Remove source from ModuleEnv (#7839)

remove source from ModuleEnv

src/base/ModuleEnv.zig

@@ -27,20 +27,11 @@ types_store: type_mod.Store,
 /// and then use these line starts to calculate the line number and column number as required.
 /// this is a more compact representation at the expense of extra computation only when generating error diagnostics.
 line_starts: std.ArrayList(u32),
-/// The original source bytes. We use these to generate error diagnostics.
-/// TODO think about how we will manage this using the cache. Should we only
-/// read these when we need them to report an error? instead of keeping all of this in memory.
-/// This implementation here is simple, but let's us progress with working snapshot tests
-/// and we can validate the error messages and region information there.
-source: std.ArrayList(u8),
 
 /// Initialize the module environment.
-pub fn init(gpa: std.mem.Allocator, source_bytes: []const u8) Self {
+pub fn init(gpa: std.mem.Allocator) Self {
     // TODO: maybe wire in smarter default based on the initial input text size.
 
-    var source = std.ArrayList(u8).init(gpa);
-    source.appendSlice(source_bytes) catch |err| exitOnOom(err);
-
     return Self{
         .gpa = gpa,
         .idents = Ident.Store.initCapacity(gpa, 1024),
@@ -48,7 +39,6 @@ pub fn init(gpa: std.mem.Allocator, source_bytes: []const u8) Self {
         .strings = StringLiteral.Store.initCapacityBytes(gpa, 4096),
         .types_store = type_mod.Store.initCapacity(gpa, 2048, 512),
         .line_starts = std.ArrayList(u32).init(gpa),
-        .source = source,
     };
 }
 
@@ -59,7 +49,6 @@ pub fn deinit(self: *Self) void {
     self.strings.deinit(self.gpa);
     self.types_store.deinit();
     self.line_starts.deinit();
-    self.source.deinit();
 }
 
 /// Calculate and store line starts from the source text

src/check/canonicalize/CIR.zig

@@ -28,6 +28,8 @@ const CIR = @This();
 env: *base.ModuleEnv,
 store: NodeStore,
 ingested_files: IngestedFile.List,
+/// Temporary source text used during SExpr generation for region info calculation
+temp_source_for_sexpr: ?[]const u8 = null,
 imports: ModuleImport.Store,
 top_level_defs: Def.Span,
 
@@ -1479,7 +1481,10 @@ pub const ExhaustiveMark = TypeVar;
 /// and write it to the given writer.
 ///
 /// If a single expression is provided we only print that expression
-pub fn toSExprStr(ir: *CIR, writer: std.io.AnyWriter, maybe_expr_idx: ?Expr.Idx) !void {
+pub fn toSExprStr(ir: *CIR, writer: std.io.AnyWriter, maybe_expr_idx: ?Expr.Idx, source: []const u8) !void {
+    // Set temporary source for region info calculation during SExpr generation
+    ir.temp_source_for_sexpr = source;
+    defer ir.temp_source_for_sexpr = null;
     const gpa = ir.env.gpa;
 
     if (maybe_expr_idx) |expr_idx| {
@@ -1527,17 +1532,24 @@ test "NodeStore - init and deinit" {
 /// This is a standalone utility function that takes the source text as a parameter
 /// to avoid storing it in the cacheable IR structure.
 pub fn calcRegionInfo(self: *const CIR, region: Region) base.RegionInfo {
+    const empty = base.RegionInfo{
+        .start_line_idx = 0,
+        .start_col_idx = 0,
+        .end_line_idx = 0,
+        .end_col_idx = 0,
+        .line_text = "",
+    };
+
     // In the Can IR, regions store byte offsets directly, not token indices.
     // We can use these offsets directly to calculate the diagnostic position.
-    const info = base.RegionInfo.position(self.env.source.items, self.env.line_starts.items, region.start.offset, region.end.offset) catch {
+    const source = self.temp_source_for_sexpr orelse {
+        // No source available, return empty region info
+        return empty;
+    };
+
+    const info = base.RegionInfo.position(source, self.env.line_starts.items, region.start.offset, region.end.offset) catch {
         // Return a zero position if we can't calculate it
-        return .{
-            .start_line_idx = 0,
-            .start_col_idx = 0,
-            .end_line_idx = 0,
-            .end_col_idx = 0,
-            .line_text = "",
-        };
+        return empty;
     };
 
     return info;

src/check/check_types/unify.zig

@@ -1785,7 +1785,7 @@ const TestEnv = struct {
     /// slight more verbose setup for each test
     fn init(gpa: std.mem.Allocator) Self {
         const module_env = gpa.create(base.ModuleEnv) catch |e| exitOnOutOfMemory(e);
-        module_env.* = base.ModuleEnv.init(gpa, &[_]u8{});
+        module_env.* = base.ModuleEnv.init(gpa);
         return .{
             .module_env = module_env,
             .types_store = &module_env.types_store,

src/check/parse/tokenize.zig

@@ -1560,7 +1560,7 @@ pub const Tokenizer = struct {
 fn testTokenization(gpa: std.mem.Allocator, input: []const u8, expected: []const Token.Tag) !void {
     var messages: [10]Diagnostic = undefined;
 
