topic Performance problem with the search in Alfresco Archive

Performance problem with the search

djimenez — Tue, 21 Aug 2007 09:09:04 GMT

Hello,I have a problem when I list contents using webscripts and the freemarker APi. I realize a probe with 700 contents in the repository, In listing 700 contents the doclist webscript spent 53 seconds. I modify this webscript and I limit to 20 the number of contents that the webscript show, it wa

Re: Performance problem with the search

witho — Fri, 11 Sep 2009 07:59:43 GMT

I have the same problem with a doclit portlet I have created to find the last 20 documents uploaded to the repoitory. The thing is that I have to search every docs in the repository order by date and only print the first 20 documents and the others just set at display:none.

Is there any better solution in order to minimize the waiting time?

Thanks a lot!!!

Re: Performance problem with the search

rogier_oudshoor — Fri, 11 Sep 2009 12:32:57 GMT

That's because you're using Freemarker to do the sorting. That means that all 700 nodes are loaded into the template to be sorted.

What you SHOULD do, is code a Java-backed webscript in which Lucene does the sorting (!) and you have full control over which attributes from which nodes you access. I'm quite sure from experience you'll be able to drop it to a second when only touching 20 nodes. (Lucene doesn't have to load a node to sort)

Re: Performance problem with the search

witho — Wed, 16 Sep 2009 08:04:51 GMT

The thing is that I want to have the same functionality that in teh doclist portlet, so I would have to edit or replicate the WebScriportlet. The problem is that I don't know where in the WebScriptPortlet can I put my code to only get the 20 last documents added to the repository.

Re: Performance problem with the search

witho — Thu, 17 Sep 2009 09:23:47 GMT

Hi,

Finally I got a valid solution for me. I replicate the method getChildrenByLuceneSearch the class LuceneSearchResultsMap and also the method get inside this class. I make a new lucene search that sort the documents by created date and get only the 20 first. This is my method get:


    public Object get(Object key)
    {
        // execute the search
        String search = key.toString();
       List<TemplateNode> nodes = null;
        HashSet<NodeRef> nodeRefs = new HashSet<NodeRef>();

        // check if a full Lucene search string has been supplied or extracted from XML
        if (search != null && search.length() != 0)
        {
            // perform the search against the repo
            ResultSet results = null;
            try
            {
                SearchParameters sp = new SearchParameters();
                sp.addStore(this.parent.getNodeRef().getStoreRef());
                sp.setLanguage(SearchService.LANGUAGE_LUCENE);
                sp.setQuery(search);
                sp.addSort("CREATED", false);
                sp.setLimitBy(LimitBy.FINAL_SIZE);
                int limit = 20;
                sp.setLimit(limit);

                results = this.services.getSearchService().query(sp);

                if (results.length() != 0)
                {
                    nodes = new ArrayList<TemplateNode>(results.length());
                    for (ResultSetRow row : results)
                    {
                        NodeRef nodeRef = row.getNodeRef();
                        if (!nodeRefs.contains(nodeRef))
                        {
                            nodes.add(new TemplateNode(nodeRef, services, this.parent.getImageResolver()));
                            nodeRefs.add(nodeRef);
                        }
                    }
                }
            }
            catch (Throwable err)
            {
                throw new AlfrescoRuntimeException("Failed to execute search: " + search, err);
            }
            finally
            {
                if (results != null)
                {
                    results.close();
                }
            }
        }

        return nodes != null ? nodes : (List) Collections.emptyList();
    }
‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

I hope this may help somebody too.