enh(llvm) Improve grammar and align assembly like grammars (#2830)

-  Variables are now highlighted as `variable`
-  `i8`, `i32`, etc now highlighted as `type` 
-  Labels are now highlighted as `symbol` (matching our other grammars)
-  Better comment detection
-  Restore relevance to strings
-  prevent false positives on `;` line endings as "comments"
-  Add very minimal `operator` and `punctuation` support

CHANGES.md

@@ -30,6 +30,7 @@ Language Improvements:
 - enh(php) highlight variables (#2785) [Taufik Nurrohman][]
 - fix(python) Handle comments on decorators (#2804) [Jonathan Sharpe][]
 - enh(diff) improve highlighting of diff for git patches [Florian Bezdeka][]
+- fix(llvm) lots of small improvements and fixes (#2830) [Josh Goebel][]
 - enh(mathematica) Rework entire implementation [Patrick Scheibe][]
   - Correct matching of the many variations of Mathematica's numbers
   - Matching of named-characters aka special symbols like `\[Gamma]`

src/languages/llvm.js

@@ -4,12 +4,67 @@ Author: Michael Rodler <contact@f0rki.at>
 Description: language used as intermediate representation in the LLVM compiler framework
 Website: https://llvm.org/docs/LangRef.html
 Category: assembler
+Audit: 2020
 */
 
+import * as regex from '../lib/regex.js';
+
+/** @type LanguageFn */
 export default function(hljs) {
-  var identifier = '([-a-zA-Z$._][\\w\\-$.]*)';
+  const IDENT_RE = /([-a-zA-Z$._][\w$.-]*)/;
+  const TYPE = {
+    className: 'type',
+    begin: /\bi\d+(?=\s|\b)/
+  };
+  const OPERATOR = {
+    className: 'operator',
+    relevance: 0,
+    begin: /=/
+  };
+  const PUNCTUATION = {
+    className: 'punctuation',
+    relevance: 0,
+    begin: /,/
+  };
+  const NUMBER = {
+    className: 'number',
+    variants: [
+        { begin: /0[xX][a-fA-F0-9]+/ },
+        { begin: /-?\d+(?:[.]\d+)?(?:[eE][-+]?\d+(?:[.]\d+)?)?/ }
+    ],
+    relevance: 0
+  };
+  const LABEL = {
+    className: 'symbol',
+    variants: [
+        { begin: /^\s*[a-z]+:/ }, // labels
+    ],
+    relevance: 0
+  };
+  const VARIABLE = {
+    className: 'variable',
+    variants: [
+      { begin: regex.concat(/%/, IDENT_RE) },
+      { begin: /%\d+/ },
+      { begin: /#\d+/ },
+    ]
+  };
+  const FUNCTION = {
+    className: 'title',
+    variants: [
+      { begin: regex.concat(/@/, IDENT_RE) },
+      { begin: /@\d+/ },
+      { begin: regex.concat(/!/, IDENT_RE) },
+      { begin: regex.concat(/!\d+/, IDENT_RE) },
+      // https://llvm.org/docs/LangRef.html#namedmetadatastructure
+      // obviously a single digit can also be used in this fashion
+      { begin: /!\d+/ }
+    ]
+  };
+
   return {
     name: 'LLVM IR',
+    // TODO: split into different categories of keywords
     keywords:
       'begin end true false declare define global ' +
       'constant private linker_private internal ' +
@@ -49,48 +104,26 @@ export default function(hljs) {
       'extractvalue insertvalue atomicrmw cmpxchg fence ' +
       'argmemonly double',
     contains: [
-      {
-        className: 'keyword',
-        begin: 'i\\d+'
-      },
-      hljs.COMMENT(
-        ';', '\\n', {relevance: 0}
-      ),
-      // Double quote string
+      TYPE,
+      // this matches "empty comments"...
+      // ...because it's far more likely this is a statement terminator in
+      // another language than an actual comment
+      hljs.COMMENT(/;\s*$/, null, { relevance: 0 }),
+      hljs.COMMENT(/;/, /$/),
       hljs.QUOTE_STRING_MODE,
       {
         className: 'string',
         variants: [
           // Double-quoted string
-          { begin: '"', end: '[^\\\\]"' },
-        ],
-        relevance: 0
-      },
-      {
-        className: 'title',
-        variants: [
-          { begin: '@' + identifier },
-          { begin: '@\\d+' },
-          { begin: '!' + identifier },
-          { begin: '!\\d+' + identifier }
+          { begin: /"/, end: /[^\\]"/ },
         ]
       },
-      {
-        className: 'symbol',
-        variants: [
-          { begin: '%' + identifier },
-          { begin: '%\\d+' },
-          { begin: '#\\d+' },
-        ]
-      },
-      {
-        className: 'number',
-        variants: [
-            { begin: '0[xX][a-fA-F0-9]+' },
-            { begin: '-?\\d+(?:[.]\\d+)?(?:[eE][-+]?\\d+(?:[.]\\d+)?)?' }
-        ],
-        relevance: 0
-      },
+      FUNCTION,
+      PUNCTUATION,
+      OPERATOR,
+      VARIABLE,
+      LABEL,
+      NUMBER
     ]
   };
 }

