compact compiler ids and spans to reduce memory usage

TODO.md

@@ -4,7 +4,6 @@
 
 # compiler hardening follow-ups
 
-- migrate spans, AST/HIR/IR ids, and diagnostics from `int` to compact integer types
 - prune unreachable function specializations before HIR construction and emission
 - support unary minus, including the signed i64 minimum literal boundary
 - move ignored example binaries into a dedicated build directory and remove `.review_tmp`

benchmarks/symbols/README.md

@@ -29,3 +29,9 @@ The measured run reduced token size by 14.3% and peak tracked memory by 9.8%.
 After the iterative compiler-hardening work on 2026-06-12, the same benchmark
 reported 13,222,443 peak bytes and 40,117 allocations. The reusable traversal
 stacks keep allocation count effectively unchanged from the interning baseline.
+
+After migrating persistent compiler references and spans to compact IDs on
+2026-06-12, the benchmark reported 8,651,659 peak bytes and 40,117 allocations.
+`Span` is 12 bytes and `Token` is 24 bytes; AST expressions, HIR expressions,
+and IR instructions are 64, 88, and 88 bytes respectively. These sizes are
+also printed by the benchmark to catch layout regressions.

benchmarks/symbols/main.odin

@@ -3,6 +3,7 @@ package main
 import "../../compiler/ast"
 import "../../compiler/checker"
 import "../../compiler/hir"
+import "../../compiler/ir"
 import "../../compiler/lexer"
 import "../../compiler/parser"
 import "../../compiler/source"
@@ -72,6 +73,10 @@ main :: proc() {
 	fmt.printf("allocation_count: %d\n", tracker.total_allocation_count)
 	fmt.printf("token_count: %d\n", metrics.token_count)
 	fmt.printf("token_size_bytes: %d\n", size_of(token.Token))
+	fmt.printf("span_size_bytes: %d\n", size_of(source.Span))
+	fmt.printf("ast_expr_size_bytes: %d\n", size_of(ast.Expr))
+	fmt.printf("hir_expr_size_bytes: %d\n", size_of(hir.Expr))
+	fmt.printf("ir_instruction_size_bytes: %d\n", size_of(ir.Instruction))
 	fmt.printf("unique_symbol_count: %d\n", metrics.unique_symbol_count)
 	fmt.printf("stored_symbol_bytes: %d\n", metrics.stored_symbol_bytes)
 	fmt.printf("diagnostic_count: %d\n", metrics.diagnostic_count)

compiler/ast/ast.odin

@@ -4,9 +4,63 @@ import "../source"
 import "../symbol"
 import "core:mem"
 
-INVALID_ID :: -1
+Expr_Id :: distinct u32
+Stmt_Id :: distinct u32
+Function_Id :: distinct u32
+Global_Id :: distinct u32
+Import_Id :: distinct u32
+File_Id :: distinct u32
+Package_Id :: distinct u32
 
-Type_Syntax :: enum {
+INVALID_EXPR :: Expr_Id(0xffff_ffff)
+INVALID_STMT :: Stmt_Id(0xffff_ffff)
+INVALID_FUNCTION :: Function_Id(0xffff_ffff)
+INVALID_GLOBAL :: Global_Id(0xffff_ffff)
+INVALID_IMPORT :: Import_Id(0xffff_ffff)
+INVALID_FILE :: File_Id(0xffff_ffff)
+INVALID_PACKAGE :: Package_Id(0xffff_ffff)
+
+expr_id :: proc(index: int) -> Expr_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_EXPR))
+	return Expr_Id(index)
+}
+
+stmt_id :: proc(index: int) -> Stmt_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_STMT))
+	return Stmt_Id(index)
+}
+
+function_id :: proc(index: int) -> Function_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_FUNCTION))
+	return Function_Id(index)
+}
+
+global_id :: proc(index: int) -> Global_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_GLOBAL))
+	return Global_Id(index)
+}
+
+import_id :: proc(index: int) -> Import_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_IMPORT))
+	return Import_Id(index)
+}
+
+file_id :: proc(index: int) -> File_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_FILE))
+	return File_Id(index)
+}
+
+package_id :: proc(index: int) -> Package_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_PACKAGE))
+	return Package_Id(index)
+}
+
+index :: proc(id: $T, invalid: T, count: int) -> (int, bool) {
+	value := int(id)
+	return value, id != invalid && value < count
+}
+
+Type_Syntax :: enum u8 {
 	Invalid,
 	Int,
 	I8,
@@ -16,7 +70,7 @@ Type_Syntax :: enum {
 	Void,
 }
 
-Expr_Kind :: enum {
+Expr_Kind :: enum u8 {
 	Invalid,
 	Integer,
 	Name,
@@ -25,15 +79,15 @@ Expr_Kind :: enum {
 }
 
 Expr :: struct {
-	kind:       Expr_Kind,
 	span:       source.Span,
+	integer:    i64,
+	args:       []Expr_Id,
 	qualifier:  symbol.Id,
 	name:       symbol.Id,
-	integer:    i64,
+	left:       Expr_Id,
-	left:       int,
+	right:      Expr_Id,
-	right:      int,
+	diagnostic: source.Diagnostic_Id,
-	args:       []int,
+	kind:       Expr_Kind,
-	diagnostic: int,
 }
 
 Param :: struct {
@@ -42,7 +96,7 @@ Param :: struct {
 	type: Type_Syntax,
 }
 
-Stmt_Kind :: enum {
+Stmt_Kind :: enum u8 {
 	Invalid,
 	Declaration,
 	Assignment,
@@ -56,49 +110,49 @@ Stmt :: struct {
 	name:       symbol.Id,
 	type:       Type_Syntax,
 	immutable:  bool,
-	expr:       int,
+	expr:       Expr_Id,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
 }
 
 Function :: struct {
 	span:       source.Span,
 	name:       symbol.Id,
-	pkg:        int,
+	pkg:        Package_Id,
-	file:       int,
+	file:       File_Id,
 	c_abi:      bool,
 	has_body:   bool,
 	params:     []Param,
 	result:     Type_Syntax,
-	body:       []int,
+	body:       []Stmt_Id,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
 }
 
 Global :: struct {
 	span:       source.Span,
 	name:       symbol.Id,
-	pkg:        int,
+	pkg:        Package_Id,
-	file:       int,
+	file:       File_Id,
 	type:       Type_Syntax,
 	immutable:  bool,
-	expr:       int,
+	expr:       Expr_Id,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
 }
 
 Import :: struct {
 	span:       source.Span,
 	alias:      symbol.Id,
 	path:       string,
-	pkg:        int,
+	pkg:        Package_Id,
-	file:       int,
+	file:       File_Id,
-	target:     int,
+	target:     Package_Id,
 	valid:      bool,
 	used:       bool,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
 }
 
 File :: struct {
-	source:  int,
+	source:  source.Source_Id,
-	pkg:     int,
+	pkg:     Package_Id,
 }
 
 Package :: struct {

compiler/compiler.odin

@@ -56,6 +56,7 @@ compile_package :: proc(input_path, output_path: string, link_arguments: []linke
 		vmem.arena_allocator(&parser_arena),
 	)
 	if !loaded {
+		source.print_all(&diagnostics)
 		fmt.eprintln("failed to load root package directory:", input_path)
 		return 2
 	}

compiler/hir/hir.odin

@@ -5,24 +5,73 @@ import "../symbol"
 import "../types"
 import "core:mem"
 
-INVALID_ID :: -1
+Expr_Id :: distinct u32
+Stmt_Id :: distinct u32
+Local_Id :: distinct u32
+Function_Id :: distinct u32
+Global_Id :: distinct u32
+Ref :: distinct u32
 
-Calling_Convention :: enum {
+INVALID_EXPR :: Expr_Id(0xffff_ffff)
+INVALID_STMT :: Stmt_Id(0xffff_ffff)
+INVALID_LOCAL :: Local_Id(0xffff_ffff)
+INVALID_FUNCTION :: Function_Id(0xffff_ffff)
+INVALID_GLOBAL :: Global_Id(0xffff_ffff)
+INVALID_REF :: Ref(0xffff_ffff)
+
+expr_id :: proc(index: int) -> Expr_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_EXPR))
+	return Expr_Id(index)
+}
+
+stmt_id :: proc(index: int) -> Stmt_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_STMT))
+	return Stmt_Id(index)
+}
+
+local_id :: proc(index: int) -> Local_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_LOCAL))
+	return Local_Id(index)
+}
+
+function_id :: proc(index: int) -> Function_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_FUNCTION))
+	return Function_Id(index)
+}
+
+global_id :: proc(index: int) -> Global_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_GLOBAL))
+	return Global_Id(index)
+}
+
+index :: proc(id: $T, invalid: T, count: int) -> (int, bool) {
+	value := int(id)
+	return value, id != invalid && value < count
+}
+
+local_ref :: proc(id: Local_Id) -> Ref {return Ref(id)}
+function_ref :: proc(id: Function_Id) -> Ref {return Ref(id)}
+global_ref :: proc(id: Global_Id) -> Ref {return Ref(id)}
+as_local :: proc(ref: Ref) -> Local_Id {return Local_Id(ref)}
+as_function :: proc(ref: Ref) -> Function_Id {return Function_Id(ref)}
+as_global :: proc(ref: Ref) -> Global_Id {return Global_Id(ref)}
+
+Calling_Convention :: enum u8 {
 	Brolang,
 	C,
 }
 
-Implementation :: enum {
+Implementation :: enum u8 {
 	Definition,
 	Declaration,
 }
 
-Linkage :: enum {
+Linkage :: enum u8 {
 	Internal,
 	External,
 }
 
-Expr_Kind :: enum {
+Expr_Kind :: enum u8 {
 	Invalid,
 	Integer,
 	Local,
@@ -33,15 +82,15 @@ Expr_Kind :: enum {
 }
 
 Expr :: struct {
-	kind:       Expr_Kind,
 	span:       source.Span,
 	type:       types.Type,
 	integer:    i64,
-	target:     int,
+	args:       []Expr_Id,
-	left:       int,
+	target:     Ref,
-	right:      int,
+	left:       Expr_Id,
-	args:       []int,
+	right:      Expr_Id,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
+	kind:       Expr_Kind,
 }
 
 Local :: struct {
@@ -51,7 +100,7 @@ Local :: struct {
 	parameter: bool,
 }
 
-Stmt_Kind :: enum {
+Stmt_Kind :: enum u8 {
 	Declaration,
 	Assignment,
 	Return,
@@ -63,9 +112,9 @@ Stmt_Kind :: enum {
 Stmt :: struct {
 	kind:       Stmt_Kind,
 	span:       source.Span,
-	local:      int,
+	local:      Local_Id,
-	expr:       int,
+	expr:       Expr_Id,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
 }
 
