Why Functional Programming Matters


Baishampayan "BG" Ghose
CTO/Co-founder - Helpshift, Inc.


Why Functional Programming Matters

(today)


Baishampayan "BG" Ghose
CTO/Co-founder - Helpshift, Inc.





A Quick Poll

Which programming language do you use at work?


  • Java
  • Ruby
  • Python
  • JavaScript
  • Any functional language?

Does your code run perfectly the first time?

Yes/No


Why not?
How do we get there?

About Me

  • Started my career hacking Lisp
  • Want to write correct code that runs fast
  • Career polyglot programmer with focus on Functional Programming
  • CTO/Co-founder at Helpshift, Inc.



A Brief Interlude

 


von Neumann's Core Ideas


  • Stored program model
  • Computation through mutation of memory cells
  • Assumes single thread of execution
  • Abstractions to distort reality in other cases





And then, reality strikes back...





Clearly, we have run out of options



Functional Programming


A Primer

History


Core Ideas


  • Computation based on evaluation of λ-expressions
  • Immutability
  • First class, and hence higher-order functions
  • Sequence transformations instead of mutation of memory cells
  • Referential transparency

Immutability


x = [1, 2, 3, 4];
print x; // [1, 2, 3, 4]
y = x.append(5);
print y; // [1, 2, 3, 4, 5]
print x; // [1, 2, 3, 4] 

Because: Maths!

First Class Functions


def make_adder(x):
    # close over x
    return lambda y: x + y

add5 = make_adder(5) # I am returning a function here

add5(10) # 15

###
["Church","Turing","Gödel"].sort(key=lower)

Sequence Transformation


nums = [77, 45, 2, 98, -12]

# square all numbers
nums2 = map(square, nums) # [5929, 2025, 4, 9604, 144]

# keep only > 1000
print filter(lambda n: n > 1000, nums2)[5929, 2025, 9604]

Referential Transparency


square(5) == 25 # a pure function

square(5) + square(5) + square(5)

# compiler can optimize this as

25 + 25 + 25

Because: square is "referentially transparent"



Benefits

  • Higher levels of abstractions
  • Increased modularity
  • Lazy evaluation
  • Concurrency
  • Parallelism
  • Memoization
  • Testability
  • Composability

FP Languages













           ML








FP Languages











Interesting Ideas That Work Well With FP


  • Communicating Sequential Processes
  • Generative Testing
  • Logic/Relational Programming
  • Parsers





Thanks!

 

We're hiring: www.helpshift.com

Why Functional Programming Matters (Today)

By Baishampayan Ghose

Why Functional Programming Matters (Today)

  • 3,810