wip

Author: rtfeldman

src/fmt/fmt2.zig

@@ -237,7 +237,7 @@ pub fn binOp(self: *const Ast, idx: Node.Idx) collections.NodeSlices(Node.Idx).B
 
 /// Special accessor for arrow nodes which may have multiple parameters
 /// Returns the first parameter and return type as a BinOp for compatibility
-pub fn arrowBinOp(self: *const Ast, idx: Node.Idx) collections.NodeSlices(Node.Idx).BinOp {
+fn arrowBinOp(self: *const Ast, idx: Node.Idx) collections.NodeSlices(Node.Idx).BinOp {
     const node_tag = self.tag(idx);
     std.debug.assert(node_tag == .binop_thin_arrow or node_tag == .binop_thick_arrow);
 
@@ -275,7 +275,7 @@ pub fn arrowBinOp(self: *const Ast, idx: Node.Idx) collections.NodeSlices(Node.I
 
 /// Returns an iterator over all parameters and return type of an arrow node
 /// The last node in the iteration is the return type, all others are parameters
-pub fn arrowNodes(self: *const Ast, idx: Node.Idx) collections.NodeSlices(Node.Idx).Iterator {
+fn arrowNodes(self: *const Ast, idx: Node.Idx) collections.NodeSlices(Node.Idx).Iterator {
     const node_tag = self.tag(idx);
     std.debug.assert(node_tag == .binop_thin_arrow or node_tag == .binop_thick_arrow);
 
@@ -291,15 +291,15 @@ pub fn arrowNodes(self: *const Ast, idx: Node.Idx) collections.NodeSlices(Node.I
 }
 
 /// Given the idx to a lambda, return the region of just its args (the `| ... |` including the pipes)
-pub fn lambdaArgsRegion(self: *const Ast, idx: Node.Idx) Region {
+fn lambdaArgsRegion(self: *const Ast, idx: Node.Idx) Region {
     // This function returns the region of lambda args
     // Since we now store complete regions, just return the lambda's region
     // This function isn't actually used anywhere in the codebase
     return self.getRegion(idx);
 }
 
 /// Given the idx to a BinOp, return the region of just its symbol (e.g. "*" or "==") etc.
-pub fn binOpSymbolRegion(self: *const Ast, idx: Node.Idx) Region {
+fn binOpSymbolRegion(self: *const Ast, idx: Node.Idx) Region {
     const binop = self.binOp(idx);
 
     // To find the binop symbol itself, we scan from lhs_end to either rhs_start or first whitespace.

src/parse/AST2.zig

@@ -24,9 +24,15 @@ const Ident = base.Ident;
 const MAX_PARSE_DIAGNOSTICS: usize = 1_000;
 const MAX_PARSE_STACK_SIZE: usize = 10_000;
 
-/// Temporary helper to create a Region from start and end positions
-/// TODO: Update all callers to compute proper end positions
+/// Helper to create a Region from start and end positions
 fn makeRegion(start: Position, end: Position) Region {
+    // In debug mode, catch bugs where end is before start
+    if (std.debug.runtime_safety) {
+        if (end.offset < start.offset) {
+            std.debug.print("makeRegion error: end.offset ({}) < start.offset ({})\n", .{ end.offset, start.offset });
+            std.debug.assert(end.offset >= start.offset);
+        }
+    }
     return Region{ .start = start, .end = end };
 }
 
@@ -564,7 +570,7 @@ fn parseModuleApply(self: *Parser) Error!Node.Idx {
         close_paren_region = self.currentRegion();
         self.advance();
     } else {
-        // Use last position as approximation
+        // Use current position for missing close paren
         close_paren_region = makeRegion(self.currentPosition(), self.currentPosition());
     }
 
@@ -689,12 +695,15 @@ pub fn pushMalformed(self: *Parser, tag: AST.Diagnostic.Tag, start_pos: Position
         }
     } else start_pos;
 
-    const end_pos = self.currentPosition();
+    var end_pos = self.currentPosition();
 
     if (self.peek() != .EndOfFile) {
         self.advance();
+        end_pos = self.currentPosition();
     }
 
+    std.debug.assert(end_pos.offset >= actual_start_pos.offset);
+
     // Only add diagnostics if we haven't hit the limit
     if (self.diagnostics.items.len < MAX_PARSE_DIAGNOSTICS) {
         try self.pushDiagnostic(tag, actual_start_pos, end_pos);
@@ -860,6 +869,7 @@ fn parseAppHeader(self: *Parser, start_pos: Position) Error!AST.Header {
                 self.advance(); // consume 'platform' keyword
 
                 // Parse provides list after platform keyword
+                const provides_start = self.currentPosition();
                 self.expect(.OpenSquare) catch {
                     try self.pushDiagnostic(.header_expected_open_square, field_start, self.currentPosition());
                     continue;
@@ -872,8 +882,7 @@ fn parseAppHeader(self: *Parser, start_pos: Position) Error!AST.Header {
                 };
 