 Function :: struct {
@@ -75,27 +124,27 @@ Function :: struct {
 	implementation:      Implementation,
 	linkage:             Linkage,
 	is_main:             bool,
-	params:              []int,
+	params:              []Local_Id,
 	result:              types.Type,
 	locals:              []Local,
-	body:                []int,
+	body:                []Stmt_Id,
-	direct_global_reads: [dynamic]int,
+	direct_global_reads: [dynamic]Global_Id,
-	calls:               []int,
+	calls:               []Function_Id,
 	problematic:         bool,
-	diagnostic:          int,
+	diagnostic:          source.Diagnostic_Id,
 }
 
 Global :: struct {
 	name:          symbol.Id,
 	type:          types.Type,
-	expr:          int,
+	expr:          Expr_Id,
 	static_value:  i64,
 	is_static:     bool,
-	dependencies:  [dynamic]int,
+	dependencies:  [dynamic]Global_Id,
-	calls:         []int,
+	calls:         []Function_Id,
 	direct_problem: bool,
 	problematic:   bool,
-	diagnostic:    int,
+	diagnostic:    source.Diagnostic_Id,
 }
 
 Module :: struct {

compiler/ir/ir.odin

@@ -5,24 +5,66 @@ import "../symbol"
 import "../types"
 import "core:mem"
 
-INVALID_ID :: -1
+Instruction_Id :: distinct u32
+Local_Id :: distinct u32
+Function_Id :: distinct u32
+Global_Id :: distinct u32
+Ref :: distinct u32
 
-Calling_Convention :: enum {
+INVALID_INSTRUCTION :: Instruction_Id(0xffff_ffff)
+INVALID_LOCAL :: Local_Id(0xffff_ffff)
+INVALID_FUNCTION :: Function_Id(0xffff_ffff)
+INVALID_GLOBAL :: Global_Id(0xffff_ffff)
+INVALID_REF :: Ref(0xffff_ffff)
+
+instruction_id :: proc(index: int) -> Instruction_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_INSTRUCTION))
+	return Instruction_Id(index)
+}
+
+local_id :: proc(index: int) -> Local_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_LOCAL))
+	return Local_Id(index)
+}
+
+function_id :: proc(index: int) -> Function_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_FUNCTION))
+	return Function_Id(index)
+}
+
+global_id :: proc(index: int) -> Global_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_GLOBAL))
+	return Global_Id(index)
+}
+
+index :: proc(id: $T, invalid: T, count: int) -> (int, bool) {
+	value := int(id)
+	return value, id != invalid && value < count
+}
+
+local_ref :: proc(id: Local_Id) -> Ref {return Ref(id)}
+function_ref :: proc(id: Function_Id) -> Ref {return Ref(id)}
+global_ref :: proc(id: Global_Id) -> Ref {return Ref(id)}
+as_local :: proc(ref: Ref) -> Local_Id {return Local_Id(ref)}
+as_function :: proc(ref: Ref) -> Function_Id {return Function_Id(ref)}
+as_global :: proc(ref: Ref) -> Global_Id {return Global_Id(ref)}
+
+Calling_Convention :: enum u8 {
 	Brolang,
 	C,
 }
 
-Implementation :: enum {
+Implementation :: enum u8 {
 	Definition,
 	Declaration,
 }
 
-Linkage :: enum {
+Linkage :: enum u8 {
 	Internal,
 	External,
 }
 
-Opcode :: enum {
+Opcode :: enum u8 {
 	Param,
 	Const,
 	Load_Global,
@@ -38,15 +80,15 @@ Opcode :: enum {
 }
 
 Instruction :: struct {
-	op:         Opcode,
 	span:       source.Span,
 	type:       types.Type,
 	integer:    i64,
-	target:     int,
+	args:       []Instruction_Id,
-	a:          int,
+	target:     Ref,
-	b:          int,
+	a:          Instruction_Id,
-	args:       []int,
+	b:          Instruction_Id,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
+	op:         Opcode,
 }
 
 Function :: struct {
@@ -68,7 +110,7 @@ Global :: struct {
 	static_value: i64,
 	initializer:  []Instruction,
 	problematic:  bool,
-	diagnostic:   int,
+	diagnostic:   source.Diagnostic_Id,
 }
 
 Module :: struct {

compiler/lexer/lexer.odin

@@ -34,11 +34,11 @@ append_token :: proc(
 	kind: token.Kind,
 	start, end: int,
 	id := symbol.INVALID,
-	diagnostic := -1,
+	diagnostic := source.INVALID_DIAGNOSTIC,
 ) {
 	append(&stream.items, token.Token{
 		kind=kind,
-		span=source.Span{file=source_file.id, start=start, end=end},
+		span=source.Span{file=source_file.id, start=source.Offset(start), end=source.Offset(end)},
 		symbol=id,
 		diagnostic=diagnostic,
 	})
@@ -74,7 +74,7 @@ lex :: proc(
 				cursor += 1
 				append_token(&stream, source_file, .Colon_Colon, start, cursor)
 			} else {
-				id := source.add(diagnostics, source.Span{file=source_file.id, start=start, end=cursor}, "expected a second ':'")
+				id := source.add(diagnostics, source.Span{file=source_file.id, start=source.Offset(start), end=source.Offset(cursor)}, "expected a second ':'")
 				append_token(&stream, source_file, .Invalid, start, cursor, diagnostic=id)
 			}
 		case '=':
@@ -111,7 +111,7 @@ lex :: proc(
 					if cursor >= len(bytes) || (bytes[cursor] != '\\' && bytes[cursor] != '"') {
 						source.add(
 							diagnostics,
-							source.Span{file=source_file.id, start=max(cursor-1, start), end=min(cursor+1, len(bytes))},
+							source.Span{file=source_file.id, start=source.Offset(max(cursor-1, start)), end=source.Offset(min(cursor+1, len(bytes)))},
 							"import strings only support '\\\\' and '\\\"' escapes",
 						)
 						valid = false
@@ -127,7 +127,7 @@ lex :: proc(
 			} else {
 				id := source.add(
 					diagnostics,
-					source.Span{file=source_file.id, start=start, end=cursor},
+					source.Span{file=source_file.id, start=source.Offset(start), end=source.Offset(cursor)},
 					"unterminated import string",
 				)
 				append_token(&stream, source_file, .Invalid, start, cursor, diagnostic=id)
@@ -135,7 +135,7 @@ lex :: proc(
 		case ';':
 			id := source.add(
 				diagnostics,
-				source.Span{file=source_file.id, start=cursor, end=cursor+1},
+				source.Span{file=source_file.id, start=source.Offset(cursor), end=source.Offset(cursor+1)},
 				"semicolons are invalid; terminate statements with a newline",
 			)
 			append_token(&stream, source_file, .Invalid, cursor, cursor+1, diagnostic=id)
@@ -162,7 +162,7 @@ lex :: proc(
 			} else {
 				id := source.addf(
 					diagnostics,
-					source.Span{file=source_file.id, start=cursor, end=cursor+1},
+					source.Span{file=source_file.id, start=source.Offset(cursor), end=source.Offset(cursor+1)},
 					"invalid source byte 0x%02x",
 					value,
 				)

compiler/llvm/llvm.odin

@@ -49,11 +49,11 @@ sentinel :: proc(value_type: types.Type) -> i64 {
 	}
 }
 
-valid_instruction :: proc(instructions: []ir.Instruction, instruction_id: int) -> bool {
+valid_instruction :: proc(instructions: []ir.Instruction, instruction_id: ir.Instruction_Id) -> bool {
-	return instruction_id >= 0 && instruction_id < len(instructions)
+	return instruction_id != ir.INVALID_INSTRUCTION && int(instruction_id) < len(instructions)
 }
 
-valid_value :: proc(instructions: []ir.Instruction, value_id: int, expected: types.Type) -> bool {
+valid_value :: proc(instructions: []ir.Instruction, value_id: ir.Instruction_Id, expected: types.Type) -> bool {
 	if !valid_instruction(instructions, value_id) ||
 	   !types.is_concrete_integer(expected) ||
 	   !types.equal(instructions[value_id].type, expected) {
@@ -68,7 +68,7 @@ valid_value :: proc(instructions: []ir.Instruction, value_id: int, expected: typ
 	return false
 }
 
-write_operand :: proc(builder: ^strings.Builder, instructions: []ir.Instruction, value_id: int, expected: types.Type) {
+write_operand :: proc(builder: ^strings.Builder, instructions: []ir.Instruction, value_id: ir.Instruction_Id, expected: types.Type) {
 	if !valid_value(instructions, value_id, expected) {
 		fmt.sbprintf(builder, "%d", sentinel(expected))
 		return
@@ -88,8 +88,8 @@ register_message :: proc(emitter: ^Emitter, text: string) -> int {
 	return id
 }
 
-diagnostic_message :: proc(emitter: ^Emitter, diagnostic: int, span: source.Span, fallback: string) -> int {
+diagnostic_message :: proc(emitter: ^Emitter, diagnostic: source.Diagnostic_Id, span: source.Span, fallback: string) -> int {
-	if diagnostic >= 0 && diagnostic < len(emitter.diagnostics.items) {
+	if _, ok := source.diagnostic_index(diagnostic, len(emitter.diagnostics.items)); ok {
 		message := source.format(emitter.diagnostics, diagnostic, emitter.allocator)
 		id := register_message(emitter, message)
 		delete(message, emitter.allocator)
@@ -137,7 +137,7 @@ emit_recovery_value :: proc(emitter: ^Emitter, instruction_id: int, instruction:
 	}
 }
 
