Implement filter for .7z files. Based on rclzip and rcltar

src/filters/rcl7z

@@ -0,0 +1,127 @@
+#!/usr/bin/env python
+
+# 7-Zip file filter for Recoll
+
+# Thanks to Recoll user Martin Ziegler
+# This is a modified version of rclzip, with some help from rcltar
+# Python pylzma library required. See http://www.joachim-bauch.de/projects/pylzma/ 
+
+
+import os
+import fnmatch
+import rclexecm
+
+try:
+    import pylzma
+    from py7zlib import Archive7z
+except:
+    print "RECFILTERROR HELPERNOTFOUND python:pylzma"
+    sys.exit(1);
+
+try:
+    from recoll import rclconfig
+    hasrclconfig = True
+except:
+    hasrclconfig = False
+# As a temporary measure, we also look for rclconfig as a bare
+# module. This is so that the intermediate releases of the filter can
+# ship and use rclconfig.py with the filter code
+if not hasrclconfig:
+    try:
+        import rclconfig
+        hasrclconfig = True
+    except:
+        pass
+
+class SevenZipExtractor:
+    def __init__(self, em):
+        self.currentindex = 0
+        self.em = em
+            
+    def extractone(self, ipath):
+        #self.em.rclog("extractone: [%s]" % ipath)
+        docdata = ""
+        try:
+            docdata = self.sevenzip.getmember(ipath).read()
+            ok = True
+        except Exception, err:
+            self.em.rclog("extractone: failed: [%s]" % err)
+            ok = False
+        iseof = rclexecm.RclExecM.noteof
+        if self.currentindex >= len(self.sevenzip.getnames()) -1:
+            iseof = rclexecm.RclExecM.eofnext
+        if isinstance(ipath, unicode):
+            ipath = ipath.encode("utf-8")
+        return (ok, docdata, ipath, iseof)
+
+    ###### File type handler api, used by rclexecm ---------->
+    def openfile(self, params):
+        filename = params["filename:"]
+        self.currentindex = -1
+        self.skiplist = []
+
+        if hasrclconfig:
+            config = rclconfig.RclConfig()
+            config.setKeyDir(os.path.dirname(filename))
+            skipped = config.getConfParam("zipSkippedNames")
+            if skipped is not None:
+                self.skiplist = skipped.split(" ")
+
+        try:
+            fp = open(filename, 'rb')
+            self.sevenzip = Archive7z(fp)
+            return True
+        except Exception, err:
+            self.em.rclog("openfile: failed: [%s]" % err)
+            return False
+
+    def getipath(self, params):
+        ipath = params["ipath:"]
+        ok, data, ipath, eof = self.extractone(ipath)
+        if ok:
+            return (ok, data, ipath, eof)
+        # Not found. Maybe we need to decode the path?
+        try:
+            ipath = ipath.decode("utf-8")
+            return self.extractone(ipath)
+        except Exception, err:
+            return (ok, data, ipath, eof)
+        
+    def getnext(self, params):
+        if self.currentindex == -1:
+            # Return "self" doc
+            self.currentindex = 0
+            self.em.setmimetype('text/plain')
+            if len(self.sevenzip.getnames()) == 0:
+                eof = rclexecm.RclExecM.eofnext
+            else:
+                eof = rclexecm.RclExecM.noteof
+            return (True, "", "", eof)
+
+        if self.currentindex >= len(self.sevenzip.getnames()):
+            #self.em.rclog("getnext: EOF hit")
+            return (False, "", "", rclexecm.RclExecM.eofnow)
+        else:
+            entryname = self.sevenzip.getnames()[self.currentindex]
+
+            if hasrclconfig and len(self.skiplist) != 0:
+                while self.currentindex < len(self.sevenzip.getnames()):
+                    entryname = self.sevenzip.getnames()[self.currentindex]
+                    for pat in self.skiplist:
+                        if fnmatch.fnmatch(entryname, pat):
+                            entryname = None
+                            break
+                    if entryname is not None:
+                        break
+                    self.currentindex += 1
+                if entryname is None:
+                    return (False, "", "", rclexecm.RclExecM.eofnow)
+                
+            ret= self.extractone(entryname)
+            self.currentindex += 1
+            return ret
+
+# Main program: create protocol handler and extractor and run them
+proto = rclexecm.RclExecM()
+extract = SevenZipExtractor(proto)
+rclexecm.main(proto, extract)