compiler: Support record spread

Thsi feels even messier than in the interpreter but I am not really sure
how else to do this. Maybe with some better static analysis that
computes record layouts at compile-time.

Author: tekknolagi

compiler.py

@@ -220,9 +220,19 @@ def try_match(self, env: Env, arg: str, pattern: Object, fallthrough: str) -> En
         if isinstance(pattern, Record):
             self._emit(f"if (!is_record({arg})) {{ goto {fallthrough}; }}")
             updates = {}
+            seen_key_indices: list[int] = []
             for key, pattern_value in pattern.data.items():
-                assert not isinstance(pattern_value, Spread), "record spread not yet supported"
+                if isinstance(pattern_value, Spread):
+                    use_spread = True
+                    if pattern_value.name:
+                        num_seen_keys = len(seen_key_indices)
+                        self._emit(
+                            f"size_t seen_keys[{num_seen_keys}] = {{ {', '.join(map(str, seen_key_indices))} }};"
+                        )
+                        updates[pattern_value.name] = self._mktemp(f"record_rest({arg}, seen_keys, {num_seen_keys})")
+                    break
                 key_idx = self.record_key(key)
+                seen_key_indices.append(key_idx)
                 record_value = self._mktemp(f"record_get({arg}, {key_idx})")
                 self._emit(f"if ({record_value} == NULL) {{ goto {fallthrough}; }}")
                 updates.update(self.try_match(env, record_value, pattern_value, fallthrough))

compiler_tests.py

@@ -52,12 +52,17 @@ def test_match_list(self) -> None:
     def test_match_list_spread(self) -> None:
         self.assertEqual(self._run("f [4, 5] . f = | [_, ...xs] -> xs"), "[5]\n")
 
+    def test_match_list_spread_empty(self) -> None:
+        self.assertEqual(self._run("f [4] . f = | [_, ...xs] -> xs"), "[]\n")
+
     def test_match_record(self) -> None:
         self.assertEqual(self._run("f {a = 4, b = 5} . f = | {a = 1, b = 2} -> 3 | {a = 4, b = 5} -> 6"), "6\n")
 
-    @unittest.skip("TODO")
     def test_match_record_spread(self) -> None:
-        self.assertEqual(self._run("f {a=1, b=2, c=3} . f = | {a=1, ...rest} -> rest"), "[5]\n")
+        self.assertEqual(self._run("f {a=1, b=2, c=3} . f = | {a=1, ...rest} -> rest"), "{b = 2, c = 3}\n")
+
+    def test_match_record_spread_empty(self) -> None:
+        self.assertEqual(self._run("f {a=1} . f = | {a=1, ...rest} -> rest"), "{}\n")
 
     def test_match_hole(self) -> None:
         self.assertEqual(self._run("f () . f = | 1 -> 3 | () -> 4"), "4\n")

runtime.c

@@ -471,6 +471,10 @@ struct object* record_get(struct object* record, size_t key) {
   return NULL;
 }
 
+size_t record_num_fields(struct object* record) {
+  return as_record(record)->size;
+}
+
 bool is_string(struct object* obj) {
   if (is_small_string(obj)) {
     return true;
@@ -603,6 +607,34 @@ struct object* list_cons(struct object* item, struct object* list) {
   return result;
 }
 
+bool array_contains(size_t* haystack, size_t size, size_t needle) {
+  for (size_t i = 0; i < size; i++) {
+    if (haystack[i] == needle) {
+      return true;
+    }
+  }
+  return false;
+}
+
+struct object* record_rest(struct object* record, size_t* exclude,
+                           size_t num_excluded) {
+  // NB: This is used in a match expression so it is assumed that all of the
+  // key indices in the exclude array are present in the record and that there
+  // are no duplicates in either the record or the exclude array.
+  HANDLES();
+  GC_PROTECT(record);
+  size_t num_keys = record_num_fields(record);
+  size_t num_result_keys = num_keys - num_excluded;
+  struct object* result = mkrecord(heap, num_result_keys);
+  for (size_t src = 0, dst = 0; dst < num_result_keys; src++) {
+    struct record_field field = as_record(record)->fields[src];
+    if (!array_contains(exclude, num_excluded, field.key)) {
+      record_set(result, dst++, field);
+    }
+  }
+  return result;
+}
+
 struct object* heap_string_concat(struct object* a, struct object* b) {
   uword a_size = string_length(a);
   uword b_size = string_length(b);