Big languages (Java, Kotlin, Ceylon) seem to be named after big islands. Siilinkari is barely a language, just like Siilinkari is barely an island.
In fact, the point of Siilinkari is not to write programs in it, but rather to function as a readable introduction to the world of compilers. You can browse the code, try out implementing new features and play around.
# ./gradlew clean build
# java -jar build/libs/siilinkari-all.jar
Welcome to Siilinkari! Enjoy your stay or type 'exit' to get out.
> var x = 4
> var s = ""
> if (x == 2 + 2) { var t = "It"; s = t + " worked!" }
> s
String = "It worked!"
> :dump if (x == 2 + 2) { var t = "It"; s = t + " worked!" }
0 frame[2] = global[4] ; x
1 frame[3] = Constant(4)
2 frame[2] = frame[2] == frame[3]
3 jump-if-false frame[2] 12
4 jump 5
5 frame[2] = Constant("It")
6 frame[3] = frame[2] ; (dup)
7 frame[1] = frame[3] ; store local t
8 frame[3] = Constant(" worked!")
9 frame[2] = frame[2] ++ frame[3]
10 global[5] = frame[2] ; s
11 jump 12
12 frame[2] = Constant(Unit)
13 quit frame[2]
> exit
Thank you for visiting Siilinkari, have a nice day!
The evaluator consists of the following pipeline:
Representation Stage Important classes
-------------- ----- -----------------
source code String
|
|
o--------- lexer siilinkari.lexer.{Lexer,LookaheadLexer}
|
v
tokens siilinkari.lexer.Token
|
|
o--------- parser siilinkari.parser.Parser
|
v
abstract syntax tree siilinkari.ast.{Expression, Statement}
|
|
o--------- type checker siilinkari.types.{Type, TypeChecker}, siilinkari.env.{StaticEnvironment, Binding}
|
v
typed syntax tree siilinkari.types.{TypedExpression, TypedStatement}
|
|
o--------- optimizer siilinkari.optimizer.ASTOptimizer
|
v
typed syntax tree (same as above)
|
|
o--------- translator siilinkari.translator.Translator
|
v
IR opcodes siilinkari.translator.{IR, BasicBlock}
|
|
o--------- peephole siilinkari.optimizer.peepholeOptimize
| optimization
|
v
IR opcodes (same as above)
|
|
o------- IR translation siilinkari.translator.TranslateIRToOpCodes
|
v
stack vm opcodes siilinkari.vm.{OpCode, CodeSegment}
|
|
o--------- evaluator siilinkari.vm.{Evaluator, ThreadState, DataSegment}
|
v
values siilinkari.objects.Value
Start from the important classes and either work your way through the pipeline, or just concentrate on a single transformation pass.
If you want to play around, here are some ideas to get started:
- Add new
unless (x) { ... }statement that works likeif (!x) { ... }. - Convert
if-statement to expression so you can sayvar x = if (a) b else c;. - Add a simple evaluator that evaluates either the typed AST directly instead of translating the code to stack vm opcodes.
- Implement a pretty printer: it should take an AST and write it to output properly formatted.
- Modify lexer and parser so that semicolons are not required at the end of statements.
- Modify lexer and parser to support indentation based syntax.
- Add support for higher order functions and lambdas.
If you want a taste of a bigger language, take a look at Blunt. It implements a large subset of ML/Haskell-like language (see e.g. prelude.blunt and river.blunt), but naturally the implementation is a lot more complex.