Query Builder API

The Query Builder offers an easy way of querying the content repository of AEM. The functionality is exposed through a Java API and a REST API. This document describes these APIs.

The server-side query builder (QueryBuilder) will accept a query description, create and run an XPath query, optionally filter the result set, and also extract facets, if desired.

The query description is simply a set of predicates (Predicate). Examples include a full-text predicate, which corresponds to the jcr:contains() function in XPath.

For each predicate type, there is an evaluator component (PredicateEvaluator) that knows how to handle that specific predicate for XPath, filtering, and facet extraction. It is very easy to create custom evaluators, which are plugged-in through the OSGi component runtime.

The REST API provides access to exactly the same features through HTTP with responses being sent in JSON.

Gem Session

AEM Gems is a series of technical deep dives into Adobe Experience Manager delivered by Adobe experts.

You can review the session dedicated to the query builder for an overview and use of the tool.

Sample Queries

These samples are given in Java properties style notation. To use them with the Java API, use a Java HashMap as in the API sample that follows.

For the QueryBuilder JSON Servlet, each example includes a sample link to an AEM installation (at the default location, http://<host>:<port>). Note that you have to log in to your AEM instance before using these links.

Returning All Results

The following query will return ten results (or to be precise a maximum of ten), but inform you of the Number of hits: that are actually available:

http://<host>:<port>/bin/querybuilder.json?path=/content&1_property=sling:resourceType&1_property.value=wknd/components/structure/page&1_property.operation=like&orderby=path

path=/content
1_property=sling:resourceType
1_property.value=wknd/components/structure/page
1_property.operation=like
orderby=path

The same query (with the parameter p.limit=-1) will return all results (this might be a high number depending on your instance):

http://<host>:<port>/bin/querybuilder.json?path=/content&1_property=sling:resourceType&1_property.value=wknd/components/structure/page&1_property.operation=like&orderby=path&p.limit=-1

path=/content
1_property=sling:resourceType
1_property.value=wknd/components/structure/page
1_property.operation=like
p.limit=-1
orderby=path

Using p.guessTotal to Return the Results

The purpose of the p.guessTotal parameter is to return the appropriate number of results that can be shown by combining the minimum viable p.offset and p.limit values. The advantage of using this parameter is improved performance with large result sets. This avoids calculating the full total (e.g calling result.getSize()) and reading the entire result set, optimized all the way down to the OAK engine and index. This can be a significant difference when there are hundreds of thousands of results, both in execution time and memory usage.

The disadvantage to the parameter is users do not see the exact total. But you can set a minimum number like p.guessTotal=1000 so it will always read up to 1000, so you get exact totals for smaller result sets, but if it’s more than that, you can only show “and more”.

Add p.guessTotal=true to the query below to see how it works:

http://<host>:<port>/bin/querybuilder.json?path=/content&1_property=sling:resourceType&1_property.value=wknd/components/structure/page&1_property.operation=like&p.guessTotal=true&orderby=path

path=/content
1_property=sling:resourceType
1_property.value=wknd/components/structure/page
1_property.operation=like
p.guessTotal=true
orderby=path

The query will return the p.limit default of 10 results with a 0 offset:

"success": true,
"results": 10,
"total": 10,
"more": true,
"offset": 0,

You can also use a numeric value to count up to a custom number of maximum results. Use the same query as above, but change the value of p.guessTotal to 50:

http://<host>:<port>/bin/querybuilder.json?path=/content&1_property=sling:resourceType&1_property.value=wknd/components/structure/page&1_property.operation=like&p.guessTotal=50&orderby=path

It will return a number the same default limit of 10 results with a 0 offset, but will only display a maximum of 50 results:

"success": true,
"results": 10,
"total": 50,
"more": true,
"offset": 0,

Implementing Pagination

By default the Query Builder would also provide the number of hits. Depending on the result size this might take long time as determining the accurate count involves checking every result for access control. Mostly the total is used to implement pagination for the end user UI. As determining the exact count can be slow it is recommended to make use of the guessTotal feature to implement the pagination.

For example, the UI can adapt following approach:

• Get and display the accurate count of the number of total hits (SearchResult.getTotalMatches() or total in the querybuilder.json response) are less than or equal to 100;

• Set guessTotal to 100 while making the call to the Query Builder.

• The response can have the following outcome:

  • total=43, more=false - Indicates that total number of hits is 43. The UI can show up to ten results as part of the first page and provide pagination for the next three pages. You can also use this implementation to display a descriptive text like “43 results found”.
  • total=100, more=true - Indicates that the total number of hits is greater than 100 and the exact count is not known. The UI can show up to ten as part of the first page and provide pagination for the next ten pages. You can also use this to display a text like “more than 100 results found”. As the user goes to the next pages calls made to the Query Builder would increase the limit of guessTotal and also of the offset and limit parameters.

guessTotal should also be used in cases where the UI needs to make use of infinite scrolling, in order to avoid the Query Builder from determining the exact hit count.

