Search bridges¶
Warning
All search bridge classes have been deprecated in 5.3.0. This extension has never been widely used and is now out of date, causing a bottleneck in performance and preventing the upgrade of Hibernate. A new modern search functionality is under development.
A “bridge” is the mapping between your metadata and how it is stored in
the Lucene index.
OMERO search uses one internal bridge to parse
all of your metadata for later searching. If, however, there is more
metadata that you would like to add to the index, you can implement the
org.hibernate.search.bridge.FieldBridge
interface yourself, or
subclass the helper class
components/server/src/ome/services/fulltext/BridgeHelper.java
Example¶
Assume you wanted to be able to search for a project based on the name of all images contained in that project. In the set method,
public void set(final String name, final Object value,
final Document document, final Field.Store store,
final Field.Index index, final Float boost) {
you would need to add a field to the Document
for each Image
.
Project p = (Project) value;
List<Image> images = getImages(p);
for (Image image : images) {
add(document, "image_name", image.getName(), store, index, boost);
}
Configuration¶
Custom bridges are configured in Search but can be
overridden via the standard configuration mechanisms.
The omero.search.bridges
property defines a comma-separated list of
bridge classes which will be passed to
components/server/src/ome/services/fulltext/FullTextBridge.java.
See Java deployment for how to have your bridge classes included on the server’s classpath if it doesn’t get built by the Build System.
Available bridges¶
See components/server/src/ome/services/fulltext/bridges for a list of provided (example) bridges.
Re-indexing¶
BridgeHelper
provides two methods – reindex(IObject)
and
reindexAll(List<IObject>)
– for keeping the indexes for objects in
sync.
For example, if the image.name
above were to be changed, the index
for the Project
would be stale until the Project
itself were
re-indexed. Custom bridges can call reindex(Project)
while indexing
the image to have the Project re-indexed from the backlog. Before
any new changes are processed, the backlog is always first cleared. One
caveat: while processing the backlog, no new objects can be added to it.
For bridge writers, this means concretely that implementations should check for all related types and index them in groups, rather than relying on transitivity. For example,
if (value instanceof Project) {
final Project p = (Project) value;
handleProject(p, document, store, index, boost);
for (final ProjectDatasetLink pdl : p.unmodifiableDatasetLinks()) {
final Dataset d = pdl.child();
reindex.add(d);
handleDataset(d, document, store, index, boost);
for (final DatasetImageLink dil : d.unmodifiableImageLinks()) {
final Image i = dil.child();
reindex.add(i);
handleImage(document, store, index, two_step_boost, i);
}
}
} else if (value instanceof Dataset) {
final Dataset d = (Dataset) value;
handleDataset(d, document, store, index, boost);
for (final ProjectDatasetLink pdl : d.unmodifiableProjectLinks()) {
final Project p = pdl.parent();
reindex.add(p);
handleProject(p, document, store, index, two_step_boost);
}
for (final DatasetImageLink dil : d.unmodifiableImageLinks()) {
final Image i = dil.child();
reindex.add(i);
handleImage(document, store, index, two_step_boost, i);
}
} else if (value instanceof Image) {
final Image i = (Image) value;
handleImage(document, store, index, two_step_boost, i);
for (final DatasetImageLink dil : i.unmodifiableDatasetLinks()) {
final Dataset d = dil.parent();
reindex.add(d);
handleDataset(d, document, store, index, boost);
for (final ProjectDatasetLink pdl : d
.unmodifiableProjectLinks()) {
final Project p = pdl.parent();
reindex.add(p);
handleProject(p, document, store, index, boost);
}
}
}
//
// Handle re-indexing
//
if (reindex.size() > 0) {
reindexAll(reindex);
}
}
In which case, regardless of whether an Image, Dataset, or Project is indexed, all related objects are simultaneously added to the backlog, which will be processed in the next cycle, but their indexing will not add any new values to the backlog.
See also