kraken/stdlib/ast_transformation.krak

import symbol:*
import tree:*
import vector:*
import stack:*
import map:*
import util:*
import string:*
import mem:*
import io:*
import importer:*
import ast_nodes:*
import type:*

adt search_type {
    none,
    function: vector<*type>
}

obj ast_transformation (Object) {
    var ast_to_syntax: map<*ast_node, *tree<symbol>>
    fun construct(): *ast_transformation {
        ast_to_syntax.construct()
        return this
    }
    fun copy_construct(old: *ast_transformation) {
        ast_to_syntax.copy_construct(&old->ast_to_syntax)
    }
    fun operator=(old: ref ast_transformation) {
        destruct()
        copy_construct(&old)
    }
    fun destruct() {
        ast_to_syntax.destruct()
    }
    // first pass defines all type_defs (objects and aliases), ADTs, and top-level if-comps/passthroughs
    fun first_pass(file_name: string, parse_tree: *tree<symbol>, importer: *importer): *ast_node {
        var translation_unit = ast_translation_unit_ptr(file_name)
        importer->register(file_name, parse_tree, translation_unit)
        parse_tree->children.for_each(fun(child: *tree<symbol>) {
            if (child->data.name == "type_def") {
                var name = concat_symbol_tree(get_node("identifier", child))
                var type_def_node = ast_type_def_ptr(name)
                type_def_node->type_def.self_type = type_ptr(type_def_node)
                translation_unit->translation_unit.children.add(type_def_node)
                ast_to_syntax.set(type_def_node, child)
                add_to_scope("~enclosing_scope", translation_unit, type_def_node)
                add_to_scope(name, type_def_node, translation_unit)
                // set up type - self-referential, traits, template, etc
            } else if (child->data.name == "adt_def") {
                var name = concat_symbol_tree(get_node("identifier", child))
                var adt_def_node = ast_adt_def_ptr(name)
                translation_unit->translation_unit.children.add(adt_def_node)
                ast_to_syntax.set(adt_def_node, child)
                add_to_scope("~enclosing_scope", translation_unit, adt_def_node)
                add_to_scope(name, adt_def_node, translation_unit)
            } else if (child->data.name == "if_comp") {
                var if_comp_node = transform_if_comp(child, translation_unit)
                translation_unit->translation_unit.children.add(if_comp_node)
                ast_to_syntax.set(if_comp_node, child)
            }
        })

        // now do all imports (done second so that if it imports this translation_unit,
        // this one already has all its types defined
        parse_tree->children.for_each(fun(child: *tree<symbol>) {
            if (child->data.name == "import") {
                var import_identifier_children = get_nodes("identifier", child)
                var name = concat_symbol_tree(import_identifier_children[0])
                var outside_translation_unit = importer->import_first_pass(name + ".krak")
                add_to_scope(name, outside_translation_unit, translation_unit)
                var import_node = ast_import_ptr(name, outside_translation_unit)
                translation_unit->translation_unit.children.add(import_node)
                ast_to_syntax.set(import_node, child)
                add_to_scope("~enclosing_scope", translation_unit, import_node)
                import_node->import.imported = from_vector(import_identifier_children.slice(1,-1).map(fun(ident: *tree<symbol>):string return concat_symbol_tree(ident);))
            }
        })
        return translation_unit
    }
    // defines inside of objects + ADTs, outside declaration statements, and function prototypes
    fun second_pass(parse_tree: *tree<symbol>, translation_unit: *ast_node) {
        println(string("Second Pass for ") + translation_unit->translation_unit.name)
        // we go through the parse tree for getting functions, but we're going through the ast for the things we've already set up and using the ast_to_syntax map
        parse_tree->children.for_each(fun(child: *tree<symbol>) {
            if (child->data.name == "function") {
                var function_node = second_pass_function(child, translation_unit, map<string, *type>())
                translation_unit->translation_unit.children.add(function_node)
                ast_to_syntax.set(function_node, child)
            } else if (child->data.name == "declaration_statement") {
                // second pass declaration can actually just call a normal transform (but maybe should be it's own method to do so because typedef has to do it too?)...
