Shareable constant nodes

kddnewton · kddnewton · commit 473cfed6d004 · 2024-03-14T22:00:38.000-04:00
diff --git a/config.yml b/config.yml
@@ -248,6 +248,7 @@ warnings:
   - INTEGER_IN_FLIP_FLOP
   - INVALID_CHARACTER
   - INVALID_NUMBERED_REFERENCE
+  - INVALID_SHAREABLE_CONSTANT_VALUE
   - KEYWORD_EOL
   - LITERAL_IN_CONDITION_DEFAULT
   - LITERAL_IN_CONDITION_VERBOSE
@@ -667,6 +668,15 @@ flags:
       - name: FORCED_US_ASCII_ENCODING
         comment: "internal bytes forced the encoding to US-ASCII"
     comment: Flags for regular expression and match last line nodes.
+  - name: ShareableConstantNodeFlags
+    values:
+      - name: LITERAL
+        comment: "constant writes that should be modified with shareable constant value literal"
+      - name: EXPERIMENTAL_EVERYTHING
+        comment: "constant writes that should be modified with shareable constant value experimental everything"
+      - name: EXPERIMENTAL_COPY
+        comment: "constant writes that should be modified with shareable constant value experimental copy"
+    comment: Flags for shareable constant nodes.
   - name: StringFlags
     values:
       - name: FORCED_UTF8_ENCODING
@@ -3063,6 +3073,29 @@ nodes:
 
           self
           ^^^^
+  - name: ShareableConstantNode
+    fields:
+      - name: flags
+        type: flags
+        kind: ShareableConstantNodeFlags
+      - name: write
+        type: node
+        kind:
+          - ConstantWriteNode
+          - ConstantAndWriteNode
+          - ConstantOrWriteNode
+          - ConstantOperatorWriteNode
+          - ConstantPathWriteNode
+          - ConstantPathAndWriteNode
+          - ConstantPathOrWriteNode
+          - ConstantPathOperatorWriteNode
+        comment: The constant write that should be modified with the shareability state.
+    comment: |
+      This node wraps a constant write to indicate that when the value is written, it should have its shareability state modified.
+
+          # shareable_constant_value: literal
+          C = { a: 1 }
+          ^^^^^^^^^^^^
   - name: SingletonClassNode
     fields:
       - name: locals
diff --git a/include/prism/parser.h b/include/prism/parser.h
@@ -448,6 +448,13 @@ typedef struct {
     void (*callback)(void *data, pm_parser_t *parser, pm_token_t *token);
 } pm_lex_callback_t;
 
+/** The type of shareable constant value that can be set. */
+typedef uint8_t pm_shareable_constant_value_t;
+static const pm_shareable_constant_value_t PM_SCOPE_SHAREABLE_CONSTANT_NONE = 0x0;
+static const pm_shareable_constant_value_t PM_SCOPE_SHAREABLE_CONSTANT_LITERAL = 0x1;
+static const pm_shareable_constant_value_t PM_SCOPE_SHAREABLE_CONSTANT_EXPERIMENTAL_EVERYTHING = 0x2;
+static const pm_shareable_constant_value_t PM_SCOPE_SHAREABLE_CONSTANT_EXPERIMENTAL_COPY = 0x4;
+
 /**
  * This struct represents a node in a linked list of scopes. Some scopes can see
  * into their parent scopes, while others cannot.
@@ -487,6 +494,12 @@ typedef struct pm_scope {
      */
     int8_t numbered_parameters;
 
+    /**
+     * The current state of constant shareability for this scope. This is
+     * changed by magic shareable_constant_value comments.
+     */
+    pm_shareable_constant_value_t shareable_constant;
+
     /**
      * A boolean indicating whether or not this scope can see into its parent.
      * If closed is true, then the scope cannot see into its parent.
diff --git a/lib/prism/translation/parser/compiler.rb b/lib/prism/translation/parser/compiler.rb
@@ -1423,6 +1423,11 @@ def visit_self_node(node)
           builder.self(token(node.location))
         end
 
+        # A shareable constant.
+        def visit_shareable_constant_node(node)
+          visit(node.write)
+        end
+
         # class << self; end
         # ^^^^^^^^^^^^^^^^^^
         def visit_singleton_class_node(node)
diff --git a/lib/prism/translation/ripper.rb b/lib/prism/translation/ripper.rb
@@ -2858,6 +2858,11 @@ def visit_self_node(node)
         on_var_ref(on_kw("self"))
       end
 
