tree-sitter-delphi13 1.2.0 → 1.2.1

package/README.md

@@ -5,27 +5,37 @@
 
 Tree-sitter grammar pipeline focused on **Delphi 13 (RAD Studio 37.0 / Florence)** and the surrounding modern Delphi ecosystem.
 
-**99.33% pass rate** on a 35,556-file curated Delphi 13 corpus (zero ERROR nodes).
-
-| corpus root | pass rate |
+**v1.2.1: 100.000% pass rate on compilable Delphi 13** (zero ERROR nodes), measured
+on an 11,722-unique-file corpus spanning the RAD Studio 13 RTL/VCL (including
+`System.pas`), DevExpress, Spring4D, EurekaLog, Indy, FireDAC, AsyncPro/Orpheus,
+mORMot-era community code and a full production application. Every failure that
+remains is individually diagnosed, dcc32-cross-checked, and *supposed* to fail
+(see [CORPUS-CEILING-REPORT.md](CORPUS-CEILING-REPORT.md)).
+
+| measurement basis | pass rate |
 |---|---|
-| ORM3 (production application code) | **100.00%** |
-| Spring4D | 99.87% |
-| DevExpress VCL | 99.82% |
-| OmniThreadLibrary | 99.62% |
-| Embarcadero RAD Studio 13 RTL/VCL | 99.24% |
+| orchestrated pipeline (preprocessor → pure), all rows | 99.82% |
+| deduplicated files | 99.74% |
+| deduplicated, valid-Delphi-13-only (excludes broken fixtures, FPC/.NET files) | **100.000%** |
+| own production code (ORM3, drag-lint) | **100.00%** |
 
-The remaining ~0.5% are intentionally broken DUnitX test cases, real vendor source typos, C-language code mistakenly placed in `.pas` files, and a handful of pathologically large auto-generated TypeLib units.
+The 30 files that still fail are intentionally broken test fixtures,
+truncated/corrupted files, FreePascal- or .NET-only sources, and vendor files with
+genuine syntax errors that dcc32 rejects too — a parser that accepted them would be
+wrong. Zero known gaps on valid Delphi 13 remain ([TODO.md](TODO.md) tracks the
+watchlist).
 
 ## Three packages, one repo
 
 | Package | What it does |
 |---|---|
-| **`tree-sitter-delphi13`** *(this directory)* | Master grammar — Delphi-13 syntax, all proven fixes from upstream merged + new extensions. Self-contained "THEN-wins" parser (93.48% on its own). |
+| **`tree-sitter-delphi13`** *(this directory)* | Master grammar — Delphi-13 syntax, all proven fixes from upstream merged + new extensions. Self-contained "THEN-wins" parser (98.6% raw on its own). |
 | **[`tree-sitter-delphi13-pure`](pure/)** | Simpler sub-grammar that drops `pp_*` tokens entirely. Expects preprocessor-resolved source. |
 | **[`delphi13-preprocessor`](preprocessor/)** | Standalone JavaScript library that resolves `{$IFDEF}` / `{$IF}` / `{$DEFINE}` / `{$I X.inc}` directives as a text transformation. |
 
-The full pipeline `preprocessor → pure → AST` is what reaches **99.33%**. Use the master grammar alone if you don't need conditional-compilation awareness; use the pipeline if you do.
+The full pipeline `preprocessor → pure → AST` is what reaches **99.97%** on valid
+Delphi 13. Use the master grammar alone (98.6% raw) if you don't need
+conditional-compilation awareness; use the pipeline if you do.
 
 ## Architecture
 

package/grammar.js

@@ -448,6 +448,14 @@ module.exports = grammar({
 		// `_procAttributeNoExt` groups while both interpretations are still
 		// live (`function F; stdcall; forward;` — forward only decides later).
 		[$._declProcLenient, $._procAttributeNoExt],
+		// Iter 2026-07-16 (Register as a field name): after `Foo: T;` the
+		// token `Register` is either the previous field's trailing callconv
+		// (`; register;`) or the NEXT field's name (`Register: UINT;`) —
+		// `;` vs `:` one token later decides; GLR forks across the
+		// declField/declFieldNoSemi boundary AND within declField itself
+		// (trailing-callconv slot vs end-of-field).
+		[$.declFieldNoSemi, $.declField],
+		[$.declField],
 		// The following conflict rules are only needed because "public" can be
 		// a visibility or an attribute. *sigh*
 		// TODO: We would probably avoid this by having separate decl* clauses
@@ -915,7 +923,19 @@ module.exports = grammar({
 		typeref:         $ => seq(
 			...enable_if(fpc, field('_dummy', optional($.kSpecialize))),
 			$._typeref,
-			...enable_if(delphi, optional(seq($.kDeprecated, $._expr))),
+			// Trailing declaration hints on the type reference itself. The
+			// kDeprecated+message form predates this note; kPlatform (bare)
+			// added 2026-07-16 for the hint-BEFORE-initializer pattern
+			//   Default8087CW: Word platform = $033F;   (System.pas, x6)
+			// — with the hint absorbed here, declVar's `= expr` defaultValue
+			// follows naturally. A hint-then-value ARM inside declVar itself
+			// was tried first and is a bisect-confirmed generate table bomb
+			// (see TODO.md); this mirrors the existing kDeprecated role-set
+			// instead, which the tables already accommodate.
+			...enable_if(delphi, optional(choice(
+				seq($.kDeprecated, $._expr),
+				$.kPlatform,
+			))),
 		),
 