                translation_unit->translation_unit.children.add(transform_declaration_statement(child, translation_unit))
            }
        })
        // work on the ones already started
        translation_unit->translation_unit.children.for_each(fun(node: *ast_node) {
            match(*node) {
                ast_node::type_def(backing) {
                    var type_def_syntax = ast_to_syntax[node]
                    type_def_syntax->children.for_each(fun(child: *tree<symbol>) {
                        if (child->data.name == "declaration_statement") {
                            var declaration_node = transform_declaration_statement(child, node)
                            node->type_def.variables.add(declaration_node)
                            ast_to_syntax.set(declaration_node, child)
                        } else if (child->data.name == "function") {
                            var function_node = second_pass_function(child, node, map<string, *type>())
                            node->type_def.methods.add(function_node)
                            ast_to_syntax.set(function_node, child)
                        }
                    })
                }
                ast_node::adt_def(backing) do_nothing() // actually go through and do methods inside
            }
        })
    }
    fun second_pass_function(node: *tree<symbol>, translation_unit: *ast_node, template_replacements: map<string, *type>): *ast_node {
        var function_name = concat_symbol_tree(get_node("func_identifier", node))
        // check to see if it is a template
        // figure out return type
        var typed_return_node = get_node("typed_return", node)
        // darn no ternary yet
        var return_type = null<type>()
        if (typed_return_node) return_type = transform_type(get_node("type", typed_return_node), translation_unit, template_replacements)
                         else return_type = type_ptr(base_type::void_return())
        // transform parameters
        var parameters = vector<*ast_node>()
        get_nodes("typed_parameter", node).for_each(fun(child: *tree<symbol>) {
            parameters.add(ast_identifier_ptr(concat_symbol_tree(get_node("identifier", child)), transform_type(get_node("type", child), translation_unit, template_replacements)))
        })
        // figure out function type and make function_node
        var function_node = ast_function_ptr(function_name, type_ptr(parameters.map(fun(parameter: *ast_node): *type return parameter->identifier.type;), return_type), parameters)
        // add to scope (translation_unit)
        add_to_scope(function_name, function_node, translation_unit)
        add_to_scope("~enclosing_scope", translation_unit, function_node)
        // add parameters to scope of function
        parameters.for_each(fun(parameter: *ast_node) add_to_scope(parameter->identifier.name, parameter, function_node);)
        return function_node
    }
    // The third pass finishes up by doing all function bodies (top level and methods in objects)
    fun third_pass(parse_tree: *tree<symbol>, translation_unit: *ast_node) {
        println(string("Third Pass for ") + translation_unit->translation_unit.name)
        translation_unit->translation_unit.children.for_each(fun(node: *ast_node) {
            match(*node) {
                ast_node::type_def(backing) {
                    // make sure not a template? or the method not a template?
                    // also same body problem as below
                    node->type_def.methods.for_each(fun(method: *ast_node) {
                        method->function.body_statement = transform_statement(get_node("statement", ast_to_syntax[method]), method)
                    })
                }
                ast_node::function(backing) {
                    // make sure not a template
                    // huh, I guess I can't actually assign to the backing.
                    // This is actually a little bit of a problem, maybe these should be pointers also. All the pointers!
                    node->function.body_statement = transform_statement(get_node("statement", ast_to_syntax[node]), node)
                }
            }
        })
    }
    // The fourth pass generates the class templates that have not yet been generated in a "chaotic iteration" loop
    fun fourth_pass(parse_tree: *tree<symbol>, translation_unit: *ast_node) {
        println(string("Fourth Pass for ") + translation_unit->translation_unit.name)
    }
    fun transform_type(node: *tree<symbol>, scope: *ast_node, template_replacements: map<string, *type>): *type {
        // check for references and step down
        // always get to pre-reffed level
        var real_node = get_node("pre_reffed", node)
        // check for indirection and step down
        var indirection = 0
        while (get_node("pre_reffed", real_node)) {
            real_node = get_node("pre_reffed", real_node)
            indirection++
        }
        var type_syntax_str = concat_symbol_tree(real_node)
        println(type_syntax_str + " *************************")
        // should take into account indirection and references...
