Rather than reading the dissectors line by line and searching for the

registration and handoff routines in each line, read the whole file into
memory and then do one findall() search. Store the matches in sets to avoid
duplication.

(This also simplifies the number of regexps we're looking for.)

This speeds up the generation of register.c (when there is no cache)
significantly (on my system the time taken drops from ~32 seconds to ~5
seconds).

svn path=/trunk/; revision=44954

1 files changed, 30 insertions, 35 deletions

diff --git a/tools/make-dissector-reg.py b/tools/make-dissector-reg.py
index 6f1efabcfa..a29626bf32 100755
--- a/tools/make-dissector-reg.py
+++ b/tools/make-dissector-reg.py
@@ -83,32 +83,26 @@ if len(filenames) < 1:
 
 # Look through all files, applying the regex to each line.
 # If the pattern matches, save the "symbol" section to the
-# appropriate array.
+# appropriate set.
 regs = {
-	'proto_reg': [],
-	'handoff_reg': [],
-	'wtap_register': [],
+	'proto_reg': set(),
+	'handoff_reg': set(),
+	'wtap_register': set(),
 	}
 
 # For those that don't know Python, r"" indicates a raw string,
 # devoid of Python escapes.
-proto_regex0 = r"^(?P<symbol>proto_register_[_A-Za-z0-9]+)\s*\([^;]+$"
-proto_regex1 = r"void\s+(?P<symbol>proto_register_[_A-Za-z0-9]+)\s*\([^;]+$"
+proto_regex = r"(?P<symbol>proto_register_[_A-Za-z0-9]+)\s*\(\s*void\s*\)"
 
-handoff_regex0 = r"^(?P<symbol>proto_reg_handoff_[_A-Za-z0-9]+)\s*\([^;]+$"
-handoff_regex1 = r"void\s+(?P<symbol>proto_reg_handoff_[_A-Za-z0-9]+)\s*\([^;]+$"
+handoff_regex = r"(?P<symbol>proto_reg_handoff_[_A-Za-z0-9]+)\s*\(\s*void\s*\)"
 
-wtap_reg_regex0 = r"^(?P<symbol>wtap_register_[_A-Za-z0-9]+)\s*\([^;]+$"
-wtap_reg_regex1 = r"void\s+(?P<symbol>wtap_register_[_A-Za-z0-9]+)\s*\([^;]+$"
+wtap_reg_regex = r"(?P<symbol>wtap_register_[_A-Za-z0-9]+)\s*\([^;]+$"
 
 # This table drives the pattern-matching and symbol-harvesting
 patterns = [
-	( 'proto_reg', re.compile(proto_regex0) ),
-	( 'proto_reg', re.compile(proto_regex1) ),
-	( 'handoff_reg', re.compile(handoff_regex0) ),
-	( 'handoff_reg', re.compile(handoff_regex1) ),
-	( 'wtap_register', re.compile(wtap_reg_regex0) ),
-	( 'wtap_register', re.compile(wtap_reg_regex1) ),
+	( 'proto_reg', re.compile(proto_regex, re.MULTILINE) ),
+	( 'handoff_reg', re.compile(handoff_regex, re.MULTILINE) ),
+	( 'wtap_register', re.compile(wtap_reg_regex, re.MULTILINE) ),
 	]
 
 # Open our registration symbol cache
@@ -136,9 +130,9 @@ for filename in filenames:
 		if cur_mtime == cdict['mtime']:
 			cache_hits += 1
 #			print "Pulling %s from cache" % (filename)
-			regs['proto_reg'].extend(cdict['proto_reg'])
-			regs['handoff_reg'].extend(cdict['handoff_reg'])
-			regs['wtap_register'].extend(cdict['wtap_register'])
+			regs['proto_reg'] |= set(cdict['proto_reg'])
+			regs['handoff_reg'] |= set(cdict['handoff_reg'])
+			regs['wtap_register'] |= set(cdict['wtap_register'])
 			file.close()
 			continue
 	# We don't have a cache entry
@@ -151,17 +145,19 @@ for filename in filenames:
 			'wtap_register': [],
 			}
 #	print "Searching %s" % (filename)
-	for line in file.readlines():
-		for action in patterns:
-			regex = action[1]
-			match = regex.search(line)
-			if match:
-				symbol = match.group("symbol")
-				sym_type = action[0]
-				regs[sym_type].append(symbol)
-				if cache is not None:
-#					print "Caching %s for %s: %s" % (sym_type, filename, symbol)
-					cache[filename][sym_type].append(symbol)
+	# Read the whole file into memory
+	contents = file.read()
+	for action in patterns:
+		regex = action[1]
+		for match in regex.finditer(contents):
+			symbol = match.group("symbol")
+			sym_type = action[0]
+			regs[sym_type].add(symbol)
+			if cache is not None:
+#				print "Caching %s for %s: %s" % (sym_type, filename, symbol)
+				cache[filename][sym_type].append(symbol)
+	# We're done with the file contents
+	contets = ""
 	file.close()
 
 
@@ -176,11 +172,10 @@ if len(regs['proto_reg']) < 1:
 	print("No protocol registrations found")
 	sys.exit(1)
 
-# Sort the lists to make them pretty
-# Unique-ify the lists to avoid double registrations
-regs['proto_reg'] = sorted(set(regs['proto_reg']))
-regs['handoff_reg'] = sorted(set(regs['handoff_reg']))
-regs['wtap_register'] = sorted(set(regs['wtap_register']))
+# Convert the sets into sorted lists to make the output pretty
+regs['proto_reg'] = sorted(regs['proto_reg'])
+regs['handoff_reg'] = sorted(regs['handoff_reg'])
+regs['wtap_register'] = sorted(regs['wtap_register'])
 
 reg_code = open(tmp_filename, "w")