Find jar Files and Order Them, Newest First

http://<host>:<port>/bin/querybuilder.json?type=nt:file&nodename=*.jar&orderby=@jcr:content/jcr:lastModified&orderby.sort=desc

type=nt:file
nodename=*.jar
orderby=@jcr:content/jcr:lastModified
orderby.sort=desc

Find All Pages and Order Them by Last Modified

http://<host>:<port>/bin/querybuilder.json?type=cq:Page&orderby=@jcr:content/cq:lastModified

type=cq:Page
orderby=@jcr:content/cq:lastModified

Find All Pages and Order Them by Last Modified, Descending

http://<host>:<port>/bin/querybuilder.json?type=cq:Page&orderby=@jcr:content/cq:lastModified&orderby.sort=desc

type=cq:Page
orderby=@jcr:content/cq:lastModified
orderby.sort=desc

Fulltext Search, Ordered by Score

http://<host>:<port>/bin/querybuilder.json?fulltext=Management&orderby=@jcr:score&orderby.sort=desc

fulltext=Management
orderby=@jcr:score
orderby.sort=desc

Search for Pages with a Certain Tag

http://<host>:<port>/bin/querybuilder.json?type=cq:Page&tagid=wknd:activity/cycling&tagid.property=jcr:content/cq:tags

type=cq:Page
tagid=wknd:activity/cycling
tagid.property=jcr:content/cq:tags

Use the tagid predicate as in the example if you know the explicit tag ID.

Use the tag predicate for the tag title path (without spaces).

Because, in the previous example, you are searching for pages (cq:Page nodes), you need to use the relative path from that node for the tagid.property predicate, which is jcr:content/cq:tags. By default, the tagid.property would simply be cq:tags.

Search Multiple Paths (using groups)

http://<host>:<port>/bin/querybuilder.json?fulltext=Experience&group.1_path=/content/wknd/us/en/magazine&group.2_path=/content/wknd/us/en/adventures&group.p.or=true

fulltext=Experience
group.p.or=true
group.1_path=/content/wknd/us/en/magazine
group.2_path=/content/wknd/us/en/adventures

This query uses a group (named group), which acts to delimit subexpressions within a query, much as parentheses do in more standard notations. For example, the previous query might be expressed in a more familiar style as:

"Experience" and ("/content/wknd/us/en/magazine?lang=en" or "/content/wknd/us/en/adventures?lang=en")

Inside the group in the example, the path predicate is used multiple times. To differentiate and order the two instances of the predicate (ordering is required for some predicates), you must prefix the predicates with N_ where N is the ordering index. In the previous example, the resulting predicates are 1_path and 2_path.

The p in p.or is a special delimiter indicating that what follows (in this case an or) is a parameter of the group, as opposed to a subpredicate of the group, such as 1_path.

If no p.or is given then all predicates are ANDed together, that is, each result must satisfy all predicates.

Search for Properties

Here you are searching for all pages of a given template, using the cq:template property:

http://<host>:<port>/bin/querybuilder.json?property=cq%3atemplate&property.value=%2fconf%2fwknd%2fsettings%2fwcm%2ftemplates%2fadventure-page-template&type=cq%3aPageContent

type=cq:PageContent
property=cq:template
property.value=/conf/wknd/settings/wcm/templates/adventure-page-template

This has the drawback that the jcr:content nodes of the pages, not the pages themselves, are returned. To solve this, you can search by relative path:

http://<host>:<port>/bin/querybuilder.json?property=jcr%3acontent%2fcq%3atemplate&property.value=%2fconf%2fwknd%2fsettings%2fwcm%2ftemplates%2fadventure-page-template&type=cq%3aPage

type=cq:Page
property=jcr:content/cq:template
property.value=/conf/wknd/settings/wcm/templates/adventure-page-template

Search for Multiple Properties

When using the property predicate multiple times, you have to add the number prefixes again:

http://<host>:<port>/bin/querybuilder.json?1_property=jcr%3acontent%2fcq%3atemplate&1_property.value=%2fconf%2fwknd%2fsettings%2fwcm%2ftemplates%2fadventure-page-template&2_property=jcr%3acontent%2fjcr%3atitle&2_property.value=Cycling%20Tuscany&type=cq%3aPage

type=cq:Page
1_property=jcr:content/cq:template
1_property.value=/conf/wknd/settings/wcm/templates/adventure-page-template
2_property=jcr:content/jcr:title
2_property.value=Cycling Tuscany

