fix(forest_admin_agent): build PolymorphicManyToOne linkage id from foreign key (#332) by bexchauveto · Pull Request #333 · ForestAdmin/agent-ruby

bexchauveto · 2026-07-21T12:37:23Z

Fixes #332

Problem

Opening the list view of a collection with a polymorphic belongs_to crashes the frontend:

UnknownClientRecordAgentError: Expected id to be a string or number, received null

The backend returns 200 OK but the payload is malformed: the relationship linkage id is an empty string.

Root cause

A PolymorphicManyToOne can't be SQL-joined, so query.rb:208 skips it while building the SELECT. The record therefore comes back with a phantom related object that has no primary key ({ "*" => nil }). ForestSerializer then reads the PK off that phantom object, producing an empty id in the JSON:API linkage.

The existing serializer specs didn't catch this because they hand-fed a fully-populated related hash, which the real datasource never produces.

Fix

Override has_one_relationship in ForestSerializer. For a PolymorphicManyToOne it rebuilds the linkage id from the foreign key (*_id + *_type) already loaded on the owner record — no extra query — and returns nil (omitting the data key) when the relation is unlinked. Both the linkage and the compound included document go through this method, so a single override covers both.

This is the workaround reported in #332, moved into the gem.

Tests

Added two specs reproducing the real datasource output (phantom related object) — both fail without the fix:

builds the linkage id from the foreign key when the related record is a phantom
omits the data key for an unlinked polymorphic relation

🤖 Generated with Claude Code

Note

Fix `PolymorphicManyToOne` linkage id serialization in `ForestSerializer`

When a polymorphic many-to-one related record is a phantom (no primary key), the serializer now rebuilds the linkage id from the owner's foreign key fields instead of returning an empty or broken relationship.

Overrides has_one_relationship in ForestSerializer to post-process PolymorphicManyToOne relationships, computing id from the foreign key value and resolving the type using foreign_key_targets.
Namespaced type values (e.g. Admin::User) are formatted to Admin__User to match collection naming conventions, and custom primary keys are respected.
Relationships with a nil foreign key or nil type omit the data member entirely.
Risk: any consumer relying on the previous (broken) phantom linkage behavior will see different output.

Changes since #333 opened

Added test verifying PolymorphicManyToOne relationship serialization [618ae03]
Clarified comments for PolymorphicManyToOne relationship handling [618ae03]

^{Macroscope summarized 870bad8.}

…oreign key (#332) A polymorphic belongs_to can't be SQL-joined, so the datasource returns a phantom related object without a primary key. The serializer read the id off that phantom, producing an empty id that crashes the frontend list view ("Expected id to be a string or number, received null"). Rebuild the linkage id from the foreign key already loaded on the owner record (no extra query), and omit the data key for unlinked records. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ation is loaded The previous commit fabricated a relationship linkage even when the relation was never projected (e.g. update/store responses expose only the FK columns, not the related object). That made the serializer resolve the type-field value as a collection name and raise "Collection not found". Guard on super returning nil: rebuild the id only for the phantom object that list projections return; otherwise keep the prior behavior. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

qltysh · 2026-07-21T12:46:54Z

1 new issue

Tool	Category	Rule	Count
qlty	Structure	Function with high complexity (count = 5): has_one_relationship	1

qltysh · 2026-07-21T12:46:55Z

+        return nil if foreign_key.nil? || foreign_type.nil?
+
+        primary_key = (field.foreign_key_targets || {})[foreign_type] || 'id'
+        (object.is_a?(Hash) ? object : {}).merge(primary_key => foreign_key)


Function with high complexity (count = 5): has_one_relationship [qlty:function-complexity]

…ets lookup foreign_key_targets is keyed by formatted collection names (Admin::User => Admin__User), but the type column stores the raw class name. For a namespaced target with a custom primary key the lookup missed and fell back to 'id', storing the FK under the wrong field and still emitting an empty linkage id. Format the type value before the lookup. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

matthv

Nit: comments a bit verbose. The foreign_key_targets keying one justifies the gsub → keep. But # nil means… mostly restates return object, and no extra query is issued. reads as pre-empting a reviewer. Could trim both, wdyt?

The specs only assert the linkage, not the included doc (#332's 2nd symptom, the empty self-link). Could be good to add one with include: 'documentable' asserting included has id => '10'?

…333 review) Add a spec asserting the compound `included` document builds its id and self-link from the foreign key (#332's second symptom). Trim the verbose comments per review, keeping only the non-obvious foreign_key_targets note. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

bexchauveto · 2026-07-21T13:55:21Z

Both done in 618ae03:

Trimmed the comments — dropped the nil means… block (now a one-line inline note) and No extra query is issued; kept the foreign_key_targets note since the gsub isn't obvious.
Added a spec that serializes with include: 'documentable' and asserts the included resource has id => '10' and self => /forest/Car/10 (the empty self-link symptom). Verified it fails without the fix (got nil).

Full suite green (724 examples).

matthv

ok for me 👍

## [1.35.4](v1.35.3...v1.35.4) (2026-07-21) ### Bug Fixes * **forest_admin_agent:** build PolymorphicManyToOne linkage id from foreign key ([#332](#332)) ([#333](#333)) ([177b78f](177b78f))

forest-bot · 2026-07-21T14:10:27Z

🎉 This PR is included in version 1.35.4 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀

bexchauveto and others added 2 commits July 21, 2026 14:37

qltysh Bot reviewed Jul 21, 2026

View reviewed changes

macroscopeapp Bot reviewed Jul 21, 2026

View reviewed changes

Comment thread packages/forest_admin_agent/lib/forest_admin_agent/serializer/forest_serializer.rb Outdated

bexchauveto mentioned this pull request Jul 21, 2026

PolymorphicManyToOne serializes an empty id in list view → frontend crashes with "Expected id to be a string or number, received null" #332

Closed

matthv reviewed Jul 21, 2026

View reviewed changes

matthv approved these changes Jul 21, 2026

View reviewed changes

bexchauveto merged commit 177b78f into main Jul 21, 2026
48 checks passed

bexchauveto deleted the fix/polymorphic-many-to-one-empty-id branch July 21, 2026 14:04

forest-bot added the released label Jul 21, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(forest_admin_agent): build PolymorphicManyToOne linkage id from foreign key (#332) - #333

fix(forest_admin_agent): build PolymorphicManyToOne linkage id from foreign key (#332)#333
bexchauveto merged 4 commits into
mainfrom
fix/polymorphic-many-to-one-empty-id

bexchauveto commented Jul 21, 2026 •

edited by macroscopeapp Bot

Loading

Uh oh!

qltysh Bot commented Jul 21, 2026

Uh oh!

qltysh Bot Jul 21, 2026

Uh oh!

Uh oh!

matthv left a comment

Uh oh!

bexchauveto commented Jul 21, 2026

Uh oh!

matthv left a comment

Uh oh!

Uh oh!

forest-bot commented Jul 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

bexchauveto commented Jul 21, 2026 • edited by macroscopeapp Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Root cause

Fix

Tests

Fix PolymorphicManyToOne linkage id serialization in ForestSerializer

Changes since #333 opened

Uh oh!

qltysh Bot commented Jul 21, 2026

1 new issue

Uh oh!

qltysh Bot Jul 21, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

matthv left a comment

Choose a reason for hiding this comment

Uh oh!

bexchauveto commented Jul 21, 2026

Uh oh!

matthv left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

forest-bot commented Jul 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

bexchauveto commented Jul 21, 2026 •

edited by macroscopeapp Bot

Loading

Fix `PolymorphicManyToOne` linkage id serialization in `ForestSerializer`