+      # A shareable constant.
+      def visit_shareable_constant_node(node)
+        visit(node.write)
+      end
+
       # class << self; end
       # ^^^^^^^^^^^^^^^^^^
       def visit_singleton_class_node(node)
diff --git a/lib/prism/translation/ruby_parser.rb b/lib/prism/translation/ruby_parser.rb
@@ -1274,6 +1274,11 @@ def visit_self_node(node)
           s(node, :self)
         end
 
+        # A shareable constant.
+        def visit_shareable_constant_node(node)
+          visit(node.write)
+        end
+
         # class << self; end
         # ^^^^^^^^^^^^^^^^^^
         def visit_singleton_class_node(node)
diff --git a/src/prism.c b/src/prism.c
@@ -5731,6 +5731,25 @@ pm_self_node_create(pm_parser_t *parser, const pm_token_t *token) {
     return node;
 }
 
+/**
+ * Allocate and initialize a new ShareableConstantNode node.
+ */
+static pm_shareable_constant_node_t *
+pm_shareable_constant_node_create(pm_parser_t *parser, pm_node_t *write, pm_shareable_constant_value_t value) {
+    pm_shareable_constant_node_t *node = PM_ALLOC_NODE(parser, pm_shareable_constant_node_t);
+
+    *node = (pm_shareable_constant_node_t) {
+        {
+            .type = PM_SHAREABLE_CONSTANT_NODE,
+            .flags = (pm_node_flags_t) value,
+            .location = PM_LOCATION_NODE_VALUE(write)
+        },
+        .write = write
+    };
+
+    return node;
+}
+
 /**
  * Allocate a new SingletonClassNode node.
  */
@@ -6745,6 +6764,7 @@ pm_parser_scope_push(pm_parser_t *parser, bool closed) {
         .locals = { 0 },
         .parameters = PM_SCOPE_PARAMETERS_NONE,
         .numbered_parameters = PM_SCOPE_NUMBERED_PARAMETERS_NONE,
+        .shareable_constant = (closed || parser->current_scope == NULL) ? PM_SCOPE_SHAREABLE_CONSTANT_NONE : parser->current_scope->shareable_constant,
         .closed = closed
     };
 
@@ -6790,6 +6810,27 @@ pm_parser_scope_forwarding_keywords_check(pm_parser_t *parser, const pm_token_t
     pm_parser_scope_forwarding_param_check(parser, token, PM_SCOPE_PARAMETERS_FORWARDING_KEYWORDS, PM_ERR_EXPECT_EXPRESSION_AFTER_SPLAT_HASH);
 }
 
+/**
+ * Get the current state of constant shareability.
+ */
+static inline pm_shareable_constant_value_t
+pm_parser_scope_shareable_constant_get(pm_parser_t *parser) {
+    return parser->current_scope->shareable_constant;
+}
+
+/**
+ * Set the current state of constant shareability. We'll set it on all of the
+ * open scopes so that reads are quick.
+ */
+static void
+pm_parser_scope_shareable_constant_set(pm_parser_t *parser, pm_shareable_constant_value_t shareable_constant) {
+    pm_scope_t *scope = parser->current_scope;
+
+    do {
+        scope->shareable_constant = shareable_constant;
+    } while (!scope->closed && (scope = scope->previous) != NULL);
+}
+
 /**
  * Save the current param name as the return value and set it to the given
  * constant id.
@@ -7345,6 +7386,28 @@ parser_lex_magic_comment(pm_parser_t *parser, bool semantic_token_seen) {
             }
         }
 
+        // If we have hit a ractor pragma, attempt to lex that.
+        uint32_t value_length = (uint32_t) (value_end - value_start);
+        if (key_length == 24 && pm_strncasecmp(key_source, (const uint8_t *) "shareable_constant_value", 24) == 0) {
+            if (value_length == 4 && pm_strncasecmp(value_start, (const uint8_t *) "none", 4) == 0) {
+                pm_parser_scope_shareable_constant_set(parser, PM_SCOPE_SHAREABLE_CONSTANT_NONE);
+            } else if (value_length == 7 && pm_strncasecmp(value_start, (const uint8_t *) "literal", 7) == 0) {
+                pm_parser_scope_shareable_constant_set(parser, PM_SCOPE_SHAREABLE_CONSTANT_LITERAL);
+            } else if (value_length == 23 && pm_strncasecmp(value_start, (const uint8_t *) "experimental_everything", 23) == 0) {
+                pm_parser_scope_shareable_constant_set(parser, PM_SCOPE_SHAREABLE_CONSTANT_EXPERIMENTAL_EVERYTHING);
+            } else if (value_length == 17 && pm_strncasecmp(value_start, (const uint8_t *) "experimental_copy", 17) == 0) {
+                pm_parser_scope_shareable_constant_set(parser, PM_SCOPE_SHAREABLE_CONSTANT_EXPERIMENTAL_COPY);
+            } else {
+                PM_PARSER_WARN_TOKEN_FORMAT(
+                    parser,
+                    parser->current,
+                    PM_WARN_INVALID_SHAREABLE_CONSTANT_VALUE,
+                    (int) value_length,
+                    (const char *) value_start
+                );
+            }
+        }
+
         // When we're done, we want to free the string in case we had to
         // allocate memory for it.
         pm_string_free(&key);
@@ -11991,6 +12054,21 @@ parse_target_validate(pm_parser_t *parser, pm_node_t *target) {
     return result;
 }
 
