NPRG075

Programming language design

Tomáš Petříček, 309 (3rd floor)
petricek@d3s.mff.cuni.cz
https://tomasp.net | @tomaspetricek

Lectures: Monday 12:20, S7
https://d3s.mff.cuni.cz/teaching/nprg075

Introduction

What? Why? How?

Making programming
(languages | experience | systems)
better!

My background

  • PhD, University of Cambridge
    Context-aware programming languages
  • Microsoft Research Cambridge
    F# and applied functional programming
  • The Alan Turing Institute, London
    Expert and non-expert tools for data science
  • University of Kent, Canterbury
    History and programming systems

Types for context-aware programming

Program as expression in small formal language

Type system determines what programs are valid

Safety proof shows no unauthorized accesses

Data science tools and languages

Result is a document not a program

Working with one concrete dataset

Different language and system requirements!

Programming systems & history

Interacting with a stateful environment

Let programmers do more in new ways...

It's not just a language!

Bringing everything together

Systems ⊃ languages

  • Programming process matters
  • Tools shape languages
  • Harder to formalize & study!

Interdisciplinary research

  • Formal language models
  • Systematic design
  • Qualitative and quantitative studies

Case study: LINQ

LINQ queries in Visual Basic .NET and C#

Dim db As New northwindDataContext
Dim ukCompanies =
  From cust In db.Customers
  Where cust.Country = "UK"
  Select cust.CompanyName, cust.City

Why confuse programmers familiar with SQL?

SELECT [CompanyName], [City]
WHERE [Country] = 'UK'
FROM dbo.[Northwind]

What to expect?

Content and materials

Many different programming systems
TypeScript, Jupyter, ML/F#, Smalltalk, BASIC

Many different research methods
Design, logic, proofs, user studies

This is a new work-in-progress course
Slides on the web, but no textbook

Credit / zápočet

Small independent
or group project

Using any of the covered method

Described in a brief report (5 pages)

Deadlines
Topic by January 8
Draft by February 28

Programming languages

Conventional topics

Paradigms and features

Language paradigms

  • Functional, OOP, Logic, etc.
  • Their fundamental concepts
  • Interesting "extreme" designs

Language features

  • Variable scoping, pointers
  • Lambda abstraction, inheritance
  • Design and implementation

Theory and implementation

Parsing and automata

  • Theory of formal grammars
  • Parser implementation
  • Computability theory

Compilers and interpreters

  • Implementation techniques
  • Register allocation
  • Meta-circular interpreters

Why is this
not enough?

Talks about "what"
but not about "how"

Treat design as a research problem!

What can we study about programming systems?

Design

As a research discipline

What is design?

Design is the intentional solution of a problem, by the creation of plans for a new sort of thing, where the plans would not be immediately seen, by a reasonable person, as an inadequate solution.

Parsons (2015)

Designerly ways

Sciences study natural world

  • By experiment, aiming at truth

Humanities study experience

  • By analogy, aiming at justice

Design studies the artificial

  • By synthesis, aiming at appropriateness

Cultures of programming

Common ways of thinking

Case study: TypeScript

Unsoundness by design

Design questions

  • What research methods to use?
  • Is partial soundness a thing?
  • Is there a better design?
  • What does "better" mean?

Cultures of programming

Engineering culture

  • Programs are complex systems
  • Tools can help us cope
  • Careful balance of trade-offs

Mathematical culture

  • Programs as formal entities
  • Like good mathematics...
  • Safe, composable, elegant

Cultures of programming

Humanistic culture

  • Augmenting human intellect
  • Programming helps us think
  • Language close to human concepts

Hacker culture

  • Programs are fundamentally bits
  • Do not restrict the programmer
  • Convenience, but full access

Type safety

Different perspectives

  • Safety is the very essence of types!
  • Useful as long as it makes programming easier
  • Sometimes, you need to break the rules
  • Does it help programmers think better?

Research methods

Interdisciplinary research

Interdisciplinary programming language research

Creating designs
Interviews, prototyping, formalism, analysis, history

Evaluating designs
Qualitative and quantitative studies, formal proofs

(Coblenz et al., 2018)

Programming language theory

Prove properties about small formal models

"Well-typed programs do not go wrong"

Discover and avoid subtle mistakes!

Human-centric system design

User studies, questionnaires, interviews, etc.

Qualitative analysis to design & test ideas

Quantitative analysis to compare designs

History of programming

What interesting past ideas were lost?

And the socio-political reasons for that?

Use history as source for new design ideas!

Conclusions

What to expect

Course outline

Preliminary structure

Design - Design and pattern languages
Usability - Human-centric language design
Semantics - Formal models of programming
Types - Types and type safety proofs
Beyond - Unexpected perspectives on types
Paradigms - History and programming systems
Complementary - Learning from past systems
Cognition - How humans think about programming

Reading

Jeremy Singer on Notebooks

Why should you read this?

  • You'll get more out of the lecture...
  • Perfect for the morning tram ride :-)
  • Notebooks are curious programming systems!

Conclusions

How to do research about
programming language design?

  • Inherently interdisciplinary topic
  • Logic, design, user studies, history & more!

Tomáš Petříček, 309 (3rd floor)
petricek@d3s.mff.cuni.cz
https://tomasp.net | @tomaspetricek
https://d3s.mff.cuni.cz/teaching/nprg075

References

Methodology

Assorted examples