Split out the decode whitespace change to a separate PR

Keep only the encoder floatstr deferral here; the decoder whitespace
skip is contentious (depends on gh-117397 and shows mixed results on
large documents) and moves to its own PR.

Author: gaborbernat

Lib/json/decoder.py

@@ -355,16 +355,10 @@ def decode(self, s, _w=WHITESPACE.match):
         containing a JSON document).
 
         """
-        # Skip the WHITESPACE.match() call (and its match-object allocation)
-        # for the common case where there is no leading whitespace.
-        idx = _w(s, 0).end() if s and s[0] in ' \t\n\r' else 0
-        obj, end = self.raw_decode(s, idx=idx)
-        # Likewise avoid the trailing-whitespace match when the parse already
-        # consumed the whole string.
+        obj, end = self.raw_decode(s, idx=_w(s, 0).end())
+        end = _w(s, end).end()
         if end != len(s):
-            end = _w(s, end).end()
-            if end != len(s):
-                raise JSONDecodeError("Extra data", s, end)
+            raise JSONDecodeError("Extra data", s, end)
         return obj
 
     def raw_decode(self, s, idx=0):

Misc/NEWS.d/next/Library/2026-06-02-15-45-00.gh-issue-150820.W7tpO7.rst

@@ -1,3 +1,3 @@
-Speed up :func:`json.loads` and :func:`json.dumps` for small documents by
-avoiding a redundant whitespace scan on decode and by building the float
-helper only on the Python encoding path. Patch by Bernát Gábor.
+Speed up :func:`json.dumps` for small documents by building the float
+formatting helper only on the slower Python encoding path instead of on every
+call. Patch by Bernát Gábor.