test/markup/llvm/simple.expect.txt

@@ -0,0 +1,19 @@
+<span class="hljs-comment">;; foooo</span>
+<span class="hljs-keyword">define</span> <span class="hljs-type">i32</span> <span class="hljs-title">@mul_add</span>(<span class="hljs-type">i32</span> <span class="hljs-variable">%x</span><span class="hljs-punctuation">,</span> <span class="hljs-type">i32</span> <span class="hljs-variable">%y</span><span class="hljs-punctuation">,</span> <span class="hljs-type">i32</span> <span class="hljs-variable">%z</span>) {
+<span class="hljs-symbol">  entry:</span>
+    <span class="hljs-variable">%tmp</span> <span class="hljs-operator">=</span> <span class="hljs-keyword">mul</span> <span class="hljs-type">i32</span> <span class="hljs-variable">%x</span><span class="hljs-punctuation">,</span> <span class="hljs-variable">%y</span>
+    <span class="hljs-variable">%tmp2</span> <span class="hljs-operator">=</span> <span class="hljs-keyword">add</span> <span class="hljs-type">i32</span> <span class="hljs-variable">%tmp</span><span class="hljs-punctuation">,</span> <span class="hljs-variable">%z</span>
+    <span class="hljs-variable">%tmp3</span> <span class="hljs-operator">=</span> <span class="hljs-keyword">add</span> <span class="hljs-type">i32</span> <span class="hljs-variable">%tmp</span><span class="hljs-punctuation">,</span> <span class="hljs-number">0</span>
+    <span class="hljs-keyword">ret</span> <span class="hljs-type">i32</span> <span class="hljs-variable">%tmp3</span>
+}
+
+<span class="hljs-comment">; Named metadata</span>
+<span class="hljs-title">!0</span> <span class="hljs-operator">=</span> !{<span class="hljs-type">i32</span> <span class="hljs-number">42</span><span class="hljs-punctuation">,</span> <span class="hljs-keyword">null</span><span class="hljs-punctuation">,</span> !<span class="hljs-string">&quot;string&quot;</span>}
+<span class="hljs-title">!foo</span> <span class="hljs-operator">=</span> !{<span class="hljs-title">!0</span>}
+
+<span class="hljs-comment">; Some unnamed metadata nodes, which are referenced by the named metadata.</span>
+<span class="hljs-title">!0</span> <span class="hljs-operator">=</span> !{!<span class="hljs-string">&quot;zero&quot;</span>}
+<span class="hljs-title">!1</span> <span class="hljs-operator">=</span> !{!<span class="hljs-string">&quot;one&quot;</span>}
+<span class="hljs-title">!2</span> <span class="hljs-operator">=</span> !{!<span class="hljs-string">&quot;two&quot;</span>}
+<span class="hljs-comment">; A named metadata.</span>
+<span class="hljs-title">!name</span> <span class="hljs-operator">=</span> !{<span class="hljs-title">!0</span><span class="hljs-punctuation">,</span> <span class="hljs-title">!1</span><span class="hljs-punctuation">,</span> <span class="hljs-title">!2</span>}

test/markup/llvm/simple.txt

@@ -0,0 +1,19 @@
+;; foooo
+define i32 @mul_add(i32 %x, i32 %y, i32 %z) {
+  entry:
+    %tmp = mul i32 %x, %y
+    %tmp2 = add i32 %tmp, %z
+    %tmp3 = add i32 %tmp, 0
+    ret i32 %tmp3
+}
+
+; Named metadata
+!0 = !{i32 42, null, !"string"}
+!foo = !{!0}
+
+; Some unnamed metadata nodes, which are referenced by the named metadata.
+!0 = !{!"zero"}
+!1 = !{!"one"}
+!2 = !{!"two"}
+; A named metadata.
+!name = !{!0, !1, !2}