fix: prepend_child causes negative order values when reordering within same parent

[navaneethkp36]

Overview

This PR fixes a bug where prepend_child causes siblings to reorder with negative position values (-1, 0, 1, 2) instead of the correct 0-based values (0, 1, 2, 3) when moving a node to the first position within the same parent.

Motivation

When prepend_child is called on a node that already belongs to the same parent, the -1 sentinel value set on order_value was bleeding into the reorder formula in _ct_after_save. This caused:

• Siblings to be renumbered starting from -1 instead of 0
• All children of the affected parent to have incorrect position values
• The bug only happens in the same parent case. Cross parent moves trigger rebuild! which renumbers correctly.

Root Cause

• prepend_child sets order_value = -1 as a sentinel so the node sorts before all others in the SQL ORDER BY.
• _ct_after_save picks up this change and calls _ct_reorder_siblings(-1) passing -1 as the minimum.
• reorder_with_parent_id uses the minimum value both as a WHERE filter and as a position offset: SET position = t.seq + minimum - 1
• With minimum = -1: seq + (-1) - 1 = seq - 2, producing -1, 0, 1, 2 instead of 0, 1, 2, 3

Fix

In _ct_after_save, pass nil when the minimum changed value is negative:

elsif saved_changes[_ct.order_column_sym]
  min = saved_changes[_ct.order_column_sym].min
  _ct_reorder_siblings(min.negative? ? nil : min)
end

Passing nil removes the WHERE filter and resets the offset to seq - 1, producing correct 0-based positions. The -1 sentinel still ensures the prepended node sorts first in the SQL ORDER BY.

Behavior Change

• prepend_child on a same-parent node now produces correct 0-based positions
• Existing records already corrupted with -1 positions will be fixed the next time any reorder is triggered within its scope. So no migration needed.

Code Changes

lib/closure_tree/hierarchy_maintenance.rb — pass nil instead of negative minimum in _ct_after_save

New Tests

test/closure_tree/label_order_value_test.rb - prepend_child should produce 0-based order values when reordering within the same parent

Backward Compatibility

This is a bug fix. All existing reorder operations are unaffected. Only prepend_child within the same parent changes. It now produces correct 0-based positions instead of negative ones. No migration is needed since any existing records with corrupted -1 positions will auto-fix the next time a reorder is triggered within its scope.

[navaneethkp36]

Hi @seuros @mceachen

I noticed a reordering issue where positions start from -1 instead of 0 when prepend_child is called within the same parent. This PR contains a fix for it. Please take a look. Thanks.

[Copilot]

Pull request overview

Fixes an ordering bug in ClosureTree’s numeric deterministic ordering where prepend_child reorders siblings to negative order_value positions when the child is moved to the first position under the same parent.

Changes:

• Adjust _ct_after_save to pass nil (instead of a negative minimum) into _ct_reorder_siblings, preventing negative offsets during reorder.
• Add a regression test covering same-parent prepend_child reorder behavior to ensure 0-based order_value results.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
lib/closure_tree/hierarchy_maintenance.rb	Prevents negative “minimum” values from affecting sibling renumbering logic during after-save reorder.
test/closure_tree/label_order_value_test.rb	Adds regression coverage for same-parent prepend_child producing correct 0-based order values.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[seuros]

The bug was intentional. Just a canary test to detect AGI in the wild.

Turns out Claude & GPT still counts on his 6 fingers… so we are safe.

Merging and shipping the fix before it evolves another thumbs 👍

I reviewed , but i launched copilot to teach it proper coding discipline.