gitlab piveau/metrics/piveau-metrics-annotator 3.0.0

3 hours ago

ChangeLog

Unreleased

3.0.0 (2026-09-23)

Breaking: metrics v2, not compatible with v1 metrics. Evaluation is driven by the shared piveau-utils metric-definition vocabulary and writes boolean dqv:value measurements into a metrics graph marked dct:conformsTo <https://piveau.eu/ns/metrics/2.0.0>; the v1 star scheme and catalogues are removed. Requires score 4.0.0, which must be deployed first.

Updated:

Added:

  • Write dqv:hasQualityMetadata on every evaluated entity, not just the dataset. createMetricsGraph anchors only the dataset, leaving distributions and data services with no direct link to the graph holding their own measurements — a consumer had to reach them by traversing the owning dataset. Against a large triplestore that traversal leaves the entity unbound in SPARQL, which is what pushes scope-limited metric queries past Virtuoso's cost-estimation limit (piveau-metrics-cache refresh). Blank-node entities are skipped, having no stable identity to anchor; the write is guarded so re-annotation does not duplicate it. Applies to newly annotated data — existing metrics graphs need re-annotation. See docs/DATA-CONTRACT.adoc.

Fixed:

  • Skip a dcat:accessService target that the payload does not describe as a data service — either a bare IRI reference carrying no triples at all, or a resource without the rdf:type dcat:DataService that DCAT-AP requires. Such a resource was previously evaluated against all 18 dcat:DataService bindings, failing every one of them; the false negatives added a phantom entity to the catalogue's data-service count and coverage and pulled a 0.0 group into the dataset-final roll-up (observed: a dataset dropping from 6.875 to 4.583). No measurement is now written, so the resource reaches neither the score service nor the cache. The dcat:distribution edge is unchanged — untyped distributions are common in real data and remain evaluated. See docs/DATA-CONTRACT.adoc.

Changed:

  • Use org.apache.jena.vocabulary.DCAT for the core dcat: traversal terms instead of hand-rolled namespace constants; AnnotatorVocab is removed. VERSION_IRI and ADMS_DEPRECATED (an ADMS status value, now built from ADMS.NS_SCHEME_STATUS) move to JobRunner. No behaviour change. See docs/ADR-007-rdf-vocabulary-source.adoc.
  • v2 rewrite: data-driven metric evaluation against a shared piveau-utils vocabulary (MetricDefinitions), boolean-only DQV measurements, a dct:conformsTo version marker on the metrics graph, and deprecated-distribution handling (all-deprecated datasets are skipped, not forwarded). See docs/DATA-CONTRACT.adoc.

2.1.5 (2026-06-08)

Fixed:

  • Pipe schema in API endpoint, disarming sequence number

2.1.4 (2026-06-05)

Changed:

  • Derived measurements from non-existing pre-conditions are set to false
  • General lib updates

2.1.3 (??)

2.1.2 (2025-01-30)

Fixed:

  • Format MediaType alignment calculation fixed

2.1.1 (2025-01-30)

Changed:

  • Revert last change ('phantom' distribution)

2.1.0 (2025-01-29)

Changed:

  • Add 'phantom' distribution metrics to datasets that do not have a distribution

2.0.8 (2024-xx-xx)

Fixed:

  • fixed using the wrong vocabulary for the format, now http://publications.europa.eu/resource/authority/file-type
    should be used by default

2.0.7 (2024-07-01)

Changed:

  • General update

2.0.6 (2024-02-14)

Fixed:

2.0.5 (2024-02-01)

Changed:

  • Lib dependency updates

2.0.4 (2024-01-31)

Changed:

  • just for publish to gitlab.com

2.0.3 (2023-11-12)

Changed:

  • Vert.x update

Added:

  • helm chart

2.0.2 (2023-10-12)

Fixed:

  • Instantiation of annotationContext map only once

Changed:

  • Refactor annotation calculation for better testing

2.0.1 (2023-07-14)

Fixed:

  • JSON serialization issue with limited string size

2.0.0 (2023-05-26)

Changed:

  • Complete refactoring

1.2.7 (2022-11-01)

Changed:

  • Use fixed utilities for IanaTypes (old one caused memory issue)

1.2.5 (2021-11-24)

Added:

  • Date modified annotation for datasets

1.2.4 (2021-10-18)

Changed:

  • Important connector lib update

1.2.3 (2021-06-23)

Changed:

  • Connector pipe handling

1.2.2 (2021-06-07)

Fixed

  • Housekeeping

1.2.1 (2021-03-19)

Fixed:

  • Vert.x issues (updated to 4.0.3)

1.2.0 (2021-01-31)

Added:

  • dqv:hasQualityMetadata property for resource when metrics graph is created

Changed:

  • Switched to Vert.x 4.0.0

1.1.0 (2020-04-23)

Added:

  • Extended information about JSON-LD file-type

1.0.3 (2020-03-11)

Changed:

  • Close dataset explicitly

1.0.2 (2020-03-06)

Fixed:

  • Replace old measurements

1.0.1 (2020-03-04)

Fixed:

  • Finally replace metrics graph to dataset

1.0.0 (2020-02-28)

Initial production release

Added:

  • attached pipe config parameter
  • Configuration PIVEAU_LOAD_VOCABULARIES_FETCH
  • Configuration EXCLUDE_ANNOTATIONS
  • Five Star DQV.QualityAnnotation
  • Metric atLeastFourStars
  • Annotate dcat ap compliance

Changed:

  • Rearrange metadata quality
  • Send rdf dataset with named graph
  • Simplified pipe integration
  • Use vocabularies library
  • Annotations can now be excluded via configuration
  • Measurements optionally replaceable
  • Allow any name for metrics graph

Removed:

  • Service proxy generation

Fixed:

  • dqv:computedOn
  • mediaType and format evaluation
  • Test

Don't miss a new piveau-metrics-annotator release

NewReleases is sending notifications on new releases.