bpo-37723: Fix performance regression on regular expression parsing. (GH-15030)

Improve performance of sre_parse._uniq function.

Lib/sre_parse.py

@@ -430,13 +430,7 @@ def _escape(source, escape, state):
     raise source.error("bad escape %s" % escape, len(escape))
 
 def _uniq(items):
-    if len(set(items)) == len(items):
-        return items
-    newitems = []
-    for item in items:
-        if item not in newitems:
-            newitems.append(item)
-    return newitems
+    return list(dict.fromkeys(items))
 
 def _parse_sub(source, state, verbose, nested):
     # parse an alternation: a|b|c

Misc/ACKS

@@ -1708,6 +1708,7 @@ Michael Urman
 Hector Urtubia
 Lukas Vacek
 Ville Vainio
+Yann Vaginay
 Andi Vajda
 Case Van Horsen
 John Mark Vandenberg

Misc/NEWS.d/next/Library/2019-07-31-16-49-01.bpo-37723.zq6tw8.rst

@@ -0,0 +1,2 @@
+Fix performance regression on regular expression parsing with huge
+character sets. Patch by Yann Vaginay.