Let RPMFinder handle more than one architecture in a single call to

the script. The list of architectures must be separated by commas.

With this commit we now retrieve the package name from the rpminfo.net
page holding details about the RPM, rather than guessing it based on
the filename.

This commit also improves version handling and package name aliasing.

2 files changed, 71 insertions, 28 deletions

diff --git a/Functions/RPM b/Functions/RPM
index a166fb0..75d6046 100644
--- a/Functions/RPM
+++ b/Functions/RPM
@@ -71,10 +71,10 @@ function thirdparty_search_remotedb() {
     local arch="$2"
     local distroname="$3"
     local distrocode="$4"
-    local result=$(RPMFinder --path="$path" --arch="$arch" --distroname="$distroname" --distrocode="$distrocode")
-    [ -z "$result" ] && [ "$arch" = "noarch" ] && result=$(RPMFinder --path="$path" --arch="$(uname -m)" --distroname="$distroname" --distrocode="$distrocode")
-    [ -z "$result" ] && return
-    echo "$result"
+    if [ "$arch" = "noarch" ]
+    then arch="${arch},$(uname -m)"
+    fi
+    RPMFinder --path="$path" --arch="$arch" --distroname="$distroname" --distrocode="$distrocode"
 }
 
 function thirdparty_uncompress() {
diff --git a/bin/RPMFinder b/bin/RPMFinder
index f8afa79..77b4505 100755
--- a/bin/RPMFinder
+++ b/bin/RPMFinder
@@ -21,6 +21,14 @@ class VersionCmp:
 		if candidate[0].isalpha() and reference[0].isalpha():
 			return self.compare(candidate, reference)
 
+		# Handle majors such as "1:xxx.yyy"
+		if ":" in candidate and ":" in reference:
+			pass
+		elif ":" in candidate:
+			candidate = ":".join(candidate.split(":")[1:])
+		elif ":" in reference:
+			reference = ":".join(reference.split(":")[1:])
+
 		c_parts = candidate.split(".")
 		r_parts = reference.split(".")
 		while True:
@@ -62,24 +70,46 @@ class RPMFind_Parser(HTMLParser):
 	'''
 	def __init__(self):
 		self.tags = []
+		self.attrs = []
+		self.names = []
+		self.infopages = []
 		self.candidates = []
 		HTMLParser.__init__(self)
 
 	def handle_starttag(self, tag, attrs):
 		self.tags.append(tag)
+		self.attrs.append(attrs)
 
 	def handle_endtag(self, tag):
 		self.tags.pop()
+		self.attrs.pop()
 
 	def handle_data(self, data):
 		if len(self.tags) and self.tags[-1] == "a" and data.find(".rpm") >= 0:
 			self.candidates.append(data)
-
-	def get_pkgnames(self):
+		elif len(self.tags) and self.tags[-1] == "a" and data.find(".html") >= 0:
+			# self.attrs[-1] = [("href", "\\'/linux/RPM/fedora/....html\\'")]
+			href = self.attrs[-1][0][1].replace("\\", "").replace("'", "")
+			self.infopages.append(href)
+		elif len(self.tags) and self.tags[-1] == "td" and data.find("Name:") >= 0:
+			pkgname = data.replace("Name:", "").strip()
+			self.names.append(pkgname)
+
+	def get_packages(self):
 		if len(self.candidates) == 0:
 			return ""
 		return self.candidates
 
+	def get_infopage(self):
+		if len(self.infopages) == 0:
+			return ""
+		return self.infopages[0]
+
+	def get_name(self):
+		if len(self.names) == 0:
+			return ""
+		return self.names[0]
+
 
 class RPMFinder:
 	def find(self, path, arch, distroname, distrocode):
@@ -92,24 +122,30 @@ class RPMFinder:
 		self.arch = arch
 		self.distroname = distroname.replace(" ", "+")
 		self.distrocode = distrocode
-		self.name = path
-		for token in [">=", ">", "<=", "<", "="]:
-			self.name = self.name.replace(token, " ")
-		self.name = self.name.split(" ")[0]
-
-		matches = self.__search_rpmfind_net()
-		if len(matches) == 0:
-			return []
-		if any(op in self.path for op in [">", "<", "="]):
-			op, version = self.__get_op_and_version()
-			return self.__filter(matches, op, version)
-		else:
-			return matches
 