-emit_call_args :: proc(builder: ^strings.Builder, instructions: []ir.Instruction, args: []int, param_types: []types.Type) {
+emit_call_args :: proc(builder: ^strings.Builder, instructions: []ir.Instruction, args: []ir.Instruction_Id, param_types: []types.Type) {
 	for arg, index in args {
 		if index > 0 {
 			strings.write_string(builder, ", ")
@@ -152,60 +152,62 @@ emit_instruction_stream :: proc(
 	instructions: []ir.Instruction,
 	function: ir.Function,
 	global_initializer := false,
-) -> int {
+) -> ir.Instruction_Id {
-	return_value := -1
+	return_value := ir.INVALID_INSTRUCTION
 	after_return := false
-	for instruction, instruction_id in instructions {
+	for instruction, instruction_index in instructions {
+		instruction_id := ir.instruction_id(instruction_index)
 		if after_return {
-			fmt.sbprintf(&emitter.builder, "recover_after_return_%d:\n", instruction_id)
+			fmt.sbprintf(&emitter.builder, "recover_after_return_%d:\n", instruction_index)
 			after_return = false
 		}
 		switch instruction.op {
 		case .Param, .Const:
 		case .Load_Global:
-			if instruction.target < 0 || instruction.target >= len(emitter.module.globals) {
+			global_id := ir.as_global(instruction.target)
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid global reference")
+			if global_id == ir.INVALID_GLOBAL || int(global_id) >= len(emitter.module.globals) {
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid global reference")
 				continue
 			}
-			global := emitter.module.globals[instruction.target]
+			global := emitter.module.globals[global_id]
 			if !types.equal(instruction.type, global.type) {
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid global reference type")
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid global reference type")
 				continue
 			}
 			if global.is_static {
 				fmt.sbprintf(
 					&emitter.builder,
 					"  %%v%d = load %s, ptr @bro.g.%d\n",
-					instruction_id,
+					instruction_index,
 					llvm_type(global.type),
-					instruction.target,
+					global_id,
 				)
 			} else {
 				fmt.sbprintf(
 					&emitter.builder,
 					"  %%v%d = call %s @bro.get.%d()\n",
-					instruction_id,
+					instruction_index,
 					llvm_type(global.type),
-					instruction.target,
+					global_id,
 				)
 			}
 		case .Alloca:
 			if !types.is_concrete_integer(instruction.type) {
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid allocation type")
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid allocation type")
 				continue
 			}
-			fmt.sbprintf(&emitter.builder, "  %%v%d = alloca %s\n", instruction_id, llvm_type(instruction.type))
+			fmt.sbprintf(&emitter.builder, "  %%v%d = alloca %s\n", instruction_index, llvm_type(instruction.type))
 		case .Load:
 			if !valid_instruction(instructions, instruction.a) ||
 			   instructions[instruction.a].op != .Alloca ||
 			   !types.equal(instructions[instruction.a].type, instruction.type) {
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid load slot")
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid load slot")
 				continue
 			}
 			fmt.sbprintf(
 				&emitter.builder,
 				"  %%v%d = load %s, ptr %%v%d\n",
-				instruction_id,
+				instruction_index,
 				llvm_type(instruction.type),
 				instruction.a,
 			)
@@ -214,7 +216,7 @@ emit_instruction_stream :: proc(
 			   instructions[instruction.a].op != .Alloca ||
 			   !types.equal(instructions[instruction.a].type, instruction.type) ||
 			   !valid_value(instructions, instruction.b, instruction.type) {
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid store operand")
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid store operand")
 				continue
 			}
 			fmt.sbprintf(&emitter.builder, "  store %s ", llvm_type(instruction.type))
@@ -225,50 +227,51 @@ emit_instruction_stream :: proc(
 			   !types.is_concrete_integer(instructions[instruction.a].type) ||
 			   !types.is_concrete_integer(instruction.type) ||
 			   instructions[instruction.a].type.bits >= instruction.type.bits {
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid widening operand")
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid widening operand")
 				continue
 			}
 			from_type := instructions[instruction.a].type
-			fmt.sbprintf(&emitter.builder, "  %%v%d = sext %s ", instruction_id, llvm_type(from_type))
+			fmt.sbprintf(&emitter.builder, "  %%v%d = sext %s ", instruction_index, llvm_type(from_type))
 			write_operand(&emitter.builder, instructions, instruction.a, from_type)
 			fmt.sbprintf(&emitter.builder, " to %s\n", llvm_type(instruction.type))
 		case .Add_Checked:
 			if !valid_value(instructions, instruction.a, instruction.type) ||
 			   !valid_value(instructions, instruction.b, instruction.type) {
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid addition operand")
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid addition operand")
 				continue
 			}
 			type_name := llvm_type(instruction.type)
-			fmt.sbprintf(&emitter.builder, "  %%pair%d = call ", instruction_id)
+			fmt.sbprintf(&emitter.builder, "  %%pair%d = call ", instruction_index)
 			strings.write_string(&emitter.builder, "{ ")
 			fmt.sbprintf(&emitter.builder, "%s, i1 } @llvm.sadd.with.overflow.%s(%s ", type_name, type_name, type_name)
 			write_operand(&emitter.builder, instructions, instruction.a, instruction.type)
 			fmt.sbprintf(&emitter.builder, ", %s ", type_name)
 			write_operand(&emitter.builder, instructions, instruction.b, instruction.type)
 			fmt.sbprintf(&emitter.builder, ")\n")
-			fmt.sbprintf(&emitter.builder, "  %%v%d = extractvalue ", instruction_id)
+			fmt.sbprintf(&emitter.builder, "  %%v%d = extractvalue ", instruction_index)
 			strings.write_string(&emitter.builder, "{ ")
-			fmt.sbprintf(&emitter.builder, "%s, i1 } %%pair%d, 0\n", type_name, instruction_id)
+			fmt.sbprintf(&emitter.builder, "%s, i1 } %%pair%d, 0\n", type_name, instruction_index)
-			fmt.sbprintf(&emitter.builder, "  %%overflow%d = extractvalue ", instruction_id)
+			fmt.sbprintf(&emitter.builder, "  %%overflow%d = extractvalue ", instruction_index)
 			strings.write_string(&emitter.builder, "{ ")
-			fmt.sbprintf(&emitter.builder, "%s, i1 } %%pair%d, 1\n", type_name, instruction_id)
+			fmt.sbprintf(&emitter.builder, "%s, i1 } %%pair%d, 1\n", type_name, instruction_index)
 			fmt.sbprintf(
 				&emitter.builder,
 				"  br i1 %%overflow%d, label %%overflow_trap%d, label %%overflow_continue%d\n",
-				instruction_id,
+				instruction_index,
-				instruction_id,
+				instruction_index,
-				instruction_id,
+				instruction_index,
 			)
-			fmt.sbprintf(&emitter.builder, "overflow_trap%d:\n", instruction_id)
+			fmt.sbprintf(&emitter.builder, "overflow_trap%d:\n", instruction_index)
-			message := diagnostic_message(emitter, -1, instruction.span, "signed integer addition overflow")
+			message := diagnostic_message(emitter, source.INVALID_DIAGNOSTIC, instruction.span, "signed integer addition overflow")
 			emit_trap_call(emitter, message)
-			fmt.sbprintf(&emitter.builder, "  unreachable\noverflow_continue%d:\n", instruction_id)
+			fmt.sbprintf(&emitter.builder, "  unreachable\noverflow_continue%d:\n", instruction_index)
 		case .Call:
-			if instruction.target < 0 || instruction.target >= len(emitter.module.functions) {
+			function_id := ir.as_function(instruction.target)
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid function specialization")
+			if function_id == ir.INVALID_FUNCTION || int(function_id) >= len(emitter.module.functions) {
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid function specialization")
 				continue
 			}
-			target := emitter.module.functions[instruction.target]
+			target := emitter.module.functions[function_id]
 			valid_args := len(instruction.args) == len(target.param_types)
 			if valid_args {
 				for arg, index in instruction.args {
@@ -283,11 +286,11 @@ emit_instruction_stream :: proc(
 				target_result = types.I32
 			}
 			if !valid_args || !types.equal(instruction.type, target_result) {
-				emit_recovery_value(emitter, instruction_id, instruction, "invalid function call operands")
+				emit_recovery_value(emitter, instruction_index, instruction, "invalid function call operands")
 				continue
 			}
 			if instruction.type.kind != .Void {
-				fmt.sbprintf(&emitter.builder, "  %%v%d = ", instruction_id)
+				fmt.sbprintf(&emitter.builder, "  %%v%d = ", instruction_index)
 			} else {
 				strings.write_string(&emitter.builder, "  ")
 			}

compiler/loader/loader.odin

@@ -39,20 +39,20 @@ is_identifier :: proc(value: string) -> bool {
 	return true
 }
 
-find_package :: proc(state: ^State, path: string) -> int {
+find_package :: proc(state: ^State, path: string) -> ast.Package_Id {
 	for pkg, id in state.module.packages {
 		if pkg.path == path {
-			return id
+			return ast.package_id(id)
 		}
 	}
-	return -1
+	return ast.INVALID_PACKAGE
 }
 
-add_placeholder :: proc(state: ^State, path: string) -> int {
+add_placeholder :: proc(state: ^State, path: string) -> ast.Package_Id {
-	if existing := find_package(state, path); existing >= 0 {
+	if existing := find_package(state, path); existing != ast.INVALID_PACKAGE {
 		return existing
 	}
-	id := len(state.module.packages)
+	id := ast.package_id(len(state.module.packages))
 	append(&state.module.packages, ast.Package{
 		path=strings.clone(path, state.allocator),
 		name=symbol.intern(state.symbols, filepath.base(path)),
@@ -103,7 +103,7 @@ resolve_import_path :: proc(state: ^State, importing_path, import_path: string)
 	return joined, false
 }
 
-load_package :: proc(state: ^State, path: string, import_span: source.Span, is_root := false) -> int {
+load_package :: proc(state: ^State, path: string, import_span: source.Span, is_root := false) -> ast.Package_Id {
 	canonical, ok := filepath.abs(path, state.allocator)
 	if !ok || !os.is_dir(path) {
 		if is_root {
@@ -111,7 +111,7 @@ load_package :: proc(state: ^State, path: string, import_span: source.Span, is_r
 			if len(canonical) > 0 {
 				delete(canonical, state.allocator)
 			}
-			return -1
+			return ast.INVALID_PACKAGE
 		}
 		placeholder := path
 		if len(canonical) > 0 {
@@ -124,12 +124,12 @@ load_package :: proc(state: ^State, path: string, import_span: source.Span, is_r
 		}
 		return id
 	}
-	if existing := find_package(state, canonical); existing >= 0 {
+	if existing := find_package(state, canonical); existing != ast.INVALID_PACKAGE {
 		delete(canonical, state.allocator)
 		return existing
 	}
 
