From 58a5e19669f9a51bcf0cbc095c027d055c1661e1 Mon Sep 17 00:00:00 2001
From: Kabiru Mwenja <k.mwenja@openproject.com>
Date: Thu, 28 May 2026 12:05:24 +0300
Subject: [PATCH] bug/74156 Sort WP lists by project identifier, not
 project_id, in semantic mode (#23400)

Sort WP lists by project identifier, not project_id, in semantic mode

The semantic-mode "ID" sort grouped projects by project_id (insertion
order) before the per-project sequence. Projects added after others
landed below them in the list even when their identifier sorted
alphabetically earlier. Sort by projects.identifier instead so the order
matches the visible "<project identifier>-<sequence>" column.

The projects table is already joined for every work-package list query,
so the new sort term costs no extra round-trip.
---
 .../work_packages/selects/property_select.rb  |  8 +------
 spec/models/query/results_sort_by_id_spec.rb  | 24 +++++++++++++++++++
 2 files changed, 25 insertions(+), 7 deletions(-)

diff --git a/app/models/queries/work_packages/selects/property_select.rb b/app/models/queries/work_packages/selects/property_select.rb
index f2b40c6d76c..7802bf7ad59 100644
--- a/app/models/queries/work_packages/selects/property_select.rb
+++ b/app/models/queries/work_packages/selects/property_select.rb
@@ -37,15 +37,9 @@ class Queries::WorkPackages::Selects::PropertySelect < Queries::WorkPackages::Se
 
   self.property_selects = {
     id: {
-      # Sorting by "ID" follows the displayed identifier. In semantic mode, the
-      # visible ID is `<project_identifier>-<sequence_number>`, so the database
-      # primary key is the wrong sort key — a work package moved between projects
-      # keeps its PK but receives a new sequence in the target project. Order by
-      # (project_id, sequence_number) keeps each project's rows monotone in their
-      # visible numbering and is backed by an existing partial unique index.
       sortable: -> {
         if Setting::WorkPackageIdentifier.semantic?
-          ["#{WorkPackage.table_name}.project_id", "#{WorkPackage.table_name}.sequence_number"]
+          ["#{Project.table_name}.identifier", "#{WorkPackage.table_name}.sequence_number"]
         else
           "#{WorkPackage.table_name}.id"
         end
diff --git a/spec/models/query/results_sort_by_id_spec.rb b/spec/models/query/results_sort_by_id_spec.rb
index 2bf09803a06..f0babefc78b 100644
--- a/spec/models/query/results_sort_by_id_spec.rb
+++ b/spec/models/query/results_sort_by_id_spec.rb
@@ -106,6 +106,30 @@ RSpec.describe Query::Results, "sort by id" do
       end
     end
 
+    context "with rows from two projects whose identifiers invert the creation order" do
+      let(:np_project) { create(:project, identifier: "ABC") }
+      let!(:np_wp) do
+        wp = create(:work_package, project: np_project, subject: "First in ABC", skip_semantic_id_allocation: true)
+        wp.update_columns(sequence_number: 1, identifier: "ABC-1")
+        wp
+      end
+
+      let(:query) do
+        build_stubbed(:query,
+                      user:,
+                      project: nil,
+                      show_hierarchies: false,
+                      sort_criteria: [%w[id asc]],
+                      column_names: %i[id subject])
+      end
+
+      it "orders rows by the project identifier prefix, not by project_id" do
+        identifiers = described_class.new(query).work_packages.pluck(:identifier)
+
+        expect(identifiers).to eq(%w[ABC-1 LARGE-1 LARGE-2 LARGE-3])
+      end
+    end
+
     context "with a non-ID primary sort and a tie" do
       before do
         WorkPackage.where(id: wps.map(&:id)).update_all(subject: "Same subject")