Search for Multiple Property Values

To avoid big groups when you want to search for multiple values of a property ("A" or "B" or "C"), you can provide multiple values to the property predicate:

http://<host>:<port>/bin/querybuilder.json?property=jcr%3atitle&property.1_value=Cycling%20Tuscany&property.2_value=Ski%20Touring&property.3_value=Whistler%20Mountain%20Biking

property=jcr:title
property.1_value=Cycling Tuscany
property.2_value=Ski Touring
property.3_value=Whistler Mountain Biking

For multi-value properties, you can also require that multiple values match ("A" and "B" and "C"):

http://<host>:<port>/bin/querybuilder.json?property=jcr%3atitle&property.and=true&property.1_value=Cycling%20Tuscany&property.2_value=Ski%20Touring&property.3_value=Whistler%20Mountain%20Biking

property=jcr:title
property.and=true
property.1_value=Cycling Tuscany
property.2_value=Ski Touring
property.3_value=Whistler Mountain Biking

Refining What is Returned

By default, the QueryBuilder JSON Servlet will return a default set of properties for each node in the search result (e.g. path, name, title, etc.). In order to gain control over which properties are returned, you can do one of the following:

Specify

p.hits=full

in which case all properties will be included for each node:

http://<host>:<port>/bin/querybuilder.json?p.hits=full&property=jcr%3atitle&property.value=Cycling%20Tuscany

property=jcr:title
property.value=Cycling Tuscany
p.hits=full

Use

p.hits=selective

and specify the properties you want to get in

p.properties

separated by a space:

http://<host>:<port>/bin/querybuilder.json?p.hits=selective&p.properties=sling%3aresourceType%20jcr%3aprimaryType&property=jcr%3atitle&property.value=Cycling%20Tuscany

property=jcr:title
property.value=Cycling Tuscany
p.hits=selective
p.properties=sling:resourceType jcr:primaryType

Another thing you can do is include child nodes in the Query Builder response. In order to do this you need to specify

p.nodedepth=n

where n is the number of levels you want the query to return. Note that in order for a child node to be returned, it must be specified by the properties selector

p.hits=full

Example:

http://<host>:<port>/bin/querybuilder.json?p.hits=full&p.nodedepth=5&property=jcr%3atitle&property.value=Cycling%20Tuscany

property=jcr:title
property.value=Cycling Tuscany
p.hits=full
p.nodedepth=5

More Predicates

For more predicates, see the Query Builder Predicate Reference page.

You can also check the Javadoc for the PredicateEvaluator classes. The Javadoc for these classes contains the list of properties that you can use.

The prefix of the class name (for example, similar in SimilarityPredicateEvaluator) is the principal property of the class. This property is also the name of the predicate to use in the query (in lower case).

For such principal properties, you can shorten the query and use similar=/content/en instead of the fully qualified variant similar.similar=/content/en. The fully qualified form must be used for all non-principal properties of a class.

Example Query Builder API Usage

   String fulltextSearchTerm = "WKND";

    // create query description as hash map (simplest way, same as form post)
    Map<String, String> map = new HashMap<String, String>();

// create query description as hash map (simplest way, same as form post)
    map.put("path", "/content");
    map.put("type", "cq:Page");
    map.put("group.p.or", "true"); // combine this group with OR
    map.put("group.1_fulltext", fulltextSearchTerm);
    map.put("group.1_fulltext.relPath", "jcr:content");
    map.put("group.2_fulltext", fulltextSearchTerm);
    map.put("group.2_fulltext.relPath", "jcr:content/@cq:tags");

    // can be done in map or with Query methods
    map.put("p.offset", "0"); // same as query.setStart(0) below
    map.put("p.limit", "20"); // same as query.setHitsPerPage(20) below

    Query query = builder.createQuery(PredicateGroup.create(map), session);
    query.setStart(0);
    query.setHitsPerPage(20);

    SearchResult result = query.getResult();

    // paging metadata
    int hitsPerPage = result.getHits().size(); // 20 (set above) or lower
    long totalMatches = result.getTotalMatches();
    long offset = result.getStartIndex();
    long numberOfPages = totalMatches / 20;

    //Place the results in XML to return to client
    DocumentBuilderFactory factory = DocumentBuilderFactory.newInstance();
    DocumentBuilder builder = factory.newDocumentBuilder();
    Document doc = builder.newDocument();

    //Start building the XML to pass back to the AEM client
    Element root = doc.createElement( "results" );
    doc.appendChild( root );

    // iterating over the results
    for (Hit hit : result.getHits()) {
       String path = hit.getPath();

      //Create a result element
      Element resultel = doc.createElement( "result" );
      root.appendChild( resultel );

      Element pathel = doc.createElement( "path" );
      pathel.appendChild( doc.createTextNode(path ) );
      resultel.appendChild( pathel );
    }

