The grpc layer rebuilt the searcher request field by field and dropped
order_by; the response cache also ignored it, so differently sorted
searches collided on the same cache entry.
(cherry picked from commit 97626bb26bd4fe8cdb74c19b519b2230c492681c)
Name is analyzed (lowercaseKeyword) and therefore a text field, which
OpenSearch refuses to sort on without fielddata; enable it in the index
template. The keyword tokenizer emits one term per document, so the
fielddata cache stays small. The schema version has not shipped yet, so
no reindex is needed.
(cherry picked from commit f1d0e7b773b170b119dbd59fa176d656237bed96)
The fan-out passes order_by through to each engine query; the merge
re-sorts the combined matches with CompareMatches and keeps the score
as tiebreaker (and as the sole criterion when no order_by is given).
(cherry picked from commit 66af40fc34fd9daffc028ab06f83f7ade8a360d1)
Sortable is every field that is indexed as a scalar and carried on the
match entity: name, size, lastModifiedDateTime, mimeType and the scalar
facet fields (photo.*, audio.*, image.*, location.*, ...). Both sets
are derived by reflection, so new facet fields become sortable
automatically. Multivalued fields (tags), bare facets and internal
index fields are rejected with invalidRequest at the graph layer.
CompareMatches provides the merge comparator for the service layer.
(cherry picked from commit bd32795194cf3945e8416ede496a147ce8d6cb5c)
Validated against the sortable-field whitelist (name, size,
lastModifiedDateTime, photo.takenDateTime) and forwarded to the search
service as order_by. Also fixes the stub search service's IndexSpace
signature (streaming response) so the suite builds again.
(cherry picked from commit 9aa6a3492de3468da7a8480a3b9ef5748cee1f5d)
(cherry picked from commit 35a6cad8ea6a7890bdc0bc98c7ac97bdcbf477cc)
Migrate the graph search query to metricDefinition/searchMetric and the
@libre.graph prefixes. Add the aggregationFilterToken round-trip: encode
terms/range/or tokens on the response, pass aggregationFilters 1:1 to the
search service, decode them and force exact case-sensitive matches in both
backends. Map the WebDAV-report facets onto search hits: tags, video,
motionPhoto, livePhoto, allowedValues (from the space permission set),
me.following (favorite), webUrl (private link, shared helper) and
thumbnails via opt-in $expand.
Regenerate the vendored libre-graph-api-go from the feat/search spec
(metricDefinition/searchMetric, @libre.graph.subAggregations,
aggregationFilterToken) and add aggregation_filters plus the
permissionsActionsAllowedValues entity field to the search protos.
The legacy OC-Signature path already rejects methods outside
PRE_SIGNED_URL_ALLOWED_HTTP_METHODS, the JWT path did not. A leaked
signed download url could be used for PUT, DELETE, MOVE or PROPFIND
as the signing user for the lifetime of the signature.
$expand=thumbnails was only honored by sharedByMe and sharedWithMe. The
driveItem stat, the children listing and the root children listing now
honor it as well, so a client that lists a folder learns which items have
a preview instead of guessing from the mime type.
The thumbnails are set from the resource info the listing already has, so
a later preview check that needs more than the mime type has a single
place to sit. The two share listings carry driveItems only, they keep
matching on the mime type but share the url building.
The cross-space merge only carried buckets, dropping metric results
(value/metricKind) from the per-space responses. Reduce metrics with
their kind's reducer, keyed by field and kind.
Metric aggregations (sum/min/max/avg) only worked as sub-aggregations
under a terms bucket. Compute top-level metrics by folding the matched
hits through the existing accumulator and allow them through the graph
layer's numeric field validation.
Range aggregations parsed from/to with ParseFloat only, so date bounds on
datetime fields like photo.takenDateTime silently degraded to unbounded
numeric ranges. Detect date-formatted bounds (RFC3339 or YYYY-MM-DD),
switch the facet to bleve date ranges and read DateRanges from the facet
result. Malformed bounds in date mode are rejected.
Implements terms, range, metric and sub-aggregations for the OpenSearch backend
via a dedicated aggs builder, wiring them through the shared search service.
Adds the graph /v1beta1/search/query endpoint, the aggregation proto messages,
the service-layer aggregation forwarding/merging, and the recursive bleve
aggregation implementation (terms, range, metric, sub-aggregations).
Rebased dschmidt/libre-graph-api feat/graph-search-full (PR #34) onto
opencloud-eu main and regenerated via the repo's woodpecker build-go recipe
(openapi-generator v7.23.0, --api-name-suffix Api).
Dominik Schmidt