-    var env = base.ModuleEnv.init(gpa, input);
+    var env = base.ModuleEnv.init(gpa);
     defer env.deinit();
 
     var tokenizer = Tokenizer.init(&env, input, &messages);
@@ -1578,7 +1578,7 @@ fn testTokenization(gpa: std.mem.Allocator, input: []const u8, expected: []const
 
 /// Assert the invariants of the tokenizer are held.
 pub fn checkTokenizerInvariants(gpa: std.mem.Allocator, input: []const u8, debug: bool) void {
-    var env = base.ModuleEnv.init(gpa, input);
+    var env = base.ModuleEnv.init(gpa);
     defer env.deinit();
 
     // Initial tokenization.

src/coordinate.zig

@@ -480,7 +480,7 @@ fn parseDependenciesFromPackageRoot(
         return .{ .failed_to_read_root_file = err };
     defer gpa.free(contents);
 
-    var env = base.ModuleEnv.init(gpa, contents);
+    var env = base.ModuleEnv.init(gpa);
     defer env.deinit();
 
     var parse_ast = parse.parse(&env, contents);

src/coordinate/ModuleGraph.zig

@@ -112,7 +112,7 @@ fn loadOrCompileCanIr(
         // this is temporary so we can generate error reporting and diagnostics/region info.
         // We should probably be reading the file on demand or something else. Leaving this
         // comment here so we discuss the plan and make the necessary changes.
-        var module_env = base.ModuleEnv.init(gpa, contents);
+        var module_env = base.ModuleEnv.init(gpa);
 
         var parse_ir = parse.parse(&module_env, contents);
         parse_ir.store.emptyScratch();

src/coordinate_simple.zig

@@ -51,7 +51,7 @@ pub fn checkSource(
     source: []const u8,
 ) ![]const CIR.Diagnostic {
     // Initialize the ModuleEnv
-    var module_env = ModuleEnv.init(gpa, source);
+    var module_env = ModuleEnv.init(gpa);
     defer module_env.deinit();
 
     // Parse the source code
@@ -88,7 +88,7 @@ pub fn processSource(
     source: []const u8,
 ) !ProcessResult {
     // Initialize the ModuleEnv
-    var module_env = ModuleEnv.init(gpa, source);
+    var module_env = ModuleEnv.init(gpa);
     defer module_env.deinit();
 
     // Parse the source code

src/fmt.zig

@@ -146,7 +146,7 @@ pub fn formatFilePath(gpa: std.mem.Allocator, base_dir: std.fs.Dir, path: []cons
     };
     defer gpa.free(contents);
 
-    var module_env = base.ModuleEnv.init(gpa, contents);
+    var module_env = base.ModuleEnv.init(gpa);
     defer module_env.deinit();
 
     var parse_ast = parse.parse(&module_env, contents);
@@ -1942,7 +1942,7 @@ pub fn moduleFmtsStable(gpa: std.mem.Allocator, input: []const u8, debug: bool)
 }
 
 fn parseAndFmt(gpa: std.mem.Allocator, input: []const u8, debug: bool) ![]const u8 {
-    var module_env = base.ModuleEnv.init(gpa, input);
+    var module_env = base.ModuleEnv.init(gpa);
     defer module_env.deinit();
 
     var parse_ast = parse.parse(&module_env, input);

src/snapshot.zig

@@ -437,7 +437,7 @@ fn processSnapshotFile(gpa: Allocator, snapshot_path: []const u8, maybe_fuzz_cor
         }
     };
 
-    var module_env = base.ModuleEnv.init(gpa, content.source);
+    var module_env = base.ModuleEnv.init(gpa);
     defer module_env.deinit();
 
     // Parse the source code
@@ -671,7 +671,7 @@ fn processSnapshotFile(gpa: Allocator, snapshot_path: []const u8, maybe_fuzz_cor
         var canonicalized = std.ArrayList(u8).init(gpa);
         defer canonicalized.deinit();
 
-        try can_ir.toSExprStr(canonicalized.writer().any(), maybe_expr_idx);
+        try can_ir.toSExprStr(canonicalized.writer().any(), maybe_expr_idx, content.source);
 
         try writer.writeAll(Section.CANONICALIZE);
         try writer.writeAll("\n");