Proposal title

Date	2009/04/20
Contact(s)	nicolas, francois
Last edited	Timestamp
Status	being discussed
Assigned to release	2.5.0
Resources	Done in geocat.ch sandbox

Overview

Log all search made on the catalogue. Create an analyzer for the logged information and add an administration view for the analyzer results.

...

Proposal Type

• Type: GUI Change, New module
• App: GeoNetwork
• Module: Lucene searcher

Links

• Documents:
• Email discussions:
  • ​http://n2.nabble.com/Choosing-a-product-to-generate-graphics-td2371468.html
• Other wiki discussions:

Voting History

• Vote proposed by X on Y, result was +/-n (m non-voting members).

---

Motivations

Improve catalogue usage for the administrator.

Proposal

To be logged:

• search criteria
• criteria values
• user IP
• metadata view (done using popularity in GeoNetwork)

Where to log :

• in all searchers : Lucene searcher, Z39.50 searcher and CatalogueSearcher

Indicators:

• more used criteria (top 10)
• more used values (top 10)
• groups selected
• type of metadata
• spatial search or not
• simple search (ie full text) or advanced
• number of results by search (average, 0)
• metadata popularity and popularity by group

Reports:

• HTML format
  • Charts for number of search by days
  • tagclouds view
  • tables
  • text file export

...

Backwards Compatibility Issues

Risks

Participants

• Nicolas (main actor)
• Francois (support)

Specifications (current implementation of statistics)

Storing and displaying information about metadata search involves several sub components or logical elements:

• The Database schema that will store search information
• The Jeeves services that will be impacted by the statistics registration
• The configuration part in which the kind of information to present is configured (to be precised, too vague for me)
• The Graphic rendering to display charts
• The HTML output to present result tables

Database schema

Geonetwork already uses a database to store configuration information. This schema is extended to add tables that will store search criteria. The DB schema is shown on the following picture:

Druid is used to define the tables that will store search criteria.

Database access is done 2 times during the statistics lifecyle:

• When a lucene query is performed, Requests and Params objects are created from the current context and the Lucene search terms (Params). These objects are store into the dabase.
• When the statistics are used in the GN, administration page: each time a statistic is displayed, a query is made in the database to get results.

Configuration/administration

The configuration of statistics can be made at the Jeeves service level: by editing the config_statistics.xml, one can change the queries used to generate stat report (when doing this, also change the XLST sheets corresponding to the service to be sure attribute names are consistent whith those returns by SQL queries.

The Graphics service can also receive HTTP parameters sent when calling the service (image size for instance). One could add an "advanced" tab to the stats page to allow administrator to set these parameters.

Jeeves services

2 kind of services are created to deal with statistics:

• pure XML services, where the SQL query to get stats is sent by a Jeeves service. result is processed by a dedicated XSLT sheet.
• Java service, when a specific processing is to be done before presenting the result:

Such services are defined to set the graphic (date range, graphic properties, graphic file writing), or to deal with SQL results

Graphic Rendering

JFreeChart lib is used to produce graphics. Graphic type (pie chart, temporal serie) is hard-coded in the Java service. An image file is written on the server (web/images/statistics) and its URL is sent to the XSLT sheet.

An optional ImageMap HTML information can be returned by the service, allowing tooltips on the graphic image.

The UI part allow to define the time range and stat type (Daily, monthly or yearly)

A caching mechanism is used to use already-generated images.

HTML output

The statistics page contains links to stat services. Their results are injected into the page (ajax call).

Tagcloud feature:

A JS method is used to generate a lightweight tagcloud based on a SQL query counting search results group by keywords. This tag is currently displayed at the beginning of the page.

Improvements - genericity

Database management

• add a thread to perform database insertion when a search is done, to avoid slowing down the search process