So you're making a new programming language
Language design
Artistic
- How does it feel to use?
- Can I express my thoughts in new ways?
Technical
- What problems does it solve?
- What are the novelities (potential innovations)?
Syntax
The compiler
Flow
The syntax tree
Expressions
Tokens
Tokens
PushToStack
Whitespace
Number
Name
NewLine
Comment
Lexer
| Character | State | State change |
|---|---|---|
| <start> | Ready | |
| "#" | Ready | Read PushToStack { # } |
| " " | Reading whitespace | |
| "1" | Reading number | Read Whitespace { " " } |
| "0" | Reading number | |
| " " | Reading whitespace | Read number { 10 } |
| "%" | Reading comment | Read Whitespace { " " } |
Lexer
| Character | State | State change |
|---|---|---|
| "%" | Reading comment | Read Whitespace { " " } |
| " " | Reading comment | |
| "p" | Reading comment | |
| "u" | Reading comment | |
| "s" | Reading comment | |
| "h" | Reading comment | |
| "\n" | Ready | Read Comment { "% push" } |
Lexer
Running...
Tokens
| Token |
|---|
| PushToStack |
| Number { value: 10 } |
| Comment { value: "push" } |
| Newline { value: "\n" } |
| Name { value: "print" } |
| Newline { value: "\n" } |
Expressions
PushToStack
Number
FunctionCall
Parsing
Parsing
Interpreter
- Runtime
- Evaluate one line at a time
- Usually run-until-error
- Usually slower
Compiler
- Errors (mostly) exist in data or compiler, not code
- Consider entire program
- Usually faster, due to optimizations
Generators just each use one branch at a time
...until you've generated everything
Tooling
Syntax highlighting
Auto-formatting
Auto Format
Inline-errors
Editor support (LSP)
On hover
Editor support (LSP)
Autocomplete
vscode - run file
Native code
Intepreted + debug
Playground
Repl
Once you have an AST, you can do anything
1 hour to think
3 hours for MVP
1 hour for LSP
1 hour for Playground
3 hours to compile to native
3 hours to debug
It's not as hard as you think
https://github.com/eeue56/tegan-lang
