Loading…
Software Development
We designed a Laravel domain for multi-generation lineage data that combines canonical identities with tenant-scoped records, controlled propagation, local overrides, drift detection, auditable synchronization, conflict-aware imports, and materialized snapshots for fast reads.
Client: Confidential Digital Marketplace
Secteur: Marketplace Technology
The Client
A confidential digital marketplace manages products whose records include complex, multi-generation lineage data. The same ancestor can appear across many products and tenant catalogs, while each organization must retain control over its own data and presentation.
The Challenge
What looks like a tree on screen behaves like a shared graph in the data model. A parentage correction can affect many descendant trees, imported records can conflict with existing identities, and tenant-specific overrides must survive synchronization. The system needed to propagate trusted changes without overwriting deliberate local edits, detect drift, prevent circular ancestry, support bulk imports, and return complete trees quickly.
A lineage diagram appears simple: one subject, two parents, four grandparents, and progressively more ancestors at each generation.
The complexity begins when the same ancestor appears in hundreds of trees. At that point, the system is no longer storing independent diagrams. It is managing a shared network of identities, relationships, historical values, tenant-specific decisions, and derived representations.
We designed the feature as a dedicated Laravel domain rather than a collection of form fields attached to a product.
A naive implementation stores an ancestor's name and attributes directly inside every product tree. That works until a correction must be made.
Updating every copied occurrence is expensive and unreliable. Updating only one occurrence creates contradictory records. Automatically updating everything can be equally dangerous when an organization has intentionally overridden a value.
The solution separates two concepts:
Nodes can link to a shared subject while retaining the contextual information required by their own tree. This makes reuse possible without pretending that every appearance must always be identical.
The marketplace serves multiple organizations, so shared data cannot ignore tenant boundaries.
The model supports both canonical subjects and tenant-scoped subjects. Canonical subjects provide reusable identities that may be referenced across permitted contexts. Tenant-scoped subjects remain owned and visible within the appropriate organization.
This distinction allows the platform to reuse verified lineage information while preserving data isolation. A tenant can work with shared identities without gaining unrestricted access to another tenant's private records.
Scope is enforced by the server during search, linking, editing, traversal, and rendering. It is not accepted as a value that a client application can choose for itself.
When the parentage of a shared subject changes, every dependent tree may become stale.
A propagation engine identifies affected roots, traverses the updated ancestry, and refreshes derived nodes. However, not every field should be overwritten.
Each node records its provenance and whether its identity or attributes have been deliberately overridden. Propagated values can be refreshed automatically, while user-authored or overridden values remain protected.
Different contexts can choose different propagation policies:
The system therefore treats synchronization as policy-driven reconciliation, not blind copying.
Automatic propagation is not appropriate for every change.
A drift detector compares a stored tree with the currently resolved ancestry and reports:
Administrators can preview the differences before applying a manual synchronization. The operation produces a structured report explaining which positions changed, which were skipped, and why.
This turns synchronization from an opaque background effect into an observable and reviewable workflow.
Bulk catalog imports introduce a second source of complexity. Imported spreadsheets may contain partial trees, repeated ancestors, new identities, inconsistent identifiers, or values that disagree with existing shared records.
The import pipeline normalizes incoming data, searches for matching subjects, detects shared-ancestor conflicts, and decides whether each position should be created, linked, merged, overridden, or rejected.
It supports partial and update-oriented operations without requiring every import to replace an entire tree. Identity-correction workflows can repair an incorrect link while preserving the surrounding lineage.
Import review reports make conflicts visible before they become silent data corruption.
A lineage engine must protect more than field formats.
The validation layer prevents a subject from becoming its own ancestor, even indirectly through multiple generations. It validates position keys, parent roles, identity compatibility, configurable attributes, and cross-tenant ownership.
Write operations run within transactions so a failure cannot leave half of a tree updated. Deletion guards prevent subjects or nodes from being removed while protected relationships still depend on them.
Audit records preserve the context of identity, parentage, and synchronization changes.
Normalized graphs are effective for editing and propagation, but recursively resolving a deep tree on every request is unnecessarily expensive.
The platform builds materialized snapshots for read-heavy surfaces. A snapshot contains the resolved tree and the references needed to determine which cached results depend on a changed subject.
When parentage or node data changes, background jobs rebuild affected snapshots and invalidate dependent caches. Multiple changes within the same operation are coalesced so the system does not repeatedly rebuild the same tree.
This separates the write model, optimized for correctness and reuse, from the read model, optimized for predictable delivery.
Large propagation operations and maintenance tasks run outside the request lifecycle.
Background jobs handle:
The jobs are designed to be repeatable so interrupted work can be retried without producing duplicate relationships or corrupting existing overrides.
The result is a lineage domain that treats shared ancestry as a data-integrity problem rather than a visual tree editor.
Canonical identities reduce duplication. Tenant-scoped records preserve ownership. Provenance and override rules protect intentional local changes. Drift detection makes synchronization reviewable. Conflict-aware imports prevent ambiguous data from silently entering the system, and materialized snapshots keep complex trees efficient to read.
The feature demonstrates how careful domain modeling can turn a fragile hierarchy of copied values into an auditable, maintainable, and scalable synchronization system.
