The generic search worker now does indexing work during the index operation,
ensuring that queries do not have to do extraneous or repeat calculations.
Change the return format slightly and fixed a bug in the GenericSearchProvider
which caused more objects than intended to be returned from the provider.
Per discussion from nasa/openmctweb#141, minimize the
interval at which new objects get indexed, instead of
presuming requirements for CPU utilization.
When indexing for generic search, issue new batches of
requests as individual requests finish (instead of waiting
for whole batches to finish) varying size to keep the
number of outstanding requests below some maximum.
nasa/openmctweb#141
This reverts commit 4b8a5ac0b257737ee33effc966816afca6c11b94.
Performance measurements indicates this is harmful for performance,
although this is not well-explained.
Move function used to listen for mutation changes (to trigger
reindexing) up in scope, to avoid retaining references to
domain objects via closure. nasa/openmctweb#141
Also, includes misc. whitespace normalization (provided by
code editor.)
Updated the generic provider test. Removed index
checking from the generic search worker, because
that checking is already done (more efficiently)
in the generic search provider.
When indexing items initially, the generic
provider listens for mutations in case an
item's composition changes, so it can then
index the new children.
Moved the elasticsearch provider to
platform/persistence/elastic. Then moved
the search aggregator and the generic
provider into a services folder within
the search folder.
