The parser is already of a reasonable size (~15kb in min+zip), but it could be further reduced by refactoring the parse_tree_builder and lexer to serialize instead of rebuild at runtime. (also should have a slight faster loading time).
The serialized data can be changed into a better format, more efficient and with less repetition. Although it's not clear that it will make much difference when zipped.
Size is affected by parser + serialized data
The parser is already of a reasonable size (~15kb in min+zip), but it could be further reduced by refactoring the parse_tree_builder and lexer to serialize instead of rebuild at runtime. (also should have a slight faster loading time).
The serialized data can be changed into a better format, more efficient and with less repetition. Although it's not clear that it will make much difference when zipped.