+/**
+ * Potentially wrap a constant write node in a shareable constant node depending
+ * on the current state.
+ */
+static pm_node_t *
+parse_shareable_constant_write(pm_parser_t *parser, pm_node_t *write) {
+    pm_shareable_constant_value_t shareable_constant = pm_parser_scope_shareable_constant_get(parser);
+
+    if (shareable_constant != PM_SCOPE_SHAREABLE_CONSTANT_NONE) {
+        return (pm_node_t *) pm_shareable_constant_node_create(parser, write, shareable_constant);
+    }
+
+    return write;
+}
+
 /**
  * Convert the given node into a valid write node.
  */
@@ -12005,15 +12083,17 @@ parse_write(pm_parser_t *parser, pm_node_t *target, pm_token_t *operator, pm_nod
             pm_node_destroy(parser, target);
             return (pm_node_t *) node;
         }
-        case PM_CONSTANT_PATH_NODE:
-            return (pm_node_t *) pm_constant_path_write_node_create(parser, (pm_constant_path_node_t *) target, operator, value);
+        case PM_CONSTANT_PATH_NODE: {
+            pm_node_t *node = (pm_node_t *) pm_constant_path_write_node_create(parser, (pm_constant_path_node_t *) target, operator, value);
+            return parse_shareable_constant_write(parser, node);
+        }
         case PM_CONSTANT_READ_NODE: {
-            pm_constant_write_node_t *node = pm_constant_write_node_create(parser, (pm_constant_read_node_t *) target, operator, value);
+            pm_node_t *node = (pm_node_t *) pm_constant_write_node_create(parser, (pm_constant_read_node_t *) target, operator, value);
             if (context_def_p(parser)) {
-                pm_parser_err_node(parser, (pm_node_t *) node, PM_ERR_WRITE_TARGET_IN_METHOD);
+                pm_parser_err_node(parser, node, PM_ERR_WRITE_TARGET_IN_METHOD);
             }
             pm_node_destroy(parser, target);
-            return (pm_node_t *) node;
+            return parse_shareable_constant_write(parser, node);
         }
         case PM_BACK_REFERENCE_READ_NODE:
         case PM_NUMBERED_REFERENCE_READ_NODE:
@@ -17954,16 +18034,18 @@ parse_expression_infix(pm_parser_t *parser, pm_node_t *node, pm_binding_power_t
                     parser_lex(parser);
 
                     pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_AMPAMPEQ);
-                    return (pm_node_t *) pm_constant_path_and_write_node_create(parser, (pm_constant_path_node_t *) node, &token, value);
+                    pm_node_t *write = (pm_node_t *) pm_constant_path_and_write_node_create(parser, (pm_constant_path_node_t *) node, &token, value);
+
+                    return parse_shareable_constant_write(parser, write);
                 }
                 case PM_CONSTANT_READ_NODE: {
                     parser_lex(parser);
 
                     pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_AMPAMPEQ);
-                    pm_node_t *result = (pm_node_t *) pm_constant_and_write_node_create(parser, (pm_constant_read_node_t *) node, &token, value);
+                    pm_node_t *write = (pm_node_t *) pm_constant_and_write_node_create(parser, (pm_constant_read_node_t *) node, &token, value);
 
                     pm_node_destroy(parser, node);
-                    return result;
+                    return parse_shareable_constant_write(parser, write);
                 }
                 case PM_INSTANCE_VARIABLE_READ_NODE: {
                     parser_lex(parser);
@@ -18065,16 +18147,18 @@ parse_expression_infix(pm_parser_t *parser, pm_node_t *node, pm_binding_power_t
                     parser_lex(parser);
 