-	pkg_id := len(state.module.packages)
+	pkg_id := ast.package_id(len(state.module.packages))
 	append(&state.module.packages, ast.Package{
 		path=canonical,
 		name=symbol.intern(state.symbols, filepath.base(canonical)),
@@ -152,13 +152,24 @@ load_package :: proc(state: ^State, path: string, import_span: source.Span, is_r
 	}
 
 	for file_info in files {
+		if file_info.size < 0 || !source.fits_source_length(u64(file_info.size)) {
+			source.addf(state.diagnostics, import_span, "source file '%s' exceeds the 4 GiB source limit", file_info.fullpath)
+			state.root_failed = true
+			continue
+		}
 		bytes, read_ok := os.read_entire_file(file_info.fullpath, state.sources.allocator)
 		if !read_ok {
 			state.root_failed = true
 			continue
 		}
+		if !source.fits_source_length(u64(len(bytes))) {
+			source.addf(state.diagnostics, import_span, "source file '%s' exceeds the 4 GiB source limit", file_info.fullpath)
+			delete(bytes, state.sources.allocator)
+			state.root_failed = true
+			continue
+		}
 		source_id := source.add_source_owned(state.sources, file_info.fullpath, bytes)
-		file_id := len(state.module.files)
+		file_id := ast.file_id(len(state.module.files))
 		append(&state.module.files, ast.File{source=source_id, pkg=pkg_id})
 		stream := lexer.lex(&state.sources.items[source_id], state.diagnostics, state.symbols, state.token_allocator)
 		parser.parse_into(&stream, &state.sources.items[source_id], state.diagnostics, state.module, pkg_id, file_id)
@@ -169,7 +180,7 @@ load_package :: proc(state: ^State, path: string, import_span: source.Span, is_r
 	import_count := len(state.module.imports)
 	for import_id in 0..<import_count {
 		import_item := state.module.imports[import_id]
-		if import_item.pkg != pkg_id || import_item.target >= 0 {
+		if import_item.pkg != pkg_id || import_item.target != ast.INVALID_PACKAGE {
 			continue
 		}
 		if filepath.is_abs(import_item.path) {
@@ -181,7 +192,7 @@ load_package :: proc(state: ^State, path: string, import_span: source.Span, is_r
 		target_path, target_ok := resolve_import_path(state, canonical, import_item.path)
 		target := load_package(state, target_path, import_item.span)
 		state.module.imports[import_id].target = target
-		if !target_ok || target < 0 || !state.module.packages[target].available {
+		if !target_ok || target == ast.INVALID_PACKAGE || !state.module.packages[target].available {
 			state.module.imports[import_id].valid = false
 		}
 		delete(target_path, state.allocator)
@@ -189,7 +200,7 @@ load_package :: proc(state: ^State, path: string, import_span: source.Span, is_r
 	return pkg_id
 }
 
-declaration_conflicts :: proc(module: ^ast.Module, pkg: int, name: symbol.Id) -> bool {
+declaration_conflicts :: proc(module: ^ast.Module, pkg: ast.Package_Id, name: symbol.Id) -> bool {
 	for function in module.functions {
 		if function.pkg == pkg && function.name == name {
 			return true
@@ -205,7 +216,7 @@ declaration_conflicts :: proc(module: ^ast.Module, pkg: int, name: symbol.Id) ->
 
 validate_imports :: proc(state: ^State) {
 	for import_item, import_id in state.module.imports {
-		if !symbol.is_valid(import_item.alias) && import_item.target >= 0 {
+		if !symbol.is_valid(import_item.alias) && import_item.target != ast.INVALID_PACKAGE {
 			state.module.imports[import_id].alias = state.module.packages[import_item.target].name
 		}
 		alias := state.module.imports[import_id].alias
@@ -260,7 +271,7 @@ load :: proc(
 		allocator=allocator,
 	}
 	root := load_package(&state, root_path, source.Span{}, true)
-	if root != 0 && root >= 0 {
+	if root != ast.Package_Id(0) && root != ast.INVALID_PACKAGE {
 		state.root_failed = true
 	}
 	validate_imports(&state)

compiler/lower/lower.odin

@@ -10,20 +10,20 @@ import "core:mem"
 State :: struct {
 	hir_module:   ^hir.Module,
 	instructions: [dynamic]ir.Instruction,
-	local_values: []int,
+	local_values: []ir.Instruction_Id,
-	local_slots:  []int,
+	local_slots:  []ir.Instruction_Id,
 	expr_stack:   [dynamic]Lower_Expr_Frame,
 	allocator:    mem.Allocator,
 }
 
-append_instruction :: proc(state: ^State, instruction: ir.Instruction) -> int {
+append_instruction :: proc(state: ^State, instruction: ir.Instruction) -> ir.Instruction_Id {
-	id := len(state.instructions)
+	id := ir.instruction_id(len(state.instructions))
 	append(&state.instructions, instruction)
 	return id
 }
 
-clone_args :: proc(values: []int, allocator: mem.Allocator) -> []int {
+clone_args :: proc(values: []ir.Instruction_Id, allocator: mem.Allocator) -> []ir.Instruction_Id {
-	result := make([]int, len(values), allocator)
+	result := make([]ir.Instruction_Id, len(values), allocator)
 	copy(result, values)
 	return result
 }
@@ -37,14 +37,19 @@ sentinel :: proc(value_type: types.Type) -> i64 {
 	}
 }
 
-append_recovery_value :: proc(state: ^State, span: source.Span, value_type: types.Type, diagnostic := -1) -> int {
+append_recovery_value :: proc(
+	state: ^State,
+	span: source.Span,
+	value_type: types.Type,
+	diagnostic := source.INVALID_DIAGNOSTIC,
+) -> ir.Instruction_Id {
 	append_instruction(state, ir.Instruction{
 		op=.Trap,
 		span=span,
 		type=types.VOID,
-		target=-1,
+		target=ir.INVALID_REF,
-		a=-1,
+		a=ir.INVALID_INSTRUCTION,
-		b=-1,
+		b=ir.INVALID_INSTRUCTION,
 		diagnostic=diagnostic,
 	})
 	fallback := value_type
@@ -56,22 +61,22 @@ append_recovery_value :: proc(state: ^State, span: source.Span, value_type: type
 		span=span,
 		type=fallback,
 		integer=sentinel(fallback),
-		target=-1,
+		target=ir.INVALID_REF,
-		a=-1,
+		a=ir.INVALID_INSTRUCTION,
-		b=-1,
+		b=ir.INVALID_INSTRUCTION,
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 }
 
 Lower_Expr_Frame :: struct {
-	expr:      int,
+	expr:      hir.Expr_Id,
 	stage:     u8,
-	left:      int,
+	left:      ir.Instruction_Id,
 	arg_index: int,
-	args:      []int,
+	args:      []ir.Instruction_Id,
 }
 
