Make "this" the implicit context object

Summary:
Today `.` is used as the implicit context object. But there's a problem...

An expression like `(func .)` has the intent of calling `func` with the value of the implicit context object. However, the parser thinks that `.` is a separator for variable lookup. The grammar is just ambiguous here.

The fix is to take a keyword we've already reserved, `this`, and make it also refer to the implicit context object.

Reviewed By: yoney

Differential Revision: D68005178

fbshipit-source-id: d34cf65e350ca70ae6dc061bda7c033bc041d78b

thrift/compiler/whisker/ast.cc

@@ -45,7 +45,7 @@ std::string to_joined_string(
 std::string variable_lookup::chain_string() const {
   return detail::variant_match(
       chain,
-      [](this_ref) -> std::string { return "."; },
+      [](this_ref) -> std::string { return "this"; },
       [](const std::vector<identifier>& ids) -> std::string {
         return to_joined_string(
             ids, '.', [](const identifier& id) -> const std::string& {

thrift/compiler/whisker/parser.cc

@@ -918,12 +918,13 @@ class parser {
         scan};
   }
 
-  // variable-lookup → { "." | (identifier ~ ("." ~ identifier)*) }
+  // variable-lookup → { "." | "this" | (identifier ~ ("." ~ identifier)*) }
   parse_result<ast::variable_lookup> parse_variable_lookup(
       parser_scan_window scan) {
     assert(scan.empty());
     const auto scan_start = scan.start;
-    if (try_consume_token(&scan, tok::dot)) {
+    if (try_consume_token(&scan, tok::dot) ||
+        try_consume_token(&scan, tok::kw_this)) {
       return {
           ast::variable_lookup{
               scan.with_start(scan_start).range(),

thrift/compiler/whisker/test/parser_test.cc

@@ -111,7 +111,7 @@ TEST_F(ParserTest, variable_is_dot) {
   EXPECT_EQ(
       to_string(*ast),
       "root [path/to/test-1.whisker]\n"
-      "|- interpolation <line:1:1, col:7> '.'\n");
+      "|- interpolation <line:1:1, col:7> 'this'\n");
 }
 
 TEST_F(ParserTest, variable_starts_with_dot) {
@@ -803,6 +803,16 @@ TEST_F(ParserTest, let_statement) {
       "| `- expression <line:1:14, col:30> '(not true_value)'\n");
 }
 
+TEST_F(ParserTest, let_statement_with_implicit_context) {
+  auto ast = parse_ast("{{#let foo = (a.b.c this)}}");
+  EXPECT_EQ(
+      to_string(*ast),
+      "root [path/to/test-1.whisker]\n"
+      "|- let-statement <line:1:1, col:28>\n"
+      "| `- identifier 'foo'\n"
+      "| `- expression <line:1:14, col:26> '(a.b.c this)'\n");
+}
+
 TEST_F(ParserTest, let_statement_dotted_name) {
   auto ast = parse_ast("{{#let foo.bar = (not true_value)}}");
   EXPECT_THAT(

thrift/doc/contributions/whisker.md

@@ -171,7 +171,7 @@ There are three types of `expression`s in Whisker:
   * `boolean-literal` — either `true` or `false`.
   * `null-literal` — *exactly* `null`.
 * ***Variable*** — values interpolated from the *context*, like `{{ person.name }}`. Variables allow scoped access into the context using a series of *identifiers* separated by dots (`.`).
-  * Additionally, the special *implicit context* variable, `{{ . }}`, refers to the *context* object itself.
+  * Additionally, the special *implicit context* variable, `{{ . }}` or `{{ this }}`, refers to the *context* object itself.
 * ***Function call*** — [lisp-like](https://en.wikipedia.org/wiki/Lisp_(programming_language)#Syntax_and_semantics) syntax for invoking functions, often referred to as [S-expressions](https://en.wikipedia.org/wiki/S-expression). Functions are defined in the [*data model*](#data-model). Examples:
   * `{{ (uppercase "hello") }}`
   * `{{ (concat (uppercase person.firstName) " " (uppercase person.lastName)) }}`
@@ -216,7 +216,7 @@ interpolation → { "{{" ~ expression ~ "}}" }
 expression    → { literal | variable | function-call }
 
 literal             → { string-literal | i64-literal | boolean-literal | null-literal }
-variable            → { "." | (identifier ~ ("." ~ identifier)*) }
+variable            → { "." | "this" | (identifier ~ ("." ~ identifier)*) }
 function-call       → { "(" ~ (builtin-call | user-defined-call) ~ ")" }
 
 string-literal  → { <see above> }