                 // Create the provides list node (use block as a container for the exposed items)
-                // The region for the provides list spans from the current position to the last exposed item
-                const provides_start = self.currentPosition();
+                // The region for the provides list spans from the start to the last exposed item
                 const provides_region = if (provides_idx != @as(collections.NodeSlices(Node.Idx).Idx, @enumFromInt(0))) blk: {
                     // Get the region of the provides list (should have items)
                     var iter = self.ast.node_slices.nodes(&provides_idx);
@@ -885,9 +894,15 @@ fn parseAppHeader(self: *Parser, start_pos: Position) Error!AST.Header {
                         const last_region = self.ast.nodes.fieldItem(.region, @as(collections.SafeMultiList(Node).Idx, @enumFromInt(@intFromEnum(last))));
                         break :blk makeRegion(provides_start, last_region.end);
                     } else {
-                        break :blk makeRegion(provides_start, provides_start);
+                        // Empty list, use current position for end
+                        const end_pos = self.currentPosition();
+                        break :blk makeRegion(provides_start, end_pos);
                     }
-                } else makeRegion(provides_start, provides_start);
+                } else blk: {
+                    // No items, use current position for end
+                    const end_pos = self.currentPosition();
+                    break :blk makeRegion(provides_start, end_pos);
+                };
                 const provides_node = try self.ast.appendNode(self.gpa, provides_region, .block, .{ .nodes = provides_idx });
 
                 // Create the platform binop: string_value platform provides_node
@@ -1850,7 +1865,7 @@ fn parseListLiteral(self: *Parser) Error!Node.Idx {
         self.advance();
     } else {
         try self.pushDiagnostic(.expected_expr_close_square_or_comma, start_pos, self.currentPosition());
-        // Use the current position as an approximation for error recovery
+        // Use current position for missing close bracket
         close_bracket_region = makeRegion(self.currentPosition(), self.currentPosition());
     }
 
@@ -2183,10 +2198,10 @@ fn parseMatch(self: *Parser) Error!Node.Idx {
         const branch_node = self.ast.nodes.get(branch_idx);
 
         if (branch_node.tag != .binop_thick_arrow) {
-            // Not a branch - restore position and break
-            // Actually we can't easily restore position, so just break
-            // This means we might have consumed a token that's not part of the match
-            // TODO: Better error recovery here
+            // Not a branch - we've parsed something that's not a valid match branch
+            // Since we don't backtrack, treat it as a malformed branch and report error
+            // Then break out of branch parsing
+            _ = try self.pushMalformed(.expr_unexpected_token, branch_node.region.start);
             break;
         }
 
@@ -2387,12 +2402,12 @@ pub fn parseTypeAnno(self: *Parser) Error!Node.Idx {
 }
 
 // Operator precedence
-const BindingPower = struct {
+pub const BindingPower = struct {
     left: u8,
     right: u8,
 };
 
-fn getBindingPower(tag: Token.Tag) BindingPower {
+pub fn getBindingPower(tag: Token.Tag) BindingPower {
     return switch (tag) {
         .OpBar => .{ .left = 10, .right = 11 },
         .OpOr => .{ .left = 20, .right = 21 },
@@ -2412,7 +2427,7 @@ fn getBindingPower(tag: Token.Tag) BindingPower {
     };
 }
 
-fn tokenToBinOpTag(tag: Token.Tag) ?Node.Tag {
+pub fn tokenToBinOpTag(tag: Token.Tag) ?Node.Tag {
     return switch (tag) {
         .OpAssign => .binop_equals,
         .OpEquals => .binop_double_equals,

src/parse/Parser2.zig

@@ -522,7 +522,6 @@ pub const Cursor = struct {
     pos: u32,
     messages: []Diagnostic,
     message_count: u32,
-    tab_width: u8 = 4, // TODO: make this configurable
 
     /// Initialize a Cursor with the given input buffer and a pre-allocated messages slice.
     pub fn init(buf: []const u8, messages: []Diagnostic) Cursor {

src/parse/tokenize_iter.zig

@@ -17,11 +17,7 @@ KwApp OpenCurly LowerIdent OpColon String KwPlatform OpenSquare LowerIdent OpBan
 ~~~
 # FORMATTED
 ~~~roc
-app # This comment is here
-{
-	pf: "../main.roc" platform [main],
-	somePkg: "../main.roc",
-}
+app { pf: "../main.roc" platform [main], somePkg: "../main.roc" }
 
 ~~~
 # EXPECTED

test/snapshots/app_header__nonempty_multiline.md

@@ -26,7 +26,8 @@ app # Comment after keyword
 {
 # Comment after packages open
 	pf: "../main.roc" platform [ # Comment after provides open
-main],
+		main
+	],
 # Comment after platform
 	other: "../../other/main.roc",
 }

test/snapshots/app_header__nonempty_multiline__commented.md

@@ -16,11 +16,7 @@ KwApp OpenCurly LowerIdent OpColon String KwPlatform OpenSquare LowerIdent OpBan
 ~~~
 # FORMATTED
 ~~~roc
-app
-{
-	pf: "../main.roc" platform [main],
-	other: "../../other/main.roc",
-}
+app { pf: "../main.roc" platform [main], other: "../../other/main.roc" }
 
 ~~~
 # EXPECTED

test/snapshots/app_header_nonempty_singleline.md

@@ -122,7 +122,7 @@ OpenRound Int OpPlus Int Comma Int OpBinaryMinus Int Comma Int OpStar Int Comma
 NIL
 # PROBLEMS
 **Parse Error**
-at 16:1 to 16:1
+at 16:1 to 16:2
 
 **Parse Error**
 at 16:2 to 16:2

test/snapshots/binops.md

@@ -162,23 +162,14 @@ main = {
 	result2 = Http.post
 	result3 = get
 	result4 = post
-	combined = Str.concat((
-		client,
-		parser,
-		helper,
-		result1,
-		result2,
-		result3,
-		result4,
-		combined,
-	))
+	combined = Str.concat((client, parser, helper, result1, result2, result3, result4, combined))
 }
 ~~~
 # EXPECTED
 NIL
 # PROBLEMS
 **Parse Error**
-at 34:5 to 34:5
+at 34:5 to 35:1
 
 **Parse Error**
 at 23:16 to 35:1