tree data structures are much faster!
● Let’s see an industrial case: Google maps
Google Maps
● With maps, we can go
from a place to another
● We need data structure
to represent points and
streets!
○ We have a lot of data!
○ Constraints: time/date/car
● We need efficient
functions to find optimal
paths
Data Organization Perspective
● From a purpose to another, we may arrange the same data in different ways
○ There are many ways to implement a data structure
● When we have much data (Millions of users on facebook), things become
much more complex and critical!
○ E.g. Search the engines for scientific or social purposes
○ E.g. If a storm hits specific list of locations, how many homes will face a power outage?
○ E.g. We are in a war, and want to destroy the minimum number of bridges in a city to
disconnect 2 points of our enemy? Rockets are expensive!
Data Structure Efficiency
● Assume we have N (10000) employees
● Is a loop over these N employees as fast as 3 nested loops?
○ No, it seems like 10000 operations vs 10^12 operations!
○ Seems the first is efficient, but the 2nd is not!
● So how to measure the efficiency of a function?
○ The complexity (asymptotic) analysis in the algorithms field answers that.
○ The efficiency can be for time and memory (space)
○ For the same problem, we may arrange data in a way that is so fast in computations, but
another is much memory efficient
■ You maybe lucky and your DS is both time and memory efficient!
○ On a mobile: you may target a memory efficient approach
○ On real-time service: you may target a time efficient approach
Why not just use built-in DS?
● Using built-in data structures as a black box without understanding their
details is big risk in real projects
○ Typically you won’t realize the time/memory order
○ Typically you will use them improperly as you don’t recognize their differences
● More critical
○ In practice, you may need to design your own data structures, but you lack the skill
○ E.g. what are the different data organization perspectives that we may use?
■ What are the pros/cons of them?
■ The time/memory differences?
■ The tradeoffs?
● Side effect
○ Your thinking skills are improved. This is good for problem-solving & algorithms course
Use DS first & learn later approach
● You may learn using built-in data structures first, before learning them
○ What not How
● Many built-in data structures exist
● It is good during a programming course to use them!
○ See my Mastering 4 Critical Skills in C++17 course
○ You can even solve many problem-solving problems on them (See the course)
■ Mainly: Black-box solving
● When it comes to learning later, things are much easier to guess
○ Even you can implement some basic data structures by yourself!
● Trap: Many students after learn the usage, the decide to skip studying!
○ Don’t be that guy
“Acquire knowledge and impart it to the people.”
“Seek knowledge from the Cradle to the Grave.”