-lower_expr :: proc(state: ^State, expr_id: int) -> int {
+lower_expr :: proc(state: ^State, expr_id: hir.Expr_Id) -> ir.Instruction_Id {
 	stack := state.expr_stack
 	clear_dynamic_array(&stack)
 	defer {
@@ -82,11 +87,11 @@ lower_expr :: proc(state: ^State, expr_id: int) -> int {
 		state.expr_stack = stack
 	}
 	append(&stack, Lower_Expr_Frame{expr=expr_id})
-	last := -1
+	last := ir.INVALID_INSTRUCTION
 	for len(stack) > 0 {
 		frame_index := len(stack)-1
 		frame := stack[frame_index]
-		if frame.expr < 0 || frame.expr >= len(state.hir_module.exprs) {
+		if frame.expr == hir.INVALID_EXPR || int(frame.expr) >= len(state.hir_module.exprs) {
 			last = append_recovery_value(state, source.Span{}, types.I64)
 			_ = pop(&stack)
 			continue
@@ -100,31 +105,33 @@ lower_expr :: proc(state: ^State, expr_id: int) -> int {
 			case .Integer:
 				last = append_instruction(state, ir.Instruction{
 					op=.Const, span=expr.span, type=expr.type, integer=expr.integer,
-					target=-1, a=-1, b=-1, diagnostic=-1,
+					target=ir.INVALID_REF, a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC,
 				})
 				_ = pop(&stack)
 			case .Local:
-				last = -1
+				last = ir.INVALID_INSTRUCTION
-				if expr.target >= 0 && expr.target < len(state.local_slots) && state.local_slots[expr.target] >= 0 {
+				local := hir.as_local(expr.target)
+				if local != hir.INVALID_LOCAL && int(local) < len(state.local_slots) && state.local_slots[local] != ir.INVALID_INSTRUCTION {
 					last = append_instruction(state, ir.Instruction{
-						op=.Load, span=expr.span, type=expr.type, target=-1,
+						op=.Load, span=expr.span, type=expr.type, target=ir.INVALID_REF,
-						a=state.local_slots[expr.target], b=-1, diagnostic=-1,
+						a=state.local_slots[local], b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC,
 					})
-				} else if expr.target >= 0 && expr.target < len(state.local_values) &&
+				} else if local != hir.INVALID_LOCAL && int(local) < len(state.local_values) &&
-				          state.local_values[expr.target] >= 0 {
+				          state.local_values[local] != ir.INVALID_INSTRUCTION {
-					last = state.local_values[expr.target]
+					last = state.local_values[local]
 				}
-				if last < 0 {
+				if last == ir.INVALID_INSTRUCTION {
 					last = append_recovery_value(state, expr.span, expr.type, expr.diagnostic)
 				}
 				_ = pop(&stack)
 			case .Global:
-				if expr.target < 0 || expr.target >= len(state.hir_module.globals) {
+				global := hir.as_global(expr.target)
+				if global == hir.INVALID_GLOBAL || int(global) >= len(state.hir_module.globals) {
 					last = append_recovery_value(state, expr.span, expr.type, expr.diagnostic)
 				} else {
 					last = append_instruction(state, ir.Instruction{
-						op=.Load_Global, span=expr.span, type=expr.type, target=expr.target,
+						op=.Load_Global, span=expr.span, type=expr.type, target=ir.global_ref(ir.Global_Id(global)),
-						a=-1, b=-1, diagnostic=-1,
+						a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC,
 					})
 				}
 				_ = pop(&stack)
@@ -135,12 +142,13 @@ lower_expr :: proc(state: ^State, expr_id: int) -> int {
 				stack[frame_index].stage = 2
 				append(&stack, Lower_Expr_Frame{expr=expr.left})
 			case .Call:
-				if expr.target < 0 || expr.target >= len(state.hir_module.functions) {
+				function := hir.as_function(expr.target)
+				if function == hir.INVALID_FUNCTION || int(function) >= len(state.hir_module.functions) {
 					last = append_recovery_value(state, expr.span, expr.type, expr.diagnostic)
 					_ = pop(&stack)
 					continue
 				}
-				stack[frame_index].args = make([]int, len(expr.args), state.allocator)
+				stack[frame_index].args = make([]ir.Instruction_Id, len(expr.args), state.allocator)
 				stack[frame_index].stage = 4
 				if len(expr.args) > 0 {
 					append(&stack, Lower_Expr_Frame{expr=expr.args[0]})
@@ -150,8 +158,8 @@ lower_expr :: proc(state: ^State, expr_id: int) -> int {
 		}
 		if frame.stage == 1 {
 			last = append_instruction(state, ir.Instruction{
-				op=.Widen, span=expr.span, type=expr.type, target=-1,
+				op=.Widen, span=expr.span, type=expr.type, target=ir.INVALID_REF,
-				a=last, b=-1, diagnostic=-1,
+				a=last, b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC,
 			})
 			_ = pop(&stack)
 			continue
@@ -164,8 +172,8 @@ lower_expr :: proc(state: ^State, expr_id: int) -> int {
 		}
 		if frame.stage == 3 {
 			last = append_instruction(state, ir.Instruction{
-				op=.Add_Checked, span=expr.span, type=expr.type, target=-1,
+				op=.Add_Checked, span=expr.span, type=expr.type, target=ir.INVALID_REF,
-				a=frame.left, b=last, diagnostic=-1,
+				a=frame.left, b=last, diagnostic=source.INVALID_DIAGNOSTIC,
 			})
 			_ = pop(&stack)
 			continue
@@ -180,8 +188,8 @@ lower_expr :: proc(state: ^State, expr_id: int) -> int {
 				}
 			}
 			last = append_instruction(state, ir.Instruction{
-				op=.Call, span=expr.span, type=expr.type, target=expr.target,
+				op=.Call, span=expr.span, type=expr.type, target=ir.function_ref(ir.Function_Id(hir.as_function(expr.target))),
-				a=-1, b=-1, args=stack[frame_index].args, diagnostic=-1,
+				a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, args=stack[frame_index].args, diagnostic=source.INVALID_DIAGNOSTIC,
 			})
 			stack[frame_index].args = nil
 			_ = pop(&stack)
@@ -194,8 +202,8 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 	state := State{
 		hir_module=hir_module,
 		allocator=allocator,
-		local_values=make([]int, len(function.locals), allocator),
+		local_values=make([]ir.Instruction_Id, len(function.locals), allocator),
-		local_slots=make([]int, len(function.locals), allocator),
+		local_slots=make([]ir.Instruction_Id, len(function.locals), allocator),
 	}
 	state.instructions.allocator = allocator
 	state.expr_stack.allocator = allocator
@@ -205,17 +213,17 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 		delete(state.expr_stack)
 	}
 	for _, index in state.local_values {
-		state.local_values[index] = -1
+		state.local_values[index] = ir.INVALID_INSTRUCTION
-		state.local_slots[index] = -1
+		state.local_slots[index] = ir.INVALID_INSTRUCTION
 	}
 	for local_id in function.params {
 		param := append_instruction(&state, ir.Instruction{
 			op=.Param,
 			type=function.locals[local_id].type,
-			target=local_id,
+			target=ir.local_ref(ir.Local_Id(local_id)),
-			a=-1,
+			a=ir.INVALID_INSTRUCTION,
-			b=-1,
+			b=ir.INVALID_INSTRUCTION,
-			diagnostic=-1,
+			diagnostic=source.INVALID_DIAGNOSTIC,
 		})
 		state.local_values[local_id] = param
 	}
@@ -225,10 +233,10 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 		switch statement.kind {
 		case .Declaration:
 			value := lower_expr(&state, statement.expr)
-			if statement.local < 0 || statement.local >= len(function.locals) {
+			if statement.local == hir.INVALID_LOCAL || int(statement.local) >= len(function.locals) {
 				append_instruction(&state, ir.Instruction{
 					op=.Trap, span=statement.span, type=types.VOID,
-					target=-1, a=-1, b=-1, diagnostic=statement.diagnostic,
+					target=ir.INVALID_REF, a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, diagnostic=statement.diagnostic,
 				})
 				continue
 			}
@@ -238,34 +246,34 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 					op=.Alloca,
 					span=statement.span,
 					type=local.type,
-					target=statement.local,
+					target=ir.local_ref(ir.Local_Id(statement.local)),
-					a=-1,
+					a=ir.INVALID_INSTRUCTION,
-					b=-1,
+					b=ir.INVALID_INSTRUCTION,
-					diagnostic=-1,
+					diagnostic=source.INVALID_DIAGNOSTIC,
 				})
 				state.local_slots[statement.local] = slot
 				append_instruction(&state, ir.Instruction{
 					op=.Store,
 					span=statement.span,
 					type=local.type,
-					target=-1,
+					target=ir.INVALID_REF,
 					a=slot,
 					b=value,
-					diagnostic=-1,
+					diagnostic=source.INVALID_DIAGNOSTIC,
 				})
 			} else {
 				state.local_values[statement.local] = value
 			}
 		case .Assignment:
 			value := lower_expr(&state, statement.expr)
-			slot := -1
+			slot := ir.INVALID_INSTRUCTION
-			if statement.local >= 0 && statement.local < len(state.local_slots) {
+			if statement.local != hir.INVALID_LOCAL && int(statement.local) < len(state.local_slots) {
 				slot = state.local_slots[statement.local]
 			}
-			if slot < 0 || statement.local < 0 || statement.local >= len(function.locals) {
+			if slot == ir.INVALID_INSTRUCTION || statement.local == hir.INVALID_LOCAL || int(statement.local) >= len(function.locals) {
 				append_instruction(&state, ir.Instruction{
 					op=.Trap, span=statement.span, type=types.VOID,
-					target=-1, a=-1, b=-1, diagnostic=statement.diagnostic,
+					target=ir.INVALID_REF, a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, diagnostic=statement.diagnostic,
 				})
 				continue
 			}
@@ -273,21 +281,21 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 				op=.Store,
 				span=statement.span,
 				type=function.locals[statement.local].type,
-				target=-1,
+				target=ir.INVALID_REF,
 				a=slot,
 				b=value,
-				diagnostic=-1,
+				diagnostic=source.INVALID_DIAGNOSTIC,
 			})
 		case .Return:
-			if statement.expr < 0 {
+			if statement.expr == hir.INVALID_EXPR {
 				append_instruction(&state, ir.Instruction{
 					op=.Return_Void,
 					span=statement.span,
 					type=types.VOID,
-					target=-1,
+					target=ir.INVALID_REF,
-					a=-1,
+					a=ir.INVALID_INSTRUCTION,
-					b=-1,
+					b=ir.INVALID_INSTRUCTION,
-					diagnostic=-1,
+					diagnostic=source.INVALID_DIAGNOSTIC,
 				})
 			} else {
 				value := lower_expr(&state, statement.expr)
@@ -295,10 +303,10 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 					op=.Return,
 					span=statement.span,
 					type=function.result,
-					target=-1,
+					target=ir.INVALID_REF,
 					a=value,
-					b=-1,
+					b=ir.INVALID_INSTRUCTION,
-					diagnostic=-1,
+					diagnostic=source.INVALID_DIAGNOSTIC,
 				})
 			}
 		case .Expression, .Sink:
@@ -308,9 +316,9 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 				op=.Trap,
 				span=statement.span,
 				type=types.VOID,
-				target=-1,
+				target=ir.INVALID_REF,
-				a=-1,
+				a=ir.INVALID_INSTRUCTION,
-				b=-1,
+				b=ir.INVALID_INSTRUCTION,
 				diagnostic=statement.diagnostic,
 			})
 		}
@@ -319,18 +327,18 @@ lower_body :: proc(hir_module: ^hir.Module, function: hir.Function, allocator: m
 	   (state.instructions[len(state.instructions)-1].op != .Return &&
 	    state.instructions[len(state.instructions)-1].op != .Return_Void) {
 		if function.result.kind == .Void {
-			append_instruction(&state, ir.Instruction{op=.Return_Void, type=types.VOID, target=-1, a=-1, b=-1, diagnostic=-1})
+			append_instruction(&state, ir.Instruction{op=.Return_Void, type=types.VOID, target=ir.INVALID_REF, a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC})
 		} else {
 			value := append_instruction(&state, ir.Instruction{
 				op=.Const,
 				type=function.result,
 				integer=sentinel(function.result),
-				target=-1,
+				target=ir.INVALID_REF,
-				a=-1,
+				a=ir.INVALID_INSTRUCTION,
-				b=-1,
+				b=ir.INVALID_INSTRUCTION,
-				diagnostic=-1,
+				diagnostic=source.INVALID_DIAGNOSTIC,
 			})
-			append_instruction(&state, ir.Instruction{op=.Return, type=function.result, target=-1, a=value, b=-1, diagnostic=-1})
+			append_instruction(&state, ir.Instruction{op=.Return, type=function.result, target=ir.INVALID_REF, a=value, b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC})
 		}
 	}
 	return state.instructions[:]
@@ -345,10 +353,10 @@ lower_global_initializer :: proc(hir_module: ^hir.Module, global: hir.Global, al
 	append_instruction(&state, ir.Instruction{
 		op=.Return,
 		type=global.type,
-		target=-1,
+		target=ir.INVALID_REF,
 		a=value,
-		b=-1,
+		b=ir.INVALID_INSTRUCTION,
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 	return state.instructions[:]
 }
@@ -356,6 +364,7 @@ lower_global_initializer :: proc(hir_module: ^hir.Module, global: hir.Global, al
 lower :: proc(hir_module: ^hir.Module, allocator := context.allocator) -> ir.Module {
 	module := ir.init_module(allocator)
 	for global in hir_module.globals {
+		_ = ir.global_id(len(module.globals))
 		append(&module.globals, ir.Global{
 			name=global.name,
 			type=global.type,
@@ -371,6 +380,7 @@ lower :: proc(hir_module: ^hir.Module, allocator := context.allocator) -> ir.Mod
 		for local_id, index in function.params {
 			param_types[index] = function.locals[local_id].type
 		}
+		_ = ir.function_id(len(module.functions))
 		append(&module.functions, ir.Function{
 			link_name=fmt.aprintf("%s", function.link_name, allocator=allocator),
 			calling_convention=.C if function.calling_convention == .C else .Brolang,

compiler/parser/parser.odin

@@ -13,8 +13,8 @@ Parser :: struct {
 	source_file: ^source.Source,
 	diagnostics: ^source.Diagnostics,
 	module:      ast.Module,
-	pkg:         int,
+	pkg:         ast.Package_Id,
-	file:        int,
+	file:        ast.File_Id,
 	cursor:      int,
 	delimiter_depth: int,
 }
@@ -22,10 +22,10 @@ Parser :: struct {
 MAX_EXPRESSION_NESTING :: 256
 
 token_text :: proc(parser: ^Parser, tok: token.Token) -> string {
-	if tok.span.start < 0 || tok.span.end < tok.span.start || tok.span.end > len(parser.source_file.text) {
+	if tok.span.end < tok.span.start || int(tok.span.end) > len(parser.source_file.text) {
 		return ""
 	}
-	return parser.source_file.text[tok.span.start:tok.span.end]
+	return parser.source_file.text[int(tok.span.start):int(tok.span.end)]
 }
 
 span_from :: proc(first, last: source.Span) -> source.Span {
@@ -61,19 +61,19 @@ skip_newlines :: proc(parser: ^Parser) {
 	}
 }
 
