Exploring the design space of programming systems
Tomas Petricek, Charles University, Prague
What my boss thinks I do...
Context-aware programming languages
Types for data science scripting
What I think I actually do...
History and philosophy of programming
Cultures of Programming
What I will talk about today...
Programming systems design & methodology
Documents and edit history
How this fits together?
Opening the software black box
- Systems that let users look inside
- Easy to use, understand and adapt!
- What happened to the 1970s attempts?
- How to talk about interactive systems?
Programming systems
and their technical dimensions
Programming Languages
Programming is
writing code
Formal semantics, implementation, paradigms, types
We know how
to study this!
Programming Systems
Interacting with a stateful system
Feedback, liveness, interactive user interfaces
But how do we
study this?
Paradigm shift in 1990s
Scientific paradigm shifts
- The past is a foreign country
- Paradigms are incommensurable
- Old notions stop making any sense
From systems to languages
- From CLOS and Flavors to mixins
- From running system to code
- From state & interaction to semantics
Looking at programming systems
They are interactive
- Hard to see interaction "on paper"
- Screencast, interactive essay
- What makes for a rigorous demo?
What to say about them
- Here is a cool new system!
- Essential characteristics?
-
How to compare, advance and
stand on the shoulders of giants?
Technical dimensions
What is a dimension
- Capture interesting properties
- Cover old and new systems
- Allow qualitative comparison
Example dimensions
- Modes of interaction
- Uniformity, conceptual structure
- Learnability & sociability
Technical dimensions catalogue
Analysis of:
LISP machines,
UNIX, Hypercard, Spreadsheets,
Haskell, Boxer,
Web, Dark, etc.
How to use dimensions
Analysis tool
- What makes system interesting?
- Compare against existing
- Does not say what's good!
Design space mapping
- Discover design opportunities
- Are there blank areas?
- Graphical self-sustainable systems!
Design exploration
Documents and edit history
What I want?
A programming system that is
- Accessible and simple on first encounter
- But allows power-users to modify the system
- Can be used in local-first collaborative ways
- Open and can explain its working
Two ideas
Transparent document format
- Document with code and data
- Smalltalk image easier to navigate
- Also contain evaluation trace!
Program as a sequence of edits
- Can be replayed to get the document
- Makes merging documents easier
- Record and/or analyse past edits?
Demo: Conference planning
Demo
Add speaker & refactor list
Sample edits
Shared baseline with multiple sequences of edits added
Typical local-first workflow, with independent edits
Two ways
of merging
Do they result in equivalent documents?
Can we have a conflict that imposes order?
Formulas
Code as document elements
- Store formulas as trees (AST) in document
- Render in a (somewhat) nicer way
- Evaluation adds edits to the log!
- Beware interactions with edit merging
Code is data
Absolute selectors only for now
A selector
is a sequence of:
- Field name
- Index specifier
- All selector
Demo
Budget calculation & refactor list
Merging of formulas
Edit after formula add
- Formula exists in the document
-
Structure edit edits structure
and all references in the doc
Formula add after edit
-
Selectors in newly added edits
reconciled with all new non-baseline edits - Edit adds correct formula!
Evaluation
How it interacts with editing
- Evaluation adds edits that replace nodes!
- Same edit reconciliation as before
- But conflicting edit removes evaluated edit!
- Sometimes, they could adapt though...
Demo
Adding a speaker and evaluation
Future directions
-
Formally show edits can always be merged?
Maybe, but probably only for a very small subset -
Projectional editing for nicer experience
Could editors written in the document itself? -
Rethinking abstraction and copy & paste
Abstract past edits, programming by demonstration -
Add user interactions, also as edits
How soon will this become unmaintainable?
Complementary science
Learning from past systems
DEMO
How Tomas learned programming
Learning from past systems
What's cool about 90s web?
- Source code available & accessible
- Easily extractable & copyable snippets
- Poor engineering in a poor language!
Complementary science
- Do (serious) history to recover ideas
- Lost due paradigm shifts in science
- Even more powerful for computing!
Good old systems
What have we forgotten about?
- Pygmalion - program by direct manipulation
- HyperCard - usability, levels, editing = running
- Boxer - program as transparent document
- Interlisp - code = data, modifiable, live coding!
DEMO
Recovering Commodore 64 BASIC
Commodore 64
What makes this interesting?
- Editing & running as one mode of interaction
- Load code even to run a game!
- Learning by copying code from magazines
- Convenience with escape hatches (POKE)
Conclusions
Programming systems
Open positions
in Prague
Postdoc and PhD funding available
Growing PL group
with more people starting soon
Email me to
learn more!
Systems ⊃ Languages
-
Need new research methodologies
Qualitative, formal, historical, etc. -
New designs outside of LISP or Smalltalk
Document-based with edit history?
Tomas Petricek, Charles University, Prague