-	def __filter(self, matches, op, version):
+		requested_archs = self.arch.split(",")
+		for i,arch in enumerate(requested_archs):
+			infopage, search_results = self.__search_rpmfind_net(arch)
+			name = self.__get_pkgname(infopage)
+			matches = self.__filter_rpmfind_net(search_results, name, arch)
+			if len(matches) == 0 and i == len(requested_archs)-1:
+				# User possibly requested more than one architecture (e.g., "noarch,x86_64")
+				# and we had no exact package name matches. Since the RPM database holds aliases
+				# for several packages we must give a second chance to the results returned
+				# by our call to search_rpmfind_net().
+				matches = search_results
+			if len(matches) > 0:
+				if any(op in self.path for op in [">", "<", "="]):
+					op, version = self.__get_op_and_version()
+					return self.__filter(matches, op, version, name, arch)
+				else:
+					return matches
+		return []
+
+	def __filter(self, matches, op, version, name, arch):
 		filtered = []
 		for match in matches:
-			pkg_version = match.replace(self.name, "").replace("{0}.{1}.rpm".format(self.distrocode, self.arch), "").strip("-").strip(".")
+			pkg_version = match.replace(name, "").replace("{0}.{1}.rpm".format(self.distrocode, arch), "").strip("-").strip(".")
 			sys.stderr.write("package {0}: {1} {2} {3}?\n".format(match, pkg_version, op, version))
 			vcmp = VersionCmp()
 			if op == ">" and vcmp.test(pkg_version, version) > 0:
@@ -139,19 +175,26 @@ class RPMFinder:
 			sys.stderr.write("could not extract op and version from {}\n".format(self.path))
 			return None, None
 
-	def __search_rpmfind_net(self):
+	def __get_pkgname(self, infopage):
+		baseuri = "http://rpmfind.net"
+		html = subprocess.check_output(["wget", "--quiet", "{0}{1}".format(baseuri, infopage), "-O", "-"])
+		htmlparser = RPMFind_Parser()
+		htmlparser.feed(str(html))
+		return htmlparser.get_name()
+
+	def __search_rpmfind_net(self, arch):
 		path = self.path.replace("/", "%2F")
 		baseuri = "http://rpmfind.net/linux/rpm2html/search.php"
-		query = "?query={0}&submit=Search+...&system={1}&arch={2}".format(path, self.distroname, self.arch)
+		query = "?query={0}&submit=Search+...&system={1}&arch={2}".format(path, self.distroname, arch)
 		html = subprocess.check_output(["wget", "--quiet", "{0}{1}".format(baseuri, query), "-O", "-"])
+		htmlparser = RPMFind_Parser()
+		htmlparser.feed(str(html))
+		return htmlparser.get_infopage(), htmlparser.get_packages()
 
+	def __filter_rpmfind_net(self, pkgnames, name, arch):
 		# Compile a regex that catches package names derived from the basename given by self.path.
 		# Example: perl-DBICx when perl-DBI is wanted.
-		regex = re.compile(r"{0}\-[0-9]+.*{1}.{2}.rpm".format(self.name, self.distrocode, self.arch))
-
-		htmlparser = RPMFind_Parser()
-		htmlparser.feed(str(html))
-		pkgnames = htmlparser.get_pkgnames()
+		regex = re.compile(r"{0}\-[0-9]+.*{1}.{2}.rpm".format(name, self.distrocode, arch))
 		result = []
 		for pkgname in pkgnames:
 			if regex.match(pkgname):