Add C++ support

CHANGELOG.md

@@ -4,6 +4,9 @@
 
 Added a C parser.
 
+Added a C++ parser. Difftastic prefers the C++ parser for `.h`
+files. Please file a bug if you see issues.
+
 ## 0.9
 
 ### Parsing

README.md

@@ -17,6 +17,7 @@ See [the manual](http://difftastic.wilfred.me.uk/) to get started.
 Difftastic supports the following languages:
 
 * C
+* C++
 * Clojure
 * CSS
 * Elixir

build.rs

@@ -37,6 +37,11 @@ fn main() {
 
     // TODO: build these in parallel.
     build("tree-sitter-c", "vendor/tree-sitter-c-src", &[]);
+    build(
+        "tree-sitter-cpp",
+        "vendor/tree-sitter-cpp-src",
+        &["scanner.cc"],
+    );
     build("tree-sitter-clojure", "vendor/tree-sitter-clojure-src", &[]);
     build(
         "tree-sitter-css",

manual/src/introduction.md

@@ -1,9 +1,10 @@
 # Introduction
 
 Difftastic is a diff tool that understands syntax. It currently
-supports 15 languages:
+supports 16 languages:
 
 * C
+* C++
 * Clojure
 * CSS
 * Elixir

src/tree_sitter_parser.rs

@@ -27,6 +27,7 @@ pub struct TreeSitterConfig {
 extern "C" {
     fn tree_sitter_c() -> Language;
     fn tree_sitter_clojure() -> Language;
+    fn tree_sitter_cpp() -> Language;
     fn tree_sitter_css() -> Language;
     fn tree_sitter_elisp() -> Language;
     fn tree_sitter_elixir() -> Language;
@@ -46,7 +47,7 @@ pub fn from_extension(extension: &OsStr) -> Option<TreeSitterConfig> {
     // TODO: find a nice way to extract name and extension information
     // from the package.json in these parsers.
     match extension.to_string_lossy().borrow() {
-        "c" | "h" => Some(TreeSitterConfig {
+        "c" => Some(TreeSitterConfig {
             name: "C",
             language: unsafe { tree_sitter_c() },
             atom_nodes: (vec!["string_literal", "char_literal"])
@@ -55,6 +56,22 @@ pub fn from_extension(extension: &OsStr) -> Option<TreeSitterConfig> {
             // TODO: Handle array_declarator where [ is the second token.
             open_delimiter_tokens: (vec!["(", "{"]).into_iter().collect(),
         }),
+        // Treat .h as C++ rather than C. This is an arbitrary choice,
+        // but C++ is more widely used than C according to
+        // https://madnight.github.io/githut/
+        //
+        // TODO: allow users to override the association between
+        // extensions and parses.
+        "cc" | "cpp" | "h" | "hpp" => Some(TreeSitterConfig {
+            name: "C++",
+            language: unsafe { tree_sitter_cpp() },
+            // The C++ grammar extends the C grammar, so the node
+            // names are generally the same.
+            atom_nodes: (vec!["string_literal", "char_literal"])
+                .into_iter()
+                .collect(),
+            open_delimiter_tokens: (vec!["(", "{"]).into_iter().collect(),
+        }),
         "bb" | "boot" | "clj" | "cljc" | "clje" | "cljs" | "cljx" | "edn" | "joke" | "joker" => {
             Some(TreeSitterConfig {
                 name: "Clojure",

vendor/tree-sitter-cpp-src

@@ -0,0 +1 @@
+tree-sitter-cpp/src