A Smalltalk-80 compiler in Swift
This compiler is designed to work on the language specified in The Blue Book ("Smalltalk-80 The Language and it's Implementation").
The current scanner can recognise all Smalltalk-80 syntax.
This implementation includes support for byte arrays, and dynamic arrays which are not referenced in the syntax diagrams of the Blue Book.
The token scanner is fairly standard. It recognises identifiers, symbols, characters, strings, comments, block arguments, and numbers and stored the text in the token. Otherwise, the token just uses the TokenType to identify it. All tokens record the character position from the source that they were found at.
The only complication is negative numbers. A - before a digit could be a negative number or a binary method. The possible situations are checked to determine which is the appropriate way to tokenise this. The method used seems to work so far.
The parser is a hand written back-tracking parser. It has been tested on the Pharo source (about 8,000 classes with over 90,000 methods) and doesn't generate any errors. It has also now been tested on the original Smalltalk-80 sources file without errors.
The compiler uses the visitor patterns to descend the parse tree of a method and convert it to bytecode. A pre-run is done to build up the list of temporaries in order to ensure they end up in the same order as the original Smalltalk-80 system. This was done so that the output could be directly compared with the bytecodes of that system. A large number of classes have been included as tests to check this works. There are a small number of differences, though. These are in some situations where the original compiler adds the same value to the literal list twice. No attempt has been made to replicate this. It only happens for a small number of cases (less than 10).