improve flowfilter

1 files changed, 48 insertions, 64 deletions

diff --git a/mitmproxy/flowfilter.py b/mitmproxy/flowfilter.py
index 3f5afb48..b222d2a8 100644
--- a/mitmproxy/flowfilter.py
+++ b/mitmproxy/flowfilter.py
@@ -32,19 +32,17 @@
         rex         Equivalent to ~u rex
 """
 
+import functools
 import re
 import sys
-import functools
+from typing import Callable, ClassVar, Optional, Sequence, Type
+
+import pyparsing as pp
 
+from mitmproxy import flow
 from mitmproxy import http
-from mitmproxy import websocket
 from mitmproxy import tcp
-from mitmproxy import flow
-
-from mitmproxy.utils import strutils
-
-import pyparsing as pp
-from typing import Callable, Sequence, Type, Optional, ClassVar
+from mitmproxy import websocket
 
 
 def only(*types):
@@ -54,7 +52,9 @@ def only(*types):
             if isinstance(flow, types):
                 return fn(self, flow)
             return False
+
         return filter_types
+
     return decorator
 
 
@@ -146,10 +146,10 @@ class _Rex(_Action):
     def __init__(self, expr):
         self.expr = expr
         if self.is_binary:
-            expr = strutils.escaped_str_to_bytes(expr)
+            expr = expr.encode()
         try:
             self.re = re.compile(expr, self.flags)
-        except:
+        except Exception:
             raise ValueError("Cannot compile expression.")
 
 
@@ -336,6 +336,7 @@ class FUrl(_Rex):
     code = "u"
     help = "URL"
     is_binary = False
+
     # FUrl is special, because it can be "naked".
 
     @classmethod
@@ -469,68 +470,51 @@ def _make():
     # Order is important - multi-char expressions need to come before narrow
     # ones.
     parts = []
-    for klass in filter_unary:
-        f = pp.Literal("~%s" % klass.code) + pp.WordEnd()
-        f.setParseAction(klass.make)
+    for cls in filter_unary:
+        f = pp.Literal(f"~{cls.code}") + pp.WordEnd()
+        f.setParseAction(cls.make)
         parts.append(f)
 
-    simplerex = "".join(c for c in pp.printables if c not in "()~'\"")
-    alphdevanagari = pp.pyparsing_unicode.Devanagari.alphas
-    alphcyrillic = pp.pyparsing_unicode.Cyrillic.alphas
-    alphgreek = pp.pyparsing_unicode.Greek.alphas
-    alphchinese = pp.pyparsing_unicode.Chinese.alphas
-    alpharabic = pp.pyparsing_unicode.Arabic.alphas
-    alphhebrew = pp.pyparsing_unicode.Hebrew.alphas
-    alphjapanese = pp.pyparsing_unicode.Japanese.alphas
-    alphkorean = pp.pyparsing_unicode.Korean.alphas
-    alphlatin1 = pp.pyparsing_unicode.Latin1.alphas
-    alphlatinA = pp.pyparsing_unicode.LatinA.alphas
-    alphlatinB = pp.pyparsing_unicode.LatinB.alphas
-
-    rex = pp.Word(simplerex) |\
-        pp.Word(alphcyrillic) |\
-        pp.Word(alphgreek) |\
-        pp.Word(alphchinese) |\
-        pp.Word(alpharabic) |\
-        pp.Word(alphdevanagari) |\
-        pp.Word(alphhebrew) |\
-        pp.Word(alphjapanese) |\
-        pp.Word(alphkorean) |\
-        pp.Word(alphlatin1) |\
-        pp.Word(alphlatinA) |\
-        pp.Word(alphlatinB) |\
-        pp.QuotedString("\"", escChar='\\') |\
-        pp.QuotedString("'", escChar='\\')
-    for klass in filter_rex:
-        f = pp.Literal("~%s" % klass.code) + pp.WordEnd() + rex.copy()
-        f.setParseAction(klass.make)
+    # This is a bit of a hack to simulate Word(pyparsing_unicode.printables),
+    # which has a horrible performance with len(pyparsing.pyparsing_unicode.printables) == 1114060
+    unicode_words = pp.CharsNotIn("()~'\"" + pp.ParserElement.DEFAULT_WHITE_CHARS)
+    unicode_words.skipWhitespace = True
+    regex = (
+            unicode_words
+            | pp.QuotedString('"', escChar='\\')
+            | pp.QuotedString("'", escChar='\\')
+    )
+    for cls in filter_rex:
+        f = pp.Literal(f"~{cls.code}") + pp.WordEnd() + regex.copy()
+        f.setParseAction(cls.make)
         parts.append(f)
 
-    for klass in filter_int:
-        f = pp.Literal("~%s" % klass.code) + pp.WordEnd() + pp.Word(pp.nums)
-        f.setParseAction(klass.make)
+    for cls in filter_int:
+        f = pp.Literal(f"~{cls.code}") + pp.WordEnd() + pp.Word(pp.nums)
+        f.setParseAction(cls.make)
         parts.append(f)
 
     # A naked rex is a URL rex:
-    f = rex.copy()
+    f = regex.copy()
     f.setParseAction(FUrl.make)
     parts.append(f)
 
     atom = pp.MatchFirst(parts)
-    expr = pp.operatorPrecedence(atom,
-                                 [(pp.Literal("!").suppress(),
-                                   1,
-                                   pp.opAssoc.RIGHT,
-                                   lambda x: FNot(*x)),
-                                     (pp.Literal("&").suppress(),
-                                      2,
-                                      pp.opAssoc.LEFT,
-                                      lambda x: FAnd(*x)),
-                                     (pp.Literal("|").suppress(),
-                                      2,
-                                      pp.opAssoc.LEFT,
-                                      lambda x: FOr(*x)),
-                                  ])
+    expr = pp.infixNotation(
+        atom,
+        [(pp.Literal("!").suppress(),
+          1,
+          pp.opAssoc.RIGHT,
+          lambda x: FNot(*x)),
+         (pp.Literal("&").suppress(),
+          2,
+          pp.opAssoc.LEFT,
+          lambda x: FAnd(*x)),
+         (pp.Literal("|").suppress(),
+          2,
+          pp.opAssoc.LEFT,
+          lambda x: FOr(*x)),
+         ])
     expr = pp.OneOrMore(expr)
     return expr.setParseAction(lambda x: FAnd(x) if len(x) != 1 else x)
 
@@ -570,15 +554,15 @@ def match(flt, flow):
 help = []
 for a in filter_unary:
     help.append(
-        ("~%s" % a.code, a.help)
+        (f"~{a.code}", a.help)
     )
 for b in filter_rex:
     help.append(
-        ("~%s regex" % b.code, b.help)
+        (f"~{b.code} regex", b.help)
     )
 for c in filter_int:
     help.append(
-        ("~%s int" % c.code, c.help)
+        (f"~{c.code} int", c.help)
     )
 help.sort()
 help.extend(