                     pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_PIPEPIPEEQ);
-                    return (pm_node_t *) pm_constant_path_or_write_node_create(parser, (pm_constant_path_node_t *) node, &token, value);
+                    pm_node_t *write = (pm_node_t *) pm_constant_path_or_write_node_create(parser, (pm_constant_path_node_t *) node, &token, value);
+
+                    return parse_shareable_constant_write(parser, write);
                 }
                 case PM_CONSTANT_READ_NODE: {
                     parser_lex(parser);
 
                     pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_PIPEPIPEEQ);
-                    pm_node_t *result = (pm_node_t *) pm_constant_or_write_node_create(parser, (pm_constant_read_node_t *) node, &token, value);
+                    pm_node_t *write = (pm_node_t *) pm_constant_or_write_node_create(parser, (pm_constant_read_node_t *) node, &token, value);
 
                     pm_node_destroy(parser, node);
-                    return result;
+                    return parse_shareable_constant_write(parser, write);
                 }
                 case PM_INSTANCE_VARIABLE_READ_NODE: {
                     parser_lex(parser);
@@ -18186,16 +18270,18 @@ parse_expression_infix(pm_parser_t *parser, pm_node_t *node, pm_binding_power_t
                     parser_lex(parser);
 
                     pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_OPERATOR);
-                    return (pm_node_t *) pm_constant_path_operator_write_node_create(parser, (pm_constant_path_node_t *) node, &token, value);
+                    pm_node_t *write = (pm_node_t *) pm_constant_path_operator_write_node_create(parser, (pm_constant_path_node_t *) node, &token, value);
+
+                    return parse_shareable_constant_write(parser, write);
                 }
                 case PM_CONSTANT_READ_NODE: {
                     parser_lex(parser);
 
                     pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_OPERATOR);
-                    pm_node_t *result = (pm_node_t *) pm_constant_operator_write_node_create(parser, (pm_constant_read_node_t *) node, &token, value);
+                    pm_node_t *write = (pm_node_t *) pm_constant_operator_write_node_create(parser, (pm_constant_read_node_t *) node, &token, value);
 
                     pm_node_destroy(parser, node);
-                    return result;
+                    return parse_shareable_constant_write(parser, write);
                 }
                 case PM_INSTANCE_VARIABLE_READ_NODE: {
                     parser_lex(parser);
diff --git a/templates/src/diagnostic.c.erb b/templates/src/diagnostic.c.erb
@@ -327,6 +327,7 @@ static const pm_diagnostic_data_t diagnostic_messages[PM_DIAGNOSTIC_ID_MAX] = {
     [PM_WARN_FLOAT_OUT_OF_RANGE]                = { "Float %.*s%s out of range", PM_WARNING_LEVEL_VERBOSE },
     [PM_WARN_INTEGER_IN_FLIP_FLOP]              = { "integer literal in flip-flop", PM_WARNING_LEVEL_DEFAULT },
     [PM_WARN_INVALID_CHARACTER]                 = { "invalid character syntax; use %s%s%s", PM_WARNING_LEVEL_DEFAULT },
+    [PM_WARN_INVALID_SHAREABLE_CONSTANT_VALUE]  = { "invalid value for shareable_constant_value: %.*s", PM_WARNING_LEVEL_VERBOSE },
     [PM_WARN_INVALID_NUMBERED_REFERENCE]        = { "'%.*s' is too big for a number variable, always nil", PM_WARNING_LEVEL_DEFAULT },
     [PM_WARN_KEYWORD_EOL]                       = { "`%.*s` at the end of line without an expression", PM_WARNING_LEVEL_VERBOSE },
     [PM_WARN_LITERAL_IN_CONDITION_DEFAULT]      = { "%sliteral in %s", PM_WARNING_LEVEL_DEFAULT },
diff --git a/test/prism/location_test.rb b/test/prism/location_test.rb
@@ -781,6 +781,15 @@ def test_SelfNode
       assert_location(SelfNode, "self")
     end
 
+    def test_ShareableConstantNode
+      source = <<~RUBY
+        # shareable_constant_value: literal
+        C = { foo: 1 }
+      RUBY
+
+      assert_location(ShareableConstantNode, source, 36...50)
+    end
+
     def test_SingletonClassNode
       assert_location(SingletonClassNode, "class << self; end")
     end
@@ -915,8 +924,7 @@ def test_all_tested
 
     def assert_location(kind, source, expected = 0...source.length, **options)
       result = Prism.parse(source, **options)
-      assert_equal [], result.comments
-      assert_equal [], result.errors
+      assert result.success?
 
       node = result.value.statements.body.last
       node = yield node if block_given?