        if (type_syntax_str == "void")
            return type_ptr(base_type::void_return(), indirection)
        else if (type_syntax_str == "bool")
            return type_ptr(base_type::boolean(), indirection)
        else if (type_syntax_str == "int")
            return type_ptr(base_type::integer(), indirection)
        else if (type_syntax_str == "float")
            return type_ptr(base_type::floating(), indirection)
        else if (type_syntax_str == "double")
            return type_ptr(base_type::double_precision(), indirection)
        else if (type_syntax_str == "char")
            return type_ptr(base_type::character(), indirection)
        else if (/* check for function type*/ false)
            return type_ptr(base_type::function(), indirection)
        else {
            // do lookup for objects, ADTs, templates, etc
            var possibilities = scope_lookup(type_syntax_str, scope)
            print("There are "); print(possibilities.size); println(" possibilites for this object type lookup")
            for (var i = 0; i < possibilities.size; i++;) {
                match(*possibilities[i]) {
                    ast_node::type_def(backing) return backing.self_type->clone_with_indirection(indirection)
                }
            }
            println("No objects in lookup, returning none")
            return type_ptr(base_type::none(), indirection)
        }
    }
/*NodeTree<ASTData>* ASTTransformation::transform(NodeTree<Symbol>* from, NodeTree<ASTData>* scope, std::vector<Type> types, bool limitToFunction, std::map<std::string, Type*> templateTypeReplacements) {*/
    fun transform(node: *tree<symbol>, scope: *ast_node): *ast_node return transform(node, scope, search_type::none())

    fun transform(node: *tree<symbol>, scope: *ast_node, searching_for: search_type): *ast_node {
        var name = node->data.name
        if (name == "identifier" || name == "scoped_identifier") {
            return transform_identifier(node, scope, searching_for)
        } else if (name == "code_block") {
            return transform_code_block(node, scope)
        } else if (name == "if_comp") {
            return transform_if_comp(node, scope)
        } else if (name == "simple_passthrough") {
            return transform_simple_passthrough(node, scope)
        } else if (name == "statement") {
            return transform_statement(node, scope)
        } else if (name == "declaration_statement") {
            return transform_declaration_statement(node, scope)
        } else if (name == "assignment_statement") {
            return transform_assignment_statement(node, scope)
        } else if (name == "if_statement") {
            return transform_if_statement(node, scope)
        } else if (name == "while_loop") {
            return transform_while_loop(node, scope)
        } else if (name == "for_loop") {
            return transform_for_loop(node, scope)
        } else if (name == "return_statement") {
            return transform_return_statement(node, scope)
        } else if (name == "function_call") {
            return transform_function_call(node, scope)
        } else if (name == "boolean_expression" || name == "and_boolean_expression"
                || name == "bool_exp" || name == "expression"
                || name == "shiftand" || name == "term"
                || name == "factor" || name == "unarad"
                || name == "access_operation"
        ) {
            // for now, assume passthrough and just transform underneath
            return transform_expression(node, scope, searching_for)
        } else if (name == "bool" || name == "string"
                || name == "character" || name == "number"
        ) {
            println(string("transforming value: ") + name)
            return transform_value(node, scope)
        }
        print("FAILED TO TRANSFORM: "); print(name + ": "); println(concat_symbol_tree(node))
        return null<ast_node>()
    }
    fun transform_all(nodes: vector<*tree<symbol>>, scope: *ast_node): vector<*ast_node> {
        return nodes.map(fun(node: *tree<symbol>): *ast_node return transform(node, scope);)
    }
    fun transform_identifier(node: *tree<symbol>, scope: *ast_node, searching_for: search_type): *ast_node {
        // first, we check for and generate this
        var name = concat_symbol_tree(node)
        if (name == "this") {
            while (!is_type_def(scope)) scope = get_ast_scope(scope)->get(string("~enclosing_scope"))[0]
            return ast_identifier_ptr("this", scope->type_def.self_type->clone_with_indirection(1))
        }
        /*return ast_identifier_ptr(name, type_ptr(base_type::void_return()))*/
        match (searching_for) {
            search_type::none() return identifier_lookup(name, scope)
            search_type::function(type_vec) return function_lookup(name, scope, type_vec)
        }
        println("FAILED SEARCH FOR")
        return null<ast_node>()
    }
    fun transform_value(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var value_str = concat_symbol_tree(node)
        var value_type = null<type>()
        if (value_str[0] == '"')
            value_type = type_ptr(base_type::character(), 1)
        else if (value_str[0] == '\'') //' lol, comment hack for vim syntax highlighting (my fault, of course)
            value_type = type_ptr(base_type::character())
        else {
            // should differentiate between float and double...
