Add json implementation for to_bytes function. (#8723)

The query `select to_bytes(to_json("[1]"))` produces `{b'[1]'}`.

Author: dnwpark

docs/reference/stdlib/bytes.rst

@@ -193,6 +193,7 @@ Bytes
 .. TODO: Function signatures except the first need to be revealed only for v5+
 
 .. eql:function:: std::to_bytes(s: str) -> bytes
+                  std::to_bytes(j: json) -> bytes
                   std::to_bytes(val: int16, endian: Endian) -> bytes
                   std::to_bytes(val: int32, endian: Endian) -> bytes
                   std::to_bytes(val: int64, endian: Endian) -> bytes
@@ -211,6 +212,13 @@ Bytes
         db> select to_bytes('テキスト');
         {b'\xe3\x83\x86\xe3\x82\xad\xe3\x82\xb9\xe3\x83\x88'}
 
+    The json values get converted as strings using UTF-8 encoding:
+
+    .. code-block:: edgeql-repl
+
+        db> select to_bytes(to_json('{"a": 1}'));
+        {b'{"a": 1}'}
+
     The integer values can be encoded as big-endian (most significant bit
     comes first) byte strings:
 

edb/buildmeta.py

@@ -57,7 +57,7 @@
 # The merge conflict there is a nice reminder that you probably need
 # to write a patch in edb/pgsql/patches.py, and then you should preserve
 # the old value.
-EDGEDB_CATALOG_VERSION = 2025_05_12_00_00
+EDGEDB_CATALOG_VERSION = 2025_05_20_00_00
 EDGEDB_MAJOR_VERSION = 7
 
 

edb/lib/std/70-converters.edgeql

@@ -270,6 +270,15 @@ std::to_bytes(s: std::str) -> std::bytes {
 };
 
 
+CREATE FUNCTION
+std::to_bytes(j: std::json) -> std::bytes {
+    CREATE ANNOTATION std::description :=
+        'Convert a json value to a binary UTF-8 string.';
+    SET volatility := 'Immutable';
+    USING (to_bytes(to_str(j)));
+};
+
+
 CREATE SCALAR TYPE
 std::Endian EXTENDING enum<Little, Big>;
 

tests/test_edgeql_functions.py

@@ -3631,6 +3631,57 @@ async def test_edgeql_functions_string_bytes_conversion_error(self):
                 ''',
             )
 
+    async def test_edgeql_functions_json_bytes_conversion(self):
+        cases = [
+            (
+                json.dumps({"a": [1, 2, 3], "b": "foo"}),
+                json.dumps({"a": [1, 2, 3], "b": "foo"}),
+            ),
+
+            # Without ensure_ascii=False, json.dumps will escape unicode to
+            # ascii characters. For example, the character '數' (U+6578)
+            # is encoded as b'\\u6578' instead of b'\xe6\x95\xb8'.
+            # Test that both will be decoded to the correct json and then back
+            # to utf-8 encoded bytes.
+            (
+                json.dumps(
+                    {"數": [1, 2, 3], "言": "你好世界！"},
+                    ensure_ascii=True
+                ),
+                json.dumps(
+                    {"數": [1, 2, 3], "言": "你好世界！"},
+                    ensure_ascii=False
+                ),
+            ),
+            (
+                json.dumps(
+                    {"數": [1, 2, 3], "言": "你好世界！"},
+                    ensure_ascii=False
+                ),
+                json.dumps(
+                    {"數": [1, 2, 3], "言": "你好世界！"},
+                    ensure_ascii=False
+                ),
+            ),
+        ]
+
+        for input, expected in cases:
+            await self.assert_query_result(
+                r'''
+                WITH
+                    input := <bytes>$input,
+                    as_json := to_json(to_str(input)),
+                    as_bytes := to_bytes(as_json),
+                SELECT
+                    as_bytes = <bytes>$expected;
+                ''',
+                {True},
+                variables={
+                    "input": input.encode("utf-8"),
+                    "expected": expected.encode("utf-8"),
+                },
+            )
+
     async def test_edgeql_functions_int_bytes_conversion_01(self):
         # Make sure we can convert the bytes to ints and back
         for num in range(256):