[db] Eliminating duplicate key constraint violations

In concurrent storage of two runs containing the same files leads to
duplicate key constraint violation. Some DBMSs can gracefully handle
this issue by supporting "ON CONFLICT DO NOTHING" clause at INSERT
statement.

web/server/codechecker_server/api/mass_store_run.py

@@ -152,8 +152,26 @@ def add_file_record(
         return file_record.id
 
     try:
-        file_record = File(file_path, content_hash, None, None)
-        session.add(file_record)
+        # Parallel storage of runs containing common file paths results a
+        # "duplicate key violation" error. This is handled by CodeChecker, so
+        # practically it causes no problem. The INSERT command of the second
+        # transaction will be thrown away. However, some DB systems are
+        # supporting "ON CONFLICT DO NOTHING" clause in INSERT statement which
+        # solves the same issue gracefully.
+        # TODO: "ON CONFLICT DO NOTHING" feature is available for SQLite engine
+        # too in SQLAlchemy 1.4.
+        if session.bind.dialect.name == 'postgresql':
+            insert_stmt = sqlalchemy.dialects.postgresql.insert(File).values(
+                filepath=file_path,
+                filename=os.path.basename(file_path),
+                content_hash=content_hash).on_conflict_do_nothing(
+                    index_elements=['id'])
+            file_id = session.execute(insert_stmt).inserted_primary_key[0]
+            session.commit()
+            return file_id
+        else:
+            file_record = File(file_path, content_hash, None, None)
+            session.add(file_record)
         session.commit()
     except sqlalchemy.exc.IntegrityError as ex:
         LOG.error(ex)