-add_expr :: proc(parser: ^Parser, expr: ast.Expr) -> int {
+add_expr :: proc(parser: ^Parser, expr: ast.Expr) -> ast.Expr_Id {
-	id := len(parser.module.exprs)
+	id := ast.expr_id(len(parser.module.exprs))
 	append(&parser.module.exprs, expr)
 	return id
 }
 
-invalid_expr :: proc(parser: ^Parser, span: source.Span, message: string) -> int {
+invalid_expr :: proc(parser: ^Parser, span: source.Span, message: string) -> ast.Expr_Id {
 	id := source.add(parser.diagnostics, span, message)
 	return add_expr(parser, ast.Expr{
 		kind=.Invalid,
 		span=span,
-		left=ast.INVALID_ID,
+		left=ast.INVALID_EXPR,
-		right=ast.INVALID_ID,
+		right=ast.INVALID_EXPR,
 		diagnostic=id,
 	})
 }
@@ -131,7 +131,7 @@ skip_parenthesized :: proc(parser: ^Parser) -> source.Span {
 	return span_from(start.span, end.span)
 }
 
-parse_call :: proc(parser: ^Parser, qualifier: symbol.Id, first, name: token.Token, nesting: int) -> int {
+parse_call :: proc(parser: ^Parser, qualifier: symbol.Id, first, name: token.Token, nesting: int) -> ast.Expr_Id {
 	if nesting >= MAX_EXPRESSION_NESTING {
 		span := skip_parenthesized(parser)
 		return invalid_expr(parser, span, "expression nesting exceeds 256 levels")
@@ -139,7 +139,7 @@ parse_call :: proc(parser: ^Parser, qualifier: symbol.Id, first, name: token.Tok
 	left_paren := advance(parser)
 	parser.delimiter_depth += 1
 	defer parser.delimiter_depth -= 1
-	args: [dynamic]int
+	args: [dynamic]ast.Expr_Id
 	args.allocator = parser.module.allocator
 	skip_newlines(parser)
 	for current(parser).kind != .Right_Paren && current(parser).kind != .Eof {
@@ -162,13 +162,13 @@ parse_call :: proc(parser: ^Parser, qualifier: symbol.Id, first, name: token.Tok
 		qualifier=qualifier,
 		name=name.symbol,
 		args=args[:],
-		left=ast.INVALID_ID,
+		left=ast.INVALID_EXPR,
-		right=ast.INVALID_ID,
+		right=ast.INVALID_EXPR,
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 }
 
-parse_primary :: proc(parser: ^Parser, nesting: int) -> int {
+parse_primary :: proc(parser: ^Parser, nesting: int) -> ast.Expr_Id {
 	tok := current(parser)
 	#partial switch tok.kind {
 	case .Integer:
@@ -181,9 +181,9 @@ parse_primary :: proc(parser: ^Parser, nesting: int) -> int {
 			kind=.Integer,
 			span=tok.span,
 			integer=value,
-			left=ast.INVALID_ID,
+			left=ast.INVALID_EXPR,
-			right=ast.INVALID_ID,
+			right=ast.INVALID_EXPR,
-			diagnostic=-1,
+			diagnostic=source.INVALID_DIAGNOSTIC,
 		})
 	case .Identifier:
 		first := advance(parser)
@@ -204,9 +204,9 @@ parse_primary :: proc(parser: ^Parser, nesting: int) -> int {
 			span=span_from(first.span, name.span),
 			qualifier=qualifier,
 			name=name.symbol,
-			left=ast.INVALID_ID,
+			left=ast.INVALID_EXPR,
-			right=ast.INVALID_ID,
+			right=ast.INVALID_EXPR,
-			diagnostic=-1,
+			diagnostic=source.INVALID_DIAGNOSTIC,
 		})
 	case .Underscore:
 		advance(parser)
@@ -231,8 +231,8 @@ parse_primary :: proc(parser: ^Parser, nesting: int) -> int {
 		return add_expr(parser, ast.Expr{
 			kind=.Invalid,
 			span=tok.span,
-			left=ast.INVALID_ID,
+			left=ast.INVALID_EXPR,
-			right=ast.INVALID_ID,
+			right=ast.INVALID_EXPR,
 			diagnostic=tok.diagnostic,
 		})
 	}
@@ -250,7 +250,7 @@ infix_binding_power :: proc(kind: token.Kind) -> (left, right: int, ok: bool) {
 	return 0, 0, false
 }
 
-parse_expression_bp :: proc(parser: ^Parser, minimum_binding_power, nesting: int) -> int {
+parse_expression_bp :: proc(parser: ^Parser, minimum_binding_power, nesting: int) -> ast.Expr_Id {
 	if nesting > MAX_EXPRESSION_NESTING {
 		tok := current(parser)
 		if tok.kind != .Newline && tok.kind != .Right_Brace && tok.kind != .Eof {
@@ -277,7 +277,7 @@ parse_expression_bp :: proc(parser: ^Parser, minimum_binding_power, nesting: int
 			span=span_from(left_expr.span, right_expr.span),
 			left=left,
 			right=right,
-			diagnostic=-1,
+			diagnostic=source.INVALID_DIAGNOSTIC,
 		})
 		if parser.delimiter_depth > 0 {
 			skip_newlines(parser)
@@ -286,17 +286,17 @@ parse_expression_bp :: proc(parser: ^Parser, minimum_binding_power, nesting: int
 	return left
 }
 
-parse_expression :: proc(parser: ^Parser) -> int {
+parse_expression :: proc(parser: ^Parser) -> ast.Expr_Id {
 	return parse_expression_bp(parser, 0, 0)
 }
 
-finish_statement :: proc(parser: ^Parser, allow_closing_brace := false) -> int {
+finish_statement :: proc(parser: ^Parser, allow_closing_brace := false) -> source.Diagnostic_Id {
 	if current(parser).kind == .Newline {
 		skip_newlines(parser)
-		return -1
+		return source.INVALID_DIAGNOSTIC
 	}
 	if current(parser).kind == .Eof || allow_closing_brace && current(parser).kind == .Right_Brace {
-		return -1
+		return source.INVALID_DIAGNOSTIC
 	}
 	diagnostic := source.add(
 		parser.diagnostics,
@@ -312,33 +312,33 @@ finish_statement :: proc(parser: ^Parser, allow_closing_brace := false) -> int {
 	return diagnostic
 }
 
-parse_return :: proc(parser: ^Parser) -> int {
+parse_return :: proc(parser: ^Parser) -> ast.Stmt_Id {
 	start := advance(parser)
 	skip_newlines(parser)
 	if current(parser).kind == .Underscore {
 		end := advance(parser)
-		id := len(parser.module.statements)
+		id := ast.stmt_id(len(parser.module.statements))
 		append(&parser.module.statements, ast.Stmt{
 			kind=.Return,
 			span=span_from(start.span, end.span),
 			name=end.symbol,
-			expr=ast.INVALID_ID,
+			expr=ast.INVALID_EXPR,
-			diagnostic=-1,
+			diagnostic=source.INVALID_DIAGNOSTIC,
 		})
 		return id
 	}
 	expr := parse_expression(parser)
-	id := len(parser.module.statements)
+	id := ast.stmt_id(len(parser.module.statements))
 	append(&parser.module.statements, ast.Stmt{
 		kind=.Return,
 		span=span_from(start.span, parser.module.exprs[expr].span),
 		expr=expr,
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 	return id
 }
 
-parse_statement :: proc(parser: ^Parser) -> int {
+parse_statement :: proc(parser: ^Parser) -> ast.Stmt_Id {
 	if current(parser).kind == .Keyword_Return {
 		return parse_return(parser)
 	}
@@ -363,7 +363,7 @@ parse_statement :: proc(parser: ^Parser) -> int {
 				kind = .Declaration
 				immutable = operator.kind == .Colon_Colon
 			}
-			id := len(parser.module.statements)
+			id := ast.stmt_id(len(parser.module.statements))
 			append(&parser.module.statements, ast.Stmt{
 				kind=kind,
 				span=span_from(name.span, parser.module.exprs[expr].span),
@@ -371,7 +371,7 @@ parse_statement :: proc(parser: ^Parser) -> int {
 				type=type_syntax,
 				immutable=immutable,
 				expr=expr,
-				diagnostic=-1,
+				diagnostic=source.INVALID_DIAGNOSTIC,
 			})
 			return id
 		}
@@ -379,12 +379,12 @@ parse_statement :: proc(parser: ^Parser) -> int {
 	}
 
 	expr := parse_expression(parser)
-	id := len(parser.module.statements)
+	id := ast.stmt_id(len(parser.module.statements))
 	append(&parser.module.statements, ast.Stmt{
 		kind=.Expression,
 		span=parser.module.exprs[expr].span,
 		expr=expr,
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 	return id
 }
@@ -446,6 +446,7 @@ parse_function :: proc(parser: ^Parser, name: token.Token, c_abi: bool) {
 		if !ended_by_newline && current(parser).kind != .Eof {
 			_ = finish_statement(parser)
 		}
+		_ = ast.function_id(len(parser.module.functions))
 		append(&parser.module.functions, ast.Function{
 			span=span_from(name.span, end.span),
 			name=name.symbol,
@@ -455,23 +456,23 @@ parse_function :: proc(parser: ^Parser, name: token.Token, c_abi: bool) {
 			has_body=false,
 			params=params,
 			result=result,
-			diagnostic=-1,
+			diagnostic=source.INVALID_DIAGNOSTIC,
 		})
 		return
 	}
 	advance(parser)
 
