#006: Dataset history via named graphs (in-place) #6

New issue

Open

opened 2026-04-05 12:58:41 +00:00 by daniel · 0 comments

daniel commented

2026-04-05 12:58:41 +00:00

Owner

Blocked by

#005 SHACL migration detection (history is triggered by migrations)

Summary

Save dataset history in-place by creating named graphs within the same dataset each time a migration or significant change is performed. Metadata in the meta-dataset indicates which named graph is the current version and which are historical snapshots.

Design sketch

When a migration is performed:

Copy the current default graph (or affected named graph) into a timestamped named graph: <urn:history:2026-04-05T10:30:00Z>

Record in the meta-dataset:

<urn:history:2026-04-05T10:30:00Z> a concon:HistoricalSnapshot ;
    concon:snapshotOf <urn:graph:default> ;
    concon:createdAt "2026-04-05T10:30:00Z"^^xsd:dateTime ;
    concon:shaclVersion <urn:shacl:v1> ;
    concon:reason "Migration: added ex:email property" .

Apply the migration to the current graph.

Pros

Self-contained: history lives in the same dataset
Queryable: SPARQL can access historical snapshots via named graphs
No additional storage infrastructure

Cons

Dataset size grows with each snapshot
Large datasets with frequent changes may become unwieldy
Cannot easily share or transfer history independently

Configuration

Enabled per-dataset via meta-dataset:

<urn:config:history> concon:historyStrategy concon:InPlaceNamedGraphs .

Tests

Unit tests

#[test]
fn test_snapshot_created_on_migration() {
    // Configure dataset with InPlaceNamedGraphs strategy
    // Insert data, run a migration
    // Verify: a timestamped named graph exists with pre-migration data
}

#[test]
fn test_snapshot_metadata_recorded() {
    // Run migration
    // Query meta-dataset for snapshot metadata
    // Verify: snapshotOf, createdAt, reason present
}

#[test]
fn test_current_graph_updated_after_migration() {
    // Run migration that adds a mandatory property
    // Verify: current graph has the new data
    // Verify: snapshot graph has the old data (without new property)
}

#[test]
fn test_query_historical_snapshot() {
    // Create multiple snapshots
    // Query a specific historical named graph
    // Verify: returns data from that point in time
}

Manual tests

Enable InPlaceNamedGraphs for a dataset
Edit SHACL, trigger migration
Query the dataset's named graphs — verify snapshot exists
Compare current vs historical data via SPARQL

# Blocked by - [\#005 SHACL migration detection](005-shacl-migration-detection.org) (history is triggered by migrations) # Summary Save dataset history in-place by creating named graphs within the same dataset each time a migration or significant change is performed. Metadata in the meta-dataset indicates which named graph is the current version and which are historical snapshots. # Design sketch When a migration is performed: 1. Copy the current default graph (or affected named graph) into a timestamped named graph: `<urn:history:2026-04-05T10:30:00Z>` 2. Record in the meta-dataset: ``` turtle <urn:history:2026-04-05T10:30:00Z> a concon:HistoricalSnapshot ; concon:snapshotOf <urn:graph:default> ; concon:createdAt "2026-04-05T10:30:00Z"^^xsd:dateTime ; concon:shaclVersion <urn:shacl:v1> ; concon:reason "Migration: added ex:email property" . ``` 3. Apply the migration to the current graph. # Pros - Self-contained: history lives in the same dataset - Queryable: SPARQL can access historical snapshots via named graphs - No additional storage infrastructure # Cons - Dataset size grows with each snapshot - Large datasets with frequent changes may become unwieldy - Cannot easily share or transfer history independently # Configuration Enabled per-dataset via meta-dataset: ``` turtle <urn:config:history> concon:historyStrategy concon:InPlaceNamedGraphs . ``` # Tests ## Unit tests ``` rust #[test] fn test_snapshot_created_on_migration() { // Configure dataset with InPlaceNamedGraphs strategy // Insert data, run a migration // Verify: a timestamped named graph exists with pre-migration data } #[test] fn test_snapshot_metadata_recorded() { // Run migration // Query meta-dataset for snapshot metadata // Verify: snapshotOf, createdAt, reason present } #[test] fn test_current_graph_updated_after_migration() { // Run migration that adds a mandatory property // Verify: current graph has the new data // Verify: snapshot graph has the old data (without new property) } #[test] fn test_query_historical_snapshot() { // Create multiple snapshots // Query a specific historical named graph // Verify: returns data from that point in time } ``` ## Manual tests 1. Enable InPlaceNamedGraphs for a dataset 2. Edit SHACL, trigger migration 3. Query the dataset's named graphs — verify snapshot exists 4. Compare current vs historical data via SPARQL

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

daniel/concon#6

No description provided.

Rows
Columns