Session 5 — Variables, Math, Conditionals & Error Handling¶

What Was Built¶

The core features that make Terse a real general-purpose programming language — variables, arithmetic, conditional logic, and a proper error handling system.

Error Handling¶

File: src/interpreter/errors.py

A new dedicated errors module with three custom exception types:

Class	Purpose
`TerseError`	Base error — message and optional line number
`TerseNameError`	Unknown node or function
`TerseSyntaxError`	Malformed statement
`TerseTypeError`	Wrong type used

Before this session, errors produced Python tracebacks with no context. Now Terse produces clean, readable messages:

TerseError: unknown node 'dgo' — has it been declared with 'know'?
TerseError: unknown function 'clasify' — has it been defined with 'to'?

Variables¶

Variables store a value under a name and persist for the life of the program.

x = 5
name = dog
count = 0

The interpreter stores variables in self.variables{} separately from the knowledge base. Variables can reference other variables:

x = 5
y = x

Implementation Note

All parser imports must be at the top of interpreter.py — not inside methods. Python's module caching causes isinstance checks to fail silently when the same class is imported from two different module instances. This was a significant bug discovered and fixed this session.

Math Expressions¶

Full arithmetic expression support with variable references.

y = x + 10
z = y * 2
result = 100 / 4
total = x + y + z

Supported operators: +, -, *, /

The parser collects everything on the right side of = as an expression list. The interpreter resolves variable names to their values, then evaluates the expression.

x = 10
y = x * 2    →    y = 20
z = x + y    →    z = 30

Conditionals¶

If/else conditional logic with full comparison operator support.

if x > 5
  know size is big
else
  know size is small

Supported operators: >, <, >=, <=, =, !=

Variables are resolved before comparison — if x > 5 uses the current value of x. The else body is optional.

Bug Fixes¶

Two significant bugs were discovered and fixed this session:

Multi-value facts — Previously know dog has fur followed by know dog has legs would overwrite fur with legs. Now both values are stored as a list: has: ['fur', 'legs']. Inference rules check against all values in the list.

Each loop type matching — After inference ran and updated a node's is value (e.g. dog going from is: animal to is: mammal), each beast in animal could no longer find the node. Fixed by introducing self.types{} — a separate registry that tracks original declared types and is never overwritten by inference.

Showcase Program — animal_mind.trs¶

A complete Terse program demonstrating all language features working together:

know dog is animal
know dog has fur
know dog has legs
know eagle is animal
know eagle has wings

when has fur then is mammal
when has wings then can fly

to describe creature
  infer creature
  return creature

each beast in animal
  describe beast

Output — all four animals correctly classified, all behavioral sequences correctly predicted.

REPL Commands¶

The REPL gained a new command this session:

Command	What it does
`variables`	Print all current variable assignments

Phase 1 Progress After Session 5¶

Component	Status
Lexer	✅ Complete
Parser	✅ Complete
Interpreter core	✅ Complete
Facts & relationships	✅ Complete
Inference rules	✅ Complete
Functions with scope	✅ Complete
Markov chain learning	✅ Complete
Each loops	✅ Complete
While loops	✅ Complete
main.py	✅ Complete
REPL	✅ Complete
Error handling	✅ Complete
Variables	✅ Complete
Math expressions	✅ Complete
Conditionals	✅ Complete
Tensors	🔵 Next session
Compression keywords	🔵 Next session
Ethics rule syntax	⚪ Planned

Interpreter completeness: ~90%