Changelog
A record of what changed between successive versions
- 1.Post-filter in read endpoints -
log
,diff
,show
. - 2.k Shortest Paths - Custom-weighted, point-in-time, shortest paths between endpoints.
- 3.Synchronous skeleton graph updates within commit transactions.
- 4.OpenTracing instrumentation, using ancillary collector foxx service. Non-intrusive and strictly optional.
- 5.Logs can be grouped by type (vertex/edge).
- 6.Explicit Commits to sync event log with writes that occurred outside of RecallGraph's API methods.
- 7.Providers for all API endpoints to let other dependent Foxx services invoke RecallGraph's service methods directly, using ArangoDB's service linking mechanism.
- 8.Purge endpoint to remove all history for nodes at a specified path.
- 9.Restore endpoint to undelete nodes that were deleted through RecallGraph's API.
- 10.Paths are returned in traverse calls, with support for path filters.
- 11.Support for slicing in show endpoint when aggregation is enabled.
- 12.Data:
- 1.Event objects have a
collection
field to identify the collection to which their corresponding document belongs. - 2.Skeleton graph objects have a
collection
field to identify the collection to which their corresponding document belongs. - 3.Origin events have
meta.id
,hops-from-origin
,collection
,ctime
andevent
fields to allow building dense indexes on them. - 4.Migration script to automatically upgrade event and skeleton trees created using older versions.
- 5.Migration script to remove stale indices.
- 1.Precision limiters for timestamp parameters.
- 2.Dedicated
filter
endpoint. All read endpoints now have their own filtering capability. - 3.Asynchronous skeleton graph updates in cron jobs.
- 4.Breaking:
- 1.
returnCommands
parameter removed fromlog
. Anything that depended on this should now use the enhanceddiff
endpoint instead. - 2.
groupSkip
andgroupLimit
removed fromdiff
endpoint, as they were deemed not useful. The newly available post-filter param can indirectly help make up for the deficit in most cases. - 3.Service collection suffixes are no longer configurable (extraneous config).
- 5.Bug:
- 1.Sort direction was not being honored in
log
whengroupBy
was specified. This has been fixed. - 2.Hard-coded service collection name in a query for
show
was removed. - 3.
traverse
no longer breaks if starting vertex is not found. - 4.Fixed operation id for the
UPDATE
route, which was incorrectly marked asreplace
.
- 1.Document operations are checked first for compatible last events before executing.
- 2.Data:
- 1.Collection origin events have their
origin-for
field renamed tocollection
to remain consistent with field naming convention followed for regular events. - 2.Certain indices have changed - primarily, sparse indexes have been replaced by their dense counterparts.
- 3.When replacing/updating an existing document with no event log record, a
created
event is recorded instead of anupdated
event.
- 3.Breaking:
- 1.Diffs are enhanced to return some event metadata along with command lists. This is to reduce overloading the
log
function with event+command-related queries. - 2.The
countsOnly
parameter now has an effect even whengroupBy
isnull
forlog
endpoints. It returns the overall total number of events based on the filters provided. - 3.
show
returns an array with total count whengroupBy
isnull
, instead of a bare object. This is for consistent post-filter application semantics. - 4.
traverse
minDepth
has been changed from0
to1
.
Last modified 2yr ago