            var contains_dot = false
            for (var i = 0; i < value_str.length(); i++;) {
                if (value_str[i] == '.') {
                    contains_dot = true
                    break
                }
            }
            if (contains_dot)
                value_type = type_ptr(base_type::floating())
            else
                value_type = type_ptr(base_type::integer())
        }
        return ast_value_ptr(value_str, value_type)
    }
    fun transform_code_block(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var new_block = ast_code_block_ptr()
        add_to_scope("~enclosing_scope", scope, new_block)
        new_block->code_block.children = transform_all(node->children, new_block)
        return new_block
    }
    fun transform_if_comp(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var new_if_comp = ast_if_comp_ptr()
        new_if_comp->if_comp.wanted_generator = concat_symbol_tree(get_node("identifier", node))
        new_if_comp->if_comp.statement = transform_statement(get_node("statement", node), scope)
        return new_if_comp
    }
    fun transform_simple_passthrough(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var new_passthrough = ast_simple_passthrough_ptr()
        // setup passthrough params and string
        new_passthrough->simple_passthrough.passthrough_str = concat_symbol_tree(get_node("triple_quoted_string", node)).slice(3,-4)
        return new_passthrough
    }
    fun transform_statement(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var new_statement = ast_statement_ptr()
        new_statement->statement.child = transform(node->children[0], scope)
        return new_statement
    }
    fun transform_declaration_statement(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var name = concat_symbol_tree(get_node("identifier", node))
        // may have type, or an expression, or both
        var type_syntax_node = get_node("type", node)
        var expression_syntax_node = get_node("boolean_expression", node)
        var ident_type = null<type>()
        var expression = null<ast_node>()
        if (type_syntax_node) ident_type = transform_type(type_syntax_node, scope, map<string, *type>())
        if (expression_syntax_node) {
            expression = transform(expression_syntax_node, scope)
            if (!type_syntax_node)
                ident_type = get_ast_type(expression)
        }
        if (!ident_type) error("declaration statement with no type or expression from which to inference type")
        var identifier = ast_identifier_ptr(name, ident_type)
        var declaration = ast_declaration_statement_ptr(identifier, expression)
        add_to_scope(name, identifier, scope)
        return declaration
    }
    fun transform_assignment_statement(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var assignment = ast_assignment_statement_ptr(transform(get_node("factor", node), scope), transform(get_node("boolean_expression", node), scope))
        return assignment
    }
    fun transform_if_statement(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var if_statement = ast_if_statement_ptr(transform_expression(get_node("boolean_expression", node), scope))
        // one variable declarations might be in a code_block-less if statement
        add_to_scope("~enclosing_scope", scope, if_statement)
        var statements = transform_all(get_nodes("statement", node), if_statement)
        if_statement->if_statement.then_part = statements[0]
        // we have an else
        if (statements.size == 2)
            if_statement->if_statement.else_part = statements[1]
        return if_statement
    }
    fun transform_while_loop(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var while_loop = ast_while_loop_ptr(transform_expression(get_node("boolean_expression", node), scope))
        add_to_scope("~enclosing_scope", scope, while_loop)
        while_loop->while_loop.statement = transform(get_node("statement", node), while_loop)
        return while_loop
    }
    fun transform_for_loop(node: *tree<symbol>, scope: *ast_node): *ast_node {
        var for_loop = ast_for_loop_ptr()
        add_to_scope("~enclosing_scope", scope, for_loop)
        var statements = get_nodes("statement", node)
        for_loop->for_loop.init = transform(statements[0], for_loop)
