# HG changeset patch
# User Denis Laxalde <denis.laxalde@logilab.fr>
# Date 1475053588 -7200
# Node ID dabbb2a4a49398b6873c334aec711ebf14746b43
# Parent  bc864c80ec3613eb241aa85f73680c52fda4a652
[datafeed] Complete the import log even if parser could not be found

It happens that if the parser could not be found, _pull_data() would just
return an empty dict without taking care to close the import log which it just
opened. This leads to misleading information in the user interface where
CWDataImport entities kept accumulating in the "imports" tab of CWSource
primary view without anything else happening.

So:

* log an error message when parser cannot be found
* always close (write logs and set "end_timestamp" attribute) import log when
  leaving _pull_data().

Closes #15505460.

diff -r bc864c80ec36 -r dabbb2a4a493 cubicweb/server/sources/datafeed.py
--- a/cubicweb/server/sources/datafeed.py	Fri Oct 21 18:03:06 2016 +0200
+++ b/cubicweb/server/sources/datafeed.py	Wed Sep 28 11:06:28 2016 +0200
@@ -238,13 +238,17 @@
                                       source_uris=source_uris,
                                       moved_uris=self.moved_uris(cnx))
         except ObjectNotFound:
-            return {}
-        if parser.process_urls(self.urls, raise_on_error):
-            self.warning("some error occurred, don't attempt to delete entities")
+            msg = 'failed to load parser for %s'
+            importlog.record_error(msg % ('source "%s"' % self.uri))
+            self.error(msg, self)
+            stats = {}
         else:
-            parser.handle_deletion(self.config, cnx, source_uris)
+            if parser.process_urls(self.urls, raise_on_error):
+                self.warning("some error occurred, don't attempt to delete entities")
+            else:
+                parser.handle_deletion(self.config, cnx, source_uris)
+            stats = parser.stats
         self.update_latest_retrieval(cnx)
-        stats = parser.stats
         if stats.get('created'):
             importlog.record_info('added %s entities' % len(stats['created']))
         if stats.get('updated'):
diff -r bc864c80ec36 -r dabbb2a4a493 cubicweb/server/test/unittest_datafeed.py
--- a/cubicweb/server/test/unittest_datafeed.py	Fri Oct 21 18:03:06 2016 +0200
+++ b/cubicweb/server/test/unittest_datafeed.py	Wed Sep 28 11:06:28 2016 +0200
@@ -52,6 +52,10 @@
                 yield self.repo.sources_by_uri[u'ô myfeed']._get_parser(session)
             else:
                 yield
+        # vreg.unregister just pops appobjects from their regid entry,
+        # completely remove the entry to ensure we have no side effect with
+        # this empty entry.
+        del self.vreg['parsers'][AParser.__regid__]
 
     def test(self):
         self.assertIn(u'ô myfeed', self.repo.sources_by_uri)
@@ -150,6 +154,17 @@
             cnx.commit()
         self.assertEqual(dfsource.urls, [u"http://pouet.com", u"http://pouet.org"])
 
+    def test_parser_not_found(self):
+        dfsource = self.repo.sources_by_uri[u'ô myfeed']
+        with self.assertLogs('cubicweb.sources.o myfeed', level='ERROR') as cm:
+            with self.repo.internal_cnx() as cnx:
+                stats = dfsource.pull_data(cnx, force=True)
+                importlog = cnx.find('CWDataImport').one().log
+        self.assertIn('failed to load parser for', cm.output[0])
+        self.assertEqual(stats, {})
+        self.assertIn(u'failed to load parser for source &quot;ô myfeed&quot;',
+                      importlog)
+
 
 class DataFeedConfigTC(CubicWebTC):