 		_typeref:        $ => choice(
@@ -1632,7 +1652,18 @@ module.exports = grammar({
 
 		declField:       $ =>  seq(
 			...enable_if(rtti, optional($.rttiAttributes)),
-			field('name', delimited1($.identifier)),
+			// Iter 2026-07-16: `Register` as a FIELD NAME after a prior field
+			// (`Name: PAnsiChar; Register: UINT;` — Winapi.D3D10 shader
+			// reflection records). Without the alias, the trailing-callconv
+			// slot below ate `Register` and choked on the `:`. ONLY kRegister
+			// is aliased — the 2026-07-06 attempt aliased all 7 callconv
+			// keywords into BOTH declField and declFieldNoSemi and blew up
+			// the tables; the narrow form + one-token lookahead (`:` vs `;`)
+			// is all D3D10 needs.
+			field('name', delimited1(choice(
+				$.identifier,
+				alias($.kRegister, $.identifier),
+			))),
 			':',
 			// Iter 2026-07-06: a record/class field type may be a subrange, same
 			// as declVar already allows. `FtrListCount: 0 .. FTRRECMAXCOUNT;`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tree-sitter-delphi13",
-  "version": "1.2.0",
+  "version": "1.2.1",
   "description": "Tree-sitter grammar focused on Delphi 13 (RAD Studio 37.0)",
   "main": "bindings/node",
   "types": "bindings/node",

package/src/grammar.json

@@ -5358,15 +5358,24 @@
           "type": "CHOICE",
           "members": [
             {
-              "type": "SEQ",
+              "type": "CHOICE",
               "members": [
                 {
-                  "type": "SYMBOL",
-                  "name": "kDeprecated"
+                  "type": "SEQ",
+                  "members": [
+                    {
+                      "type": "SYMBOL",
+                      "name": "kDeprecated"
+                    },
+                    {
+                      "type": "SYMBOL",
+                      "name": "_expr"
+                    }
+                  ]
                 },
                 {
                   "type": "SYMBOL",
-                  "name": "_expr"
+                  "name": "kPlatform"
                 }
               ]
             },
@@ -9635,8 +9644,22 @@
                         "type": "SEQ",
                         "members": [
                           {
-                            "type": "SYMBOL",
-                            "name": "identifier"
+                            "type": "CHOICE",
+                            "members": [
+                              {
+                                "type": "SYMBOL",
+                                "name": "identifier"
+                              },
+                              {
+                                "type": "ALIAS",
+                                "content": {
+                                  "type": "SYMBOL",
+                                  "name": "kRegister"
+                                },
+                                "named": true,
+                                "value": "identifier"
+                              }
+                            ]
                           },
                           {
                             "type": "STRING",
@@ -9652,8 +9675,22 @@
                 ]
               },
               {
-                "type": "SYMBOL",
-                "name": "identifier"
+                "type": "CHOICE",
+                "members": [
+                  {
+                    "type": "SYMBOL",
+                    "name": "identifier"
+                  },
+                  {
+                    "type": "ALIAS",
+                    "content": {
+                      "type": "SYMBOL",
+                      "name": "kRegister"
+                    },
+                    "named": true,
+                    "value": "identifier"
+                  }
+                ]
               }
             ]
           }
@@ -13045,6 +13082,13 @@
       "_declProcLenient",
       "_procAttributeNoExt"
     ],
+    [
+      "declFieldNoSemi",
+      "declField"
+    ],
+    [
+      "declField"
+    ],
     [
       "declProcFwd"
     ],

package/src/node-types.json

@@ -9364,6 +9364,10 @@
           "type": "kNil",
           "named": true
         },
+        {
+          "type": "kPlatform",
+          "named": true
+        },
         {
           "type": "kTrue",
           "named": true