        for_loop->for_loop.condition = transform(get_node("boolean_expression", node), for_loop)
        for_loop->for_loop.update = transform(statements[1], for_loop)
        for_loop->for_loop.body = transform(statements[2], for_loop)
        return for_loop
    }
    fun transform_return_statement(node: *tree<symbol>, scope: *ast_node): *ast_node {
        return ast_return_statement_ptr(transform(node->children[0], scope))
    }
    fun transform_function_call(node: *tree<symbol>, scope: *ast_node): *ast_node {
        // don't bother with a full transform for parameters with their own function, just get the boolean expression and transform it
        var parameters = get_nodes("parameter", node).map(fun(child: *tree<symbol>): *ast_node return transform(get_node("boolean_expression", child), scope);)
        var parameter_types = parameters.map(fun(param: *ast_node): *type return get_ast_type(param);)
        /*return ast_function_call_ptr(function_lookup(concat_symbol_tree(get_node("unarad", node)), scope), parameters)*/
        /*var f = ast_function_call_ptr(function_lookup(concat_symbol_tree(get_node("unarad", node)), scope, parameter_types), parameters)*/
        var f = ast_function_call_ptr(transform(get_node("unarad", node), scope, search_type::function(parameter_types)), parameters)
        print("function call function ")
        println(f->function_call.func)
        print("function call parameters ")
        f->function_call.parameters.for_each(fun(param: *ast_node) print(param);)
        return f
    }
    fun transform_expression(node: *tree<symbol>, scope: *ast_node): *ast_node return transform_expression(node, scope, search_type::none())
    fun transform_expression(node: *tree<symbol>, scope: *ast_node, searching_for: search_type): *ast_node {
        // figure out what the expression is, handle overloads, or you know
        // ignore everything and do a passthrough
        var func_name = string()
        var parameters = vector<*ast_node>()
        if (node->children.size == 1)
            return transform(node->children[0], scope, searching_for)
        else if (node->children.size == 2) {
            var check_if_post = concat_symbol_tree(node->children[1])
            if (check_if_post == "--" || check_if_post == "++") {
                // give the post-operators a special suffix so the c_generator knows to emit them post
                func_name = concat_symbol_tree(node->children[1]) + "p"
                parameters = vector(transform(node->children[0], scope))
            } else {
                func_name = concat_symbol_tree(node->children[0])
                parameters = vector(transform(node->children[1], scope))
            }
        } else {
            func_name = concat_symbol_tree(node->children[1])
            var first_param = transform(node->children[0], scope)
            var second_param = null<ast_node>()
            if (func_name == "." || func_name == "->") {
                println("Gonna do the internal scope thing")
                second_param = transform(node->children[2], get_ast_type(first_param)->type_def, searching_for)
            } else {
                println("Gonna do regular scope thing")
                second_param = transform(node->children[2], scope)
            }
            parameters = vector(first_param, second_param)
        }
        var parameter_types = parameters.map(fun(param: *ast_node): *type return get_ast_type(param);)
        return ast_function_call_ptr(get_builtin_function(func_name, parameter_types), parameters)
    }
    fun get_builtin_function(name: string, param_types: vector<*type>): *ast_node {
        if (name == "." || name == "->")
            return ast_function_ptr(name, type_ptr(param_types, param_types[1]), vector<*ast_node>())
        return ast_function_ptr(name, type_ptr(param_types, param_types[0]), vector<*ast_node>())
    }
    fun function_lookup(name: string, scope: *ast_node, param_types: vector<*type>): *ast_node {
        println(string("doing function lookup for: ") + name)
        var param_string = string()
        param_types.for_each(fun(t: *type) param_string += t->to_string() + ", ";)
        var results = scope_lookup(name, scope)
        print(results.size); println(" number of results")