-	body: [dynamic]int
+	body: [dynamic]ast.Stmt_Id
 	body.allocator = parser.module.allocator
 	skip_newlines(parser)
 	for current(parser).kind != .Right_Brace && current(parser).kind != .Eof {
 		append(&body, parse_statement(parser))
-		if diagnostic := finish_statement(parser, true); diagnostic >= 0 {
+		if diagnostic := finish_statement(parser, true); diagnostic != source.INVALID_DIAGNOSTIC {
-			statement_id := len(parser.module.statements)
+			statement_id := ast.stmt_id(len(parser.module.statements))
 			append(&parser.module.statements, ast.Stmt{
 				kind=.Invalid,
 				span=current(parser).span,
-				expr=ast.INVALID_ID,
+				expr=ast.INVALID_EXPR,
 				diagnostic=diagnostic,
 			})
 			append(&body, statement_id)
@@ -482,6 +483,7 @@ parse_function :: proc(parser: ^Parser, name: token.Token, c_abi: bool) {
 		source.add(parser.diagnostics, current(parser).span, "expected '}' after function body")
 		end = func_token
 	}
+	_ = ast.function_id(len(parser.module.functions))
 	append(&parser.module.functions, ast.Function{
 		span=span_from(name.span, end.span),
 		name=name.symbol,
@@ -492,7 +494,7 @@ parse_function :: proc(parser: ^Parser, name: token.Token, c_abi: bool) {
 		params=params,
 		result=result,
 		body=body[:],
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 }
 
@@ -522,28 +524,30 @@ parse_import :: proc(parser: ^Parser, alias: token.Token, start: token.Token) {
 		if path_token.kind != .Newline && path_token.kind != .Eof {
 			advance(parser)
 		}
+		_ = ast.import_id(len(parser.module.imports))
 		append(&parser.module.imports, ast.Import{
 			span=start.span,
 			alias=alias.symbol,
 			pkg=parser.pkg,
 			file=parser.file,
-			target=-1,
+			target=ast.INVALID_PACKAGE,
 			valid=false,
-			diagnostic=-1,
+			diagnostic=source.INVALID_DIAGNOSTIC,
 		})
 		_ = finish_statement(parser)
 		return
 	}
 	advance(parser)
+	_ = ast.import_id(len(parser.module.imports))
 	append(&parser.module.imports, ast.Import{
 		span=span_from(start.span, path_token.span),
 		alias=alias.symbol,
 		path=decode_import_path(parser, path_token),
 		pkg=parser.pkg,
 		file=parser.file,
-		target=-1,
+		target=ast.INVALID_PACKAGE,
 		valid=true,
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 	_ = finish_statement(parser)
 }
@@ -606,6 +610,7 @@ parse_top_level :: proc(parser: ^Parser) {
 	}
 
 	expr := parse_expression(parser)
+	_ = ast.global_id(len(parser.module.globals))
 	append(&parser.module.globals, ast.Global{
 		span=span_from(name.span, parser.module.exprs[expr].span),
 		name=name.symbol,
@@ -614,7 +619,7 @@ parse_top_level :: proc(parser: ^Parser) {
 		type=type_syntax,
 		immutable=operator.kind == .Colon_Colon,
 		expr=expr,
-		diagnostic=-1,
+		diagnostic=source.INVALID_DIAGNOSTIC,
 	})
 	_ = finish_statement(parser)
 }
@@ -644,7 +649,8 @@ parse_into :: proc(
 	source_file: ^source.Source,
 	diagnostics: ^source.Diagnostics,
 	module: ^ast.Module,
-	pkg, file: int,
+	pkg: ast.Package_Id,
+	file: ast.File_Id,
 ) {
 	parser := Parser{
 		tokens=stream,

compiler/source/source.odin

@@ -3,17 +3,48 @@ package source
 import "core:fmt"
 import "core:mem"
 
+Source_Id :: distinct u32
+Diagnostic_Id :: distinct u32
+Offset :: distinct u32
+
+INVALID_SOURCE :: Source_Id(0xffff_ffff)
+INVALID_DIAGNOSTIC :: Diagnostic_Id(0xffff_ffff)
+
+source_id :: proc(index: int) -> Source_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_SOURCE))
+	return Source_Id(index)
+}
+
+diagnostic_id :: proc(index: int) -> Diagnostic_Id {
+	assert(index >= 0 && u64(index) < u64(INVALID_DIAGNOSTIC))
+	return Diagnostic_Id(index)
+}
+
+source_index :: proc(id: Source_Id, count: int) -> (int, bool) {
+	index := int(id)
+	return index, id != INVALID_SOURCE && index < count
+}
+
+diagnostic_index :: proc(id: Diagnostic_Id, count: int) -> (int, bool) {
+	index := int(id)
+	return index, id != INVALID_DIAGNOSTIC && index < count
+}
+
+fits_source_length :: proc(length: u64) -> bool {
+	return length <= u64(max(Offset))
+}
+
 Span :: struct {
-	file:  int,
+	file:  Source_Id,
-	start: int,
+	start: Offset,
-	end:   int,
+	end:   Offset,
 }
 
 Source :: struct {
-	id:          int,
+	id:          Source_Id,
 	path:        string,
 	text:        string,
-	line_starts: []int,
+	line_starts: []Offset,
 }
 
 Store :: struct {
@@ -30,7 +61,7 @@ Diagnostics :: struct {
 	source:    ^Source,
 	store:     ^Store,
 	items:     [dynamic]Diagnostic,
-	index:     map[Diagnostic_Key]int,
+	index:     map[Diagnostic_Key]Diagnostic_Id,
 	allocator: mem.Allocator,
 }
 
@@ -55,20 +86,22 @@ destroy_store :: proc(store: ^Store) {
 	delete(store.items)
 }
 
-make_line_starts :: proc(text: string, allocator: mem.Allocator) -> []int {
+make_line_starts :: proc(text: string, allocator: mem.Allocator) -> []Offset {
-	result: [dynamic]int
+	assert(fits_source_length(u64(len(text))))
+	result: [dynamic]Offset
 	result.allocator = allocator
 	append(&result, 0)
 	for value, offset in transmute([]byte)text {
 		if value == '\n' {
-			append(&result, offset+1)
+			append(&result, Offset(offset+1))
 		}
 	}
 	return result[:]
 }
 
-add_source_owned :: proc(store: ^Store, path: string, text: []byte) -> int {
+add_source_owned :: proc(store: ^Store, path: string, text: []byte) -> Source_Id {
-	id := len(store.items)
+	assert(fits_source_length(u64(len(text))))
+	id := source_id(len(store.items))
 	owned_text := string(text)
 	append(&store.items, Source{
 		id=id,
@@ -79,7 +112,7 @@ add_source_owned :: proc(store: ^Store, path: string, text: []byte) -> int {
 	return id
 }
 
-add_source :: proc(store: ^Store, path, text: string) -> int {
+add_source :: proc(store: ^Store, path, text: string) -> Source_Id {
 	owned := make([]byte, len(text), store.allocator)
 	copy(owned, transmute([]byte)text)
 	return add_source_owned(store, path, owned)
@@ -111,51 +144,51 @@ destroy_diagnostics :: proc(diagnostics: ^Diagnostics) {
 	delete(diagnostics.items)
 }
 
-add :: proc(diagnostics: ^Diagnostics, span: Span, message: string) -> int {
+add :: proc(diagnostics: ^Diagnostics, span: Span, message: string) -> Diagnostic_Id {
 	key := Diagnostic_Key{span=span, message=message}
 	if id, ok := diagnostics.index[key]; ok {
 		return id
 	}
-	id := len(diagnostics.items)
+	id := diagnostic_id(len(diagnostics.items))
 	cloned := fmt.aprintf("%s", message, allocator=diagnostics.allocator)
 	append(&diagnostics.items, Diagnostic{span=span, message=cloned})
 	diagnostics.index[Diagnostic_Key{span=span, message=cloned}] = id
 	return id
 }
 
-addf :: proc(diagnostics: ^Diagnostics, span: Span, format: string, args: ..any) -> int {
+addf :: proc(diagnostics: ^Diagnostics, span: Span, format: string, args: ..any) -> Diagnostic_Id {
 	message := fmt.aprintf(format, ..args, allocator=diagnostics.allocator)
 	key := Diagnostic_Key{span=span, message=message}
 	if id, ok := diagnostics.index[key]; ok {
 		delete(message, diagnostics.allocator)
 		return id
 	}
-	id := len(diagnostics.items)
+	id := diagnostic_id(len(diagnostics.items))
 	append(&diagnostics.items, Diagnostic{span=span, message=message})
 	diagnostics.index[Diagnostic_Key{span=span, message=message}] = id
 	return id
 }
 
-line_and_column :: proc(source_file: ^Source, offset: int) -> (line, column: int) {
+line_and_column :: proc(source_file: ^Source, offset: Offset) -> (line, column: int) {
 	if len(source_file.line_starts) > 0 {
-		limit := min(max(offset, 0), len(source_file.text))
+		limit := min(int(offset), len(source_file.text))
 		low := 0
 		high := len(source_file.line_starts)
 		for low < high {
 			middle := low + (high-low)/2
-			if source_file.line_starts[middle] <= limit {
+			if int(source_file.line_starts[middle]) <= limit {
 				low = middle+1
 			} else {
 				high = middle
 			}
 		}
 		line = max(low, 1)
-		column = limit-source_file.line_starts[line-1]+1
+		column = limit-int(source_file.line_starts[line-1])+1
 		return
 	}
 	line = 1
 	column = 1
-	limit := min(max(offset, 0), len(source_file.text))
+	limit := min(int(offset), len(source_file.text))
 	for byte_value in transmute([]byte)source_file.text[:limit] {
 		if byte_value == '\n' {
 			line += 1
@@ -168,14 +201,17 @@ line_and_column :: proc(source_file: ^Source, offset: int) -> (line, column: int
 }
 
 source_for_span :: proc(diagnostics: ^Diagnostics, span: Span) -> ^Source {
-	if diagnostics.store != nil && span.file >= 0 && span.file < len(diagnostics.store.items) {
+	if diagnostics.store != nil {
-		return &diagnostics.store.items[span.file]
+		if index, ok := source_index(span.file, len(diagnostics.store.items)); ok {
+			return &diagnostics.store.items[index]
+		}
 	}
 	return diagnostics.source
 }
 
