GH-77273: Better bytecodes for f-strings (GH-6132)

Author: markshannon

Doc/library/dis.rst

@@ -1465,26 +1465,47 @@ iterations of the loop.
    an argument from two-byte to four-byte.
 
 
-.. opcode:: FORMAT_VALUE (flags)
+.. opcode:: CONVERT_VALUE (oparg)
 
-   Used for implementing formatted literal strings (f-strings).  Pops
-   an optional *fmt_spec* from the stack, then a required *value*.
-   *flags* is interpreted as follows:
+   Convert value to a string, depending on ``oparg``::
 
-   * ``(flags & 0x03) == 0x00``: *value* is formatted as-is.
-   * ``(flags & 0x03) == 0x01``: call :func:`str` on *value* before
-     formatting it.
-   * ``(flags & 0x03) == 0x02``: call :func:`repr` on *value* before
-     formatting it.
-   * ``(flags & 0x03) == 0x03``: call :func:`ascii` on *value* before
-     formatting it.
-   * ``(flags & 0x04) == 0x04``: pop *fmt_spec* from the stack and use
-     it, else use an empty *fmt_spec*.
+      value = STACK.pop()
+      result = func(value)
+      STACK.push(result)
 
-   Formatting is performed using :c:func:`PyObject_Format`.  The
-   result is pushed on the stack.
+   * ``oparg == 1``: call :func:`str` on *value*
+   * ``oparg == 2``: call :func:`repr` on *value*
+   * ``oparg == 3``: call :func:`ascii` on *value*
 
-   .. versionadded:: 3.6
+   Used for implementing formatted literal strings (f-strings).
+
+   .. versionadded:: 3.13
+
+
+.. opcode:: FORMAT_SIMPLE
+
+   Formats the value on top of stack::
+
+      value = STACK.pop()
+      result = value.__format__("")
+      STACK.push(result)
+
+   Used for implementing formatted literal strings (f-strings).
+
+   .. versionadded:: 3.13
+
+.. opcode:: FORMAT_SPEC
+
+   Formats the given value with the given format spec::
+
+      spec = STACK.pop()
+      value = STACK.pop()
+      result = value.__format__(spec)
+      STACK.push(result)
+
+   Used for implementing formatted literal strings (f-strings).
+
+   .. versionadded:: 3.13
 
 
 .. opcode:: MATCH_CLASS (count)

Include/internal/pycore_opcode.h

@@ -25,13 +25,8 @@
 _have_code = (types.MethodType, types.FunctionType, types.CodeType,
               classmethod, staticmethod, type)
 
-FORMAT_VALUE = opmap['FORMAT_VALUE']
-FORMAT_VALUE_CONVERTERS = (
-    (None, ''),
-    (str, 'str'),
-    (repr, 'repr'),
-    (ascii, 'ascii'),
-)
+CONVERT_VALUE = opmap['CONVERT_VALUE']
+
 SET_FUNCTION_ATTRIBUTE = opmap['SET_FUNCTION_ATTRIBUTE']
 FUNCTION_ATTR_FLAGS = ('defaults', 'kwdefaults', 'annotations', 'closure')
 
@@ -579,13 +574,9 @@ def _get_instructions_bytes(code, varname_from_oparg=None,
             elif deop in hascompare:
                 argval = cmp_op[arg>>4]
                 argrepr = argval
-            elif deop == FORMAT_VALUE:
-                argval, argrepr = FORMAT_VALUE_CONVERTERS[arg & 0x3]
-                argval = (argval, bool(arg & 0x4))
-                if argval[1]:
-                    if argrepr:
-                        argrepr += ', '
-                    argrepr += 'with format'
+            elif deop == CONVERT_VALUE:
+                argval = (None, str, repr, ascii)[arg]
+                argrepr = ('', 'str', 'repr', 'ascii')[arg]
             elif deop == SET_FUNCTION_ATTRIBUTE:
                 argrepr = ', '.join(s for i, s in enumerate(FUNCTION_ATTR_FLAGS)
                                     if arg & (1<<i))

Include/opcode.h

@@ -447,7 +447,9 @@ def _write_atomic(path, data, mode=0o666):
 #     Python 3.12b1 3531 (Add PEP 695 changes)
 #     Python 3.13a1 3550 (Plugin optimizer support)
 #     Python 3.13a1 3551 (Compact superinstructions)
-#     Python 3.13a1 3552 (Add SET_FUNCTION_ATTRIBUTE)
+#     Python 3.13a1 3552 (Remove LOAD_FAST__LOAD_CONST and LOAD_CONST__LOAD_FAST)
+#     Python 3.13a1 3553 (Add SET_FUNCTION_ATTRIBUTE)
+#     Python 3.13a1 3554 (more efficient bytecodes for f-strings)
 
 #     Python 3.14 will start with 3600
 
@@ -464,7 +466,7 @@ def _write_atomic(path, data, mode=0o666):
 # Whenever MAGIC_NUMBER is changed, the ranges in the magic_values array
 # in PC/launcher.c must also be updated.
 
-MAGIC_NUMBER = (3552).to_bytes(2, 'little') + b'\r\n'
+MAGIC_NUMBER = (3554).to_bytes(2, 'little') + b'\r\n'
 
 _RAW_MAGIC_NUMBER = int.from_bytes(MAGIC_NUMBER, 'little')  # For import.c
 

Lib/dis.py

@@ -110,6 +110,9 @@ def pseudo_op(name, op, real_ops):
 def_op('CHECK_EXC_MATCH', 36)
 def_op('CHECK_EG_MATCH', 37)
 
+def_op('FORMAT_SIMPLE', 40)
+def_op('FORMAT_WITH_SPEC', 41)
+
 def_op('WITH_EXCEPT_START', 49)
 def_op('GET_AITER', 50)
 def_op('GET_ANEXT', 51)
@@ -213,9 +216,9 @@ def pseudo_op(name, op, real_ops):
 def_op('RESUME', 151)   # This must be kept in sync with deepfreeze.py
 def_op('MATCH_CLASS', 152)
 
-def_op('FORMAT_VALUE', 155)
 def_op('BUILD_CONST_KEY_MAP', 156)
 def_op('BUILD_STRING', 157)
+def_op('CONVERT_VALUE', 158)
 
 def_op('LIST_EXTEND', 162)
 def_op('SET_UPDATE', 163)