Commit 70f503e63e255d19e78dac36a1742e18562291c7 | ix

author	dave caruso <me@paperdave.net> 2024-07-21 15:39:07-07:00
committer	dave caruso <me@paperdave.net> 2024-07-21 15:39:07-07:00
log	70f503e63e255d19e78dac36a1742e18562291c7
tree	415663b621113f59d8c9f4ec0b65ba88083cff08
parent	30c8ffc2374e7a7222bca85dac7507bf17abbe51
signature	Commit is signed but in an unrecognized format.

remove $ as a valid identifier character

The casing conventions have not been set, but there is no universe where
we allow $ as a recommended identifier character. we had this initially
because JavaScript allows it, but in practice people just use it in
frameworks to mean "magic". (see jQuery, Svelte, Qwik, etc)

4 files changed, 15 insertions(+), 23 deletions(-)

CHANGELOG.md
src
- core

CHANGELOG.md+3-2

 @@ -1,9 +1,10 @@
 ## 0.1.4
-- feat: implement string identifiers
+- feat: implement basic struct types (#42)
-- feat: implement basic struct types
+- feat: implement string identifiers (#54)
 - feat: include a `README.html` alongside compiler with documentation and code samples.
 - feat(transform): include a warning for leading zero in numbers
+- feat(tokenizer): `$` is no longer valid in identifiers
 - fix(sema): miscompilation using left side of all binary expression for both sides
 - fix(sema): handle recursive function calls
 - fix(sema): quit block analysis when hitting `noreturn`

src/core/Report.zig+1-6

 @@ -1130,12 +1130,7 @@ fn formatIdImpl(
         var i: usize = 0;
         while (i < inner.len) {
             switch (inner[i]) {
-                'a'...'z',
+                'a'...'z', 'A'...'Z', '0'...'9', '_' => {
-                'A'...'Z',
-                '0'...'9',
-                '_',
-                '$',
-                => {
                     i += 1;
                     continue;
                 },

src/core/Tokenizer.zig+3-3

...	@@ -322,7 +322,7 @@ pub fn next(t: *Tokenizer) !?Token {	...	@@ -322,7 +322,7 @@ pub fn next(t: *Tokenizer) !?Token {
322	switch (t.byte()) {	322	switch (t.byte()) {
323	' ', '\t', '\n', '\r' => unreachable, // call skipWhitespace	323	' ', '\t', '\n', '\r' => unreachable, // call skipWhitespace
324		324
325	'a'...'z', 'A'...'Z', '_', '$' => {	325	'a'...'z', 'A'...'Z', '_' => {
326	const start = t.pos;	326	const start = t.pos;
327	t.pos += 1;	327	t.pos += 1;
328	return try t.tokenizeIdentifier(start);	328	return try t.tokenizeIdentifier(start);
...	@@ -563,7 +563,7 @@ fn codePoint(t: Tokenizer) !struct { u3, u21 } {	...	@@ -563,7 +563,7 @@ fn codePoint(t: Tokenizer) !struct { u3, u21 } {
563		563
564	fn isAtIdentifierStart(t: *Tokenizer) !bool {	564	fn isAtIdentifierStart(t: *Tokenizer) !bool {
565	switch (t.byte()) {	565	switch (t.byte()) {
566	'a'...'z', 'A'...'Z', '_', '$' => return true,	566	'a'...'z', 'A'...'Z', '_' => return true,
567	else => return false,	567	else => return false,
568	0x80...0xFF => {	568	0x80...0xFF => {
569	_, const code_point = try t.codePoint();	569	_, const code_point = try t.codePoint();
...	@@ -578,7 +578,7 @@ fn tokenizeIdentifier(t: *Tokenizer, start_byte: u32) !Token {	...	@@ -578,7 +578,7 @@ fn tokenizeIdentifier(t: *Tokenizer, start_byte: u32) !Token {
578		578
579	while (true) {	579	while (true) {
580	switch (t.byte()) {	580	switch (t.byte()) {
581	'a'...'z', 'A'...'Z', '0'...'9', '_', '$' => {	581	'a'...'z', 'A'...'Z', '0'...'9', '_' => {
582	t.pos += 1;	582	t.pos += 1;
583	},	583	},
584	0x80...0xFF => {	584	0x80...0xFF => {

src/core/Transform.zig+8-12

 @@ -6,8 +6,8 @@
 //! - Annotating expression statements with `.discard` and `.ensure_void`
 //! - Flattening nested expressions into instruction sequences.
 //!
-//! Some trivial diagnostics such as unused variables, unknown result locations, and
+//! Some trivial diagnostics such as unused variables, unknown result types,
-//! invalid assignment targets are also reported here.
+//! invalid assignment targets, and other style issues are also reported here.
 gpa: Allocator,
 ast: Ast,
 insts: Inst.List = .{},
 @@ -22,9 +22,9 @@ scopes: std.ArrayListUnmanaged(Scope) = .{},
 report: Report.Builder,
 pub const Error = error{
-    OutOfMemory,
+    /// A `Report` will contain at least one error.
     TransformFail,
-};
+} || Allocator.Error;
 pub fn transform(
     ast: Ast,
 @@ -61,7 +61,7 @@ pub fn transform(
     try t.insts.setCapacity(t.gpa, t.insts.len);
-    return Teg{
+    return .{
         .insts = t.insts.toOwnedSlice(),
         .string_bytes = try t.string_bytes.toOwnedSlice(t.gpa),
         .extra = try t.extra.toOwnedStore(t.gpa),
 @@ -142,18 +142,14 @@ const Scope = struct {
     }
 };
-const ScratchType = enum {
-    ref,
-    index,
-};
 const BlockContext = struct {
     node_base: i32,
     return_target: ?*?Inst.Index = null,
     break_target: ?*BreakTarget = null,
-    /// This struct is always stored on the stack; the idea is to pass
+    /// This struct is always stored on the stack; the idea is to pass mutable
-    /// mutable pointers to it. When `break :brk` is encountered
+    /// pointers to it. When `break :brk` is encountered, it will generate the
+    /// block instruction. This way, redundant blocks are never created.
     const BreakTarget = struct {
         label: ?[]const u8,
         /// This allows `continue` as well as implicit break targeting.