        for (var i = 0; i < results.size; i++;) {
            if (is_function(results[i])) {
                var func_param_types = get_ast_type(results[i])->parameter_types
                if (func_param_types.size != param_types.size) {
                    println(string("type sizes don't match") + get_ast_type(results[i])->to_string() + " with needed " + param_string)
                    continue
                }
                var param_types_match = true
                for (var j = 0; j < param_types.size; j++;) {
                    if (*func_param_types[j] != *param_types[j]) {
                        param_types_match = false
                        break
                    }
                }
                if (param_types_match)
                    return results[i]
            }
            println(string("either isn't function or types don't match ") + get_ast_type(results[i])->to_string() + " with needed " + param_string)
        }
        println(string("function lookup failed for ") + name)
        return null<ast_node>()
    }
    fun identifier_lookup(name: string, scope: *ast_node): *ast_node {
        println(string("doing identifier lookup for: ") + name)
        var results = scope_lookup(name, scope)
        if (!results.size) {
            println(string("identifier lookup failed for ") + name)
            return null<ast_node>()
        }
        return results[0]
    }
    fun scope_lookup(name: string, scope: *ast_node): vector<*ast_node> {
        println("*****Doing a name lookup for*****")
        println(name)
        return scope_lookup_helper(name, scope)
    }
    fun scope_lookup_helper(name: string, scope: *ast_node): vector<*ast_node> {
        // need to do properly scopded lookups
        // prevent re-checking the same one...
        print("scope is: ")
        get_ast_scope(scope)->for_each(fun(key: string, value: vector<*ast_node>) print(key + " ");)
        println()
        var results = vector<*ast_node>()
        if (get_ast_scope(scope)->contains_key(name)) {
            println(name + " is in scope, adding to results")
            results += get_ast_scope(scope)->get(name)
        }
        if (get_ast_scope(scope)->contains_key(string("~enclosing_scope")))
            results += scope_lookup_helper(name, get_ast_scope(scope)->get(string("~enclosing_scope"))[0])
        if (is_translation_unit(scope)) {
            scope->translation_unit.children.for_each(fun(child: *ast_node) {
                if (is_import(child) && child->import.imported.contains(name)) {
                    println(name + " is indeed imported")
                    results += scope_lookup_helper(name, child->import.translation_unit)
                } else println(name + " is not imported (this time)")
            })
        }
        return results
    }
}

fun concat_symbol_tree(node: *tree<symbol>): string {
    var str.construct(): string
    if (node->data.data != "no_value")
        str += node->data.data
    node->children.for_each(fun(child: *tree<symbol>) str += concat_symbol_tree(child);)
    return str
}
fun get_node(lookup: *char, parent: *tree<symbol>): *tree<symbol> {
    return get_node(string(lookup), parent)
}
fun get_node(lookup: string, parent: *tree<symbol>): *tree<symbol> {
    var results = get_nodes(lookup, parent)
    if (results.size > 1)
        println("too many results!")
    if (results.size)
        return results[0]
    return null<tree<symbol>>()
}
fun get_nodes(lookup: *char, parent: *tree<symbol>): vector<*tree<symbol>> {
    return get_nodes(string(lookup), parent)
}
fun get_nodes(lookup: string, parent: *tree<symbol>): vector<*tree<symbol>> {
    return parent->children.filter(fun(node: *tree<symbol>):bool return node->data.name == lookup;)
}
fun add_to_scope(name: *char, to_add: *ast_node, add_to: *ast_node) {
    add_to_scope(string(name), to_add, add_to)
}
fun add_to_scope(name: string, to_add: *ast_node, add_to: *ast_node) {
    var add_to_map = get_ast_scope(add_to)
    if (add_to_map->contains_key(name))
        (*add_to_map)[name].add(to_add)
    else
        add_to_map->set(name, vector(to_add))
}
fun error(message: *char) error(string(message));
fun error(message: string) {
    println("****ERROR****")
    println(message)
    while (true){}
}