-format :: proc(diagnostics: ^Diagnostics, id: int, allocator := context.allocator) -> string {
+format :: proc(diagnostics: ^Diagnostics, id: Diagnostic_Id, allocator := context.allocator) -> string {
-	if id < 0 || id >= len(diagnostics.items) {
+	index, ok := diagnostic_index(id, len(diagnostics.items))
+	if !ok {
 		path := "<unknown>"
 		if diagnostics.source != nil {
 			path = diagnostics.source.path
@@ -184,7 +220,7 @@ format :: proc(diagnostics: ^Diagnostics, id: int, allocator := context.allocato
 		}
 		return fmt.aprintf("%s: compiler recovery error", path, allocator=allocator)
 	}
-	diagnostic := diagnostics.items[id]
+	diagnostic := diagnostics.items[index]
 	source_file := source_for_span(diagnostics, diagnostic.span)
 	if source_file == nil {
 		return fmt.aprintf("<unknown>: error: %s", diagnostic.message, allocator=allocator)
@@ -202,7 +238,7 @@ format :: proc(diagnostics: ^Diagnostics, id: int, allocator := context.allocato
 
 print_all :: proc(diagnostics: ^Diagnostics) {
 	for _, id in diagnostics.items {
-		message := format(diagnostics, id)
+		message := format(diagnostics, diagnostic_id(id))
 		fmt.eprintln(message)
 		delete(message)
 	}

compiler/token/token.odin

@@ -3,7 +3,7 @@ package token
 import "../source"
 import "../symbol"
 
-Kind :: enum {
+Kind :: enum u8 {
 	Invalid,
 	Eof,
 	Newline,
@@ -32,10 +32,10 @@ Kind :: enum {
 }
 
 Token :: struct {
-	kind: Kind,
 	span:       source.Span,
 	symbol:     symbol.Id,
-	diagnostic: int,
+	diagnostic: source.Diagnostic_Id,
+	kind:       Kind,
 }
 
 Stream :: struct {

compiler_tests.odin

@@ -82,6 +82,52 @@ main :: func() void { _ = value }
 	testing.expect_value(t, module.statements[module.functions[0].body[0]].name, sink_symbol)
 }
 
+@(test)
+compact_ids_reserve_invalid_values_and_preserve_layout :: proc(t: ^testing.T) {
+	testing.expect_value(t, size_of(source.Span), 12)
+	testing.expect_value(t, size_of(token.Token), 24)
+
+	source_index, source_ok := source.source_index(source.Source_Id(0), 1)
+	testing.expect_value(t, source_index, 0)
+	testing.expect(t, source_ok)
+	_, source_invalid := source.source_index(source.INVALID_SOURCE, 1)
+	testing.expect(t, !source_invalid)
+	_, source_out_of_bounds := source.source_index(source.Source_Id(1), 1)
+	testing.expect(t, !source_out_of_bounds)
+	diagnostic_index, diagnostic_ok := source.diagnostic_index(source.Diagnostic_Id(0), 1)
+	testing.expect_value(t, diagnostic_index, 0)
+	testing.expect(t, diagnostic_ok)
+	_, diagnostic_invalid := source.diagnostic_index(source.INVALID_DIAGNOSTIC, 1)
+	testing.expect(t, !diagnostic_invalid)
+
+	expr_index, expr_ok := ast.index(ast.Expr_Id(0), ast.INVALID_EXPR, 1)
+	testing.expect_value(t, expr_index, 0)
+	testing.expect(t, expr_ok)
+	_, expr_invalid := ast.index(ast.INVALID_EXPR, ast.INVALID_EXPR, 1)
+	testing.expect(t, !expr_invalid)
+
+	function_index, function_ok := hir.index(hir.Function_Id(0), hir.INVALID_FUNCTION, 1)
+	testing.expect_value(t, function_index, 0)
+	testing.expect(t, function_ok)
+	_, function_invalid := hir.index(hir.INVALID_FUNCTION, hir.INVALID_FUNCTION, 1)
+	testing.expect(t, !function_invalid)
+
+	instruction_index, instruction_ok := ir.index(ir.Instruction_Id(0), ir.INVALID_INSTRUCTION, 1)
+	testing.expect_value(t, instruction_index, 0)
+	testing.expect(t, instruction_ok)
+	_, instruction_invalid := ir.index(ir.INVALID_INSTRUCTION, ir.INVALID_INSTRUCTION, 1)
+	testing.expect(t, !instruction_invalid)
+
+	spec_index, spec_ok := checker.spec_index(checker.Spec_Id(0), 1)
+	testing.expect_value(t, spec_index, 0)
+	testing.expect(t, spec_ok)
+	_, spec_invalid := checker.spec_index(checker.INVALID_SPEC, 1)
+	testing.expect(t, !spec_invalid)
+
+	testing.expect(t, source.fits_source_length(u64(0xffff_ffff)))
+	testing.expect(t, !source.fits_source_length(u64(0x1_0000_0000)))
+}
+
 @(test)
 lexer_preserves_newlines_and_skips_comments :: proc(t: ^testing.T) {
 	source_file := source.Source{path="test.bro", text="# comment\nmain :: func() void {}\n"}
@@ -196,7 +242,7 @@ main :: func() void {}
 
 	c_symbol := symbol.intern(&symbols, "c")
 	for tok in stream.items {
-		if tok.span.start < len(text) && text[tok.span.start:tok.span.end] == "c" {
+		if int(tok.span.start) < len(text) && text[int(tok.span.start):int(tok.span.end)] == "c" {
 			testing.expect_value(t, tok.kind, token.Kind.Identifier)
 			testing.expect_value(t, tok.symbol, c_symbol)
 		}
@@ -415,8 +461,8 @@ multi_source_diagnostics_report_the_originating_file :: proc(t: ^testing.T) {
 
 	testing.expect(t, loaded)
 	found := false
-	for _, diagnostic_id in diagnostics.items {
+	for _, diagnostic_index in diagnostics.items {
-		message := source.format(&diagnostics, diagnostic_id)
+		message := source.format(&diagnostics, source.diagnostic_id(diagnostic_index))
 		if strings.contains(message, "/b.bro:2:9:") &&
 		   strings.contains(message, "unknown package alias 'math'") {
 			found = true
@@ -788,7 +834,7 @@ main :: func() void {
 	testing.expect(t, done_id >= 0)
 	testing.expect(t, main_id >= 0)
 	testing.expect_value(t, hir_module.statements[hir_module.functions[done_id].body[0]].kind, hir.Stmt_Kind.Return)
-	testing.expect_value(t, hir_module.statements[hir_module.functions[done_id].body[0]].expr, -1)
+	testing.expect_value(t, hir_module.statements[hir_module.functions[done_id].body[0]].expr, hir.INVALID_EXPR)
 	main := hir_module.functions[main_id]
 	testing.expect_value(t, hir_module.statements[main.body[0]].kind, hir.Stmt_Kind.Expression)
 	testing.expect_value(t, hir_module.statements[main.body[1]].kind, hir.Stmt_Kind.Sink)
@@ -1243,14 +1289,59 @@ maximum_signed_i64_literal_parses_exactly :: proc(t: ^testing.T) {
 	testing.expect_value(t, module.exprs[module.globals[0].expr].integer, i64(9223372036854775807))
 }
 
+@(test)
+malformed_hir_references_lower_to_valid_trapped_llvm :: proc(t: ^testing.T) {
+	hir_module := hir.init_module()
+	defer hir.destroy_module(&hir_module)
+	append(&hir_module.exprs, hir.Expr{
+		kind=.Local,
+		type=types.I8,
+		target=hir.local_ref(hir.INVALID_LOCAL),
+		left=hir.INVALID_EXPR,
+		right=hir.INVALID_EXPR,
+		diagnostic=source.INVALID_DIAGNOSTIC,
+	})
+	append(&hir_module.statements, hir.Stmt{
+		kind=.Expression,
+		expr=hir.Expr_Id(0),
+		local=hir.INVALID_LOCAL,
+		diagnostic=source.INVALID_DIAGNOSTIC,
+	})
+	body := make([]hir.Stmt_Id, 1)
+	body[0] = hir.Stmt_Id(0)
+	append(&hir_module.functions, hir.Function{
+		name=symbol.INVALID,
+		link_name=strings.clone("main"),
+		calling_convention=.C,
+		implementation=.Definition,
+		linkage=.External,
+		is_main=true,
+		result=types.VOID,
+		body=body,
+		diagnostic=source.INVALID_DIAGNOSTIC,
+	})
+	ir_module := lower.lower(&hir_module)
+	defer ir.destroy_module(&ir_module)
+	source_file := source.Source{path="test.bro", text=""}
+	diagnostics := source.init_diagnostics(&source_file)
+	defer source.destroy_diagnostics(&diagnostics)
+	symbols := symbol.init_table()
+	defer symbol.destroy_table(&symbols)
+	text := llvm.emit(&ir_module, &diagnostics, &symbols)
+	defer delete(text)
+
+	testing.expect(t, strings.contains(text, "call void @bro.trap"))
+	testing.expect(t, !strings.contains(text, "%v-1"))
+}
+
 @(test)
 malformed_ir_emits_traps_and_typed_sentinels :: proc(t: ^testing.T) {
 	module := ir.init_module()
 	defer ir.destroy_module(&module)
 	instructions := make([]ir.Instruction, 3)
-	instructions[0] = ir.Instruction{op=.Store, type=types.I8, a=-1, b=-1, diagnostic=-1}
+	instructions[0] = ir.Instruction{op=.Store, type=types.I8, a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC}
-	instructions[1] = ir.Instruction{op=.Add_Checked, type=types.I32, a=-1, b=-1, diagnostic=-1}
+	instructions[1] = ir.Instruction{op=.Add_Checked, type=types.I32, a=ir.INVALID_INSTRUCTION, b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC}
-	instructions[2] = ir.Instruction{op=.Return, type=types.I32, a=1, b=-1, diagnostic=-1}
+	instructions[2] = ir.Instruction{op=.Return, type=types.I32, a=ir.Instruction_Id(1), b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC}
 	append(&module.functions, ir.Function{
 		link_name=strings.clone("main"),
 		calling_convention=.C,
@@ -1335,14 +1426,14 @@ deep_global_cycle_detection_uses_iterative_dfs :: proc(t: ^testing.T) {
 
 	for id in 0..<count {
 		append(&ast_module.globals, ast.Global{name=name})
-		dependencies: [dynamic]int
+		dependencies: [dynamic]hir.Global_Id
 		dependencies.allocator = context.allocator
-		append(&dependencies, (id+1)%count)
+		append(&dependencies, hir.global_id((id+1)%count))
 		append(&state.module.globals, hir.Global{name=name, dependencies=dependencies})
 	}
 	states := make([]u8, count)
 	defer delete(states)
-	checker.detect_global_cycles_visit(&state, 0, states)
+	checker.detect_global_cycles_visit(&state, hir.Global_Id(0), states)
 
 	testing.expect_value(t, len(diagnostics.items), 1)
 	for global in state.module.globals {