The same query executed over HTTP using the Query Builder (JSON) Servlet:

http://<host>:<port>/bin/querybuilder.json?path=/content&type=cq:Page&group.p.or=true&group.1_fulltext=WKND&group.1_fulltext.relPath=jcr:content&group.2_fulltext=WKND&group.2_fulltext.relPath=jcr:content/@cq:tags&p.offset=0&p.limit=20

Storing and Loading Queries

Queries can be stored to the repository so that you can use them later. The QueryBuilder provides the storeQuery method with the following signature:

void storeQuery(Query query, String path, boolean createFile, Session session) throws RepositoryException, IOException;

When using the QueryBuilder#storeQuery method, the given Query is stored into the repository as a file or as a property according to the createFile argument value. The following example shows how to save a Query to the path /mypath/getfiles as a file:

builder.storeQuery(query, "/mypath/getfiles", true, session);

Any previously stored queries can be loaded from the repository by using the QueryBuilder#loadQuery method:

Query loadQuery(String path, Session session) throws RepositoryException, IOException

For example, a Query stored to the path /mypath/getfiles can be loaded by the following snippet:

Query loadedQuery = builder.loadQuery("/mypath/getfiles", session);

Testing and Debugging

For playing around and debugging Query Builder queries, you can use the Query Builder debugger console at

http://<host>:<port>/libs/cq/search/content/querydebug.html

or alternatively the Query Builder JSON servlet at

http://<host>:<port>/bin/querybuilder.json?path=/tmp

path=/tmp is only an example.

General Debugging Recommendations

Obtain Explainable XPath via Logging

Explain all queries during the development cycle against the target index set.

1. Enable DEBUG logs for QueryBuilder to obtain underlying, explainable XPath query
   • Navigate to https://<host>:<port>/system/console/slinglog. Create a new logger for com.day.cq.search.impl.builder.QueryImpl at DEBUG.
2. Once DEBUG has been enabled for the above class, the logs will display the XPath generated by Query Builder.
3. Copy the XPath query from the log entry for the associated Query Builder query, For example:
   • com.day.cq.search.impl.builder.QueryImpl XPath query: /jcr:root/content//element(*, cq:Page)[(jcr:contains(jcr:content, "WKND") or jcr:contains(jcr:content/@cq:tags, "WKND"))]
4. Paste the XPath query into Explain Query as XPath to obtain the query plan.

Obtain Explainable XPath via the Query Builder Debugger

Use the AEM Query Builder debugger to generate an explainable XPath query.

1. Provide the Query Builder query in the Query Builder debugger
2. Execute the Search
3. Obtain the generated XPath
4. Paste the XPath query into Explain Query as XPath to obtain the query plan

Debugging Queries with Logging

The log output (INFO level) of the query builder implementation when executing the query described in the previous section Testing and Debugging:

com.day.cq.search.impl.builder.QueryImpl executing query (predicate tree):
null=group: limit=20, offset=0[
    {group=group: or=true[
        {1_fulltext=fulltext: fulltext=WKND, relPath=jcr:content}
        {2_fulltext=fulltext: fulltext=WKND, relPath=jcr:content/@cq:tags}
    ]}
    {path=path: path=/content}
    {type=type: type=cq:Page}
]
com.day.cq.search.impl.builder.QueryImpl XPath query: /jcr:root/content//element(*, cq:Page)[(jcr:contains(jcr:content, "WKND") or jcr:contains(jcr:content/@cq:tags, "WKND"))]
com.day.cq.search.impl.builder.QueryImpl no filtering predicates
com.day.cq.search.impl.builder.QueryImpl query execution took 69 ms

If you have a query using predicate evaluators that filter or that use a custom order by comparator, this will also be noted in the query:

com.day.cq.search.impl.builder.QueryImpl executing query (predicate tree):
null=group: [
    {nodename=nodename: nodename=*.jar}
    {orderby=orderby: orderby=@jcr:content/jcr:lastModified}
    {type=type: type=nt:file}
]
com.day.cq.search.impl.builder.QueryImpl custom order by comparator: jcr:content/jcr:lastModified
com.day.cq.search.impl.builder.QueryImpl XPath query: //element(*, nt:file)
com.day.cq.search.impl.builder.QueryImpl filtering predicates: {nodename=nodename: nodename=*.jar}
com.day.cq.search.impl.builder.QueryImpl query execution took 272 ms

Javadoc Links

Javadoc	Description
com.day.cq.search	Basic Query Builder and Query API
com.day.cq.search.result	Result API
com.day.cq.search.facets	Facets
com.day.cq.search.facets.buckets	Buckets (contained within facets)
com.day.cq.search.eval	Predicate Evaluators
com.day.cq.search.facets.extractors	Facet Extractors (for evaluators)
com.day.cq.search.writer	JSON Result Hit Writer for Query Builder servlet (/bin/querybuilder.json)