paper_trail_diff 0.12.0 → 0.13.1

data/README.md

@@ -1,1453 +1,244 @@
 # paper_trail_diff
 
-> [!WARNING]
-> **Experimental software and AI disclosure:** This gem is experimental and may
-> contain incomplete behavior, reconstruction errors, performance problems, or
-> breaking changes. AI-assisted coding tools made significant contributions to
-> its design, implementation, tests, and documentation under maintainer
-> direction. AI involvement is not a substitute for independent review: audit
-> the code and validate it against your own PaperTrail history before relying on
-> it in production, compliance, security, or other high-stakes workflows.
+Compare past and present records using the history saved by
+[PaperTrail](https://github.com/paper-trail-gem/paper_trail). Show changed fields,
+include changes to associated records, or build a report for many records.
+Results are Ruby objects with methods for reading the changes and `to_h` for
+turning them into hashes. The gem does not generate display text or save changes.
 
-`paper_trail_diff` adds structured endpoint, checkpoint-timeline, and activity
-comparisons to
-[PaperTrail](https://github.com/paper-trail-gem/paper_trail). It returns immutable
-Ruby value objects and hashes rather than formatted text.
+If you already use `has_paper_trail` and `record.versions`, you can pass those
+versions directly to this gem. You do not need to call `reify` yourself.
 
-PaperTrail is required. Explicit, bounded association-path comparison is
-available when
-[paper_trail-association_tracking](https://github.com/westonganger/paper_trail-association_tracking)
-(PT-AT) is installed, loaded, and enabled.
+## Get started
 
-Ruby 3.1 or newer and PaperTrail 16 or 17 are supported.
-
-New to the gem? Run [`examples/demo.rb`](examples/demo.rb) to see it work in
-about ten seconds — one self-contained file, no application and no migrations,
-which fetches what it needs on first run:
+Add the gem to your Rails application:
 
 ```console
-ruby demo.rb
+bundle add paper_trail_diff
 ```
 
-Then start with the copyable [Quickstart](QUICKSTART.md), or browse
-[the demo application](https://github.com/aheathwilliams/paper_trail_summary_test_app)
-— a small Rails app that installs a published release and exercises scalar,
-nested, through-association, and HABTM history against it.
+Already have a model with `has_paper_trail`? Try the comparison below.
+Otherwise, follow the [Rails quickstart](QUICKSTART.md) to set up PaperTrail
+and create your first comparison.
 
-## Installation
+To try it without Rails, download [examples/demo.rb](examples/demo.rb) and run
+`ruby demo.rb`. It installs its dependencies on the first run. The
+[demo application](https://github.com/aheathwilliams/paper_trail_summary_test_app)
+shows reports in a Rails UI.
 
-Add the gem to your bundle:
+Ruby 3.1+ and PaperTrail 16 or 17 are supported, subject to
+[PaperTrail's Ruby and ActiveRecord requirements](https://github.com/paper-trail-gem/paper_trail#1a-compatibility).
+Comparing association history also requires
+[paper_trail-association_tracking](https://github.com/westonganger/paper_trail-association_tracking),
+an optional PaperTrail plugin. The quickstart includes its setup.
 
-```ruby
-gem "paper_trail_diff"
-```
+> **Experimental:** AI-assisted coding tools made significant contributions to
+> this project. Check results against your application's history before relying
+> on them. Changes that were never recorded cannot be recovered.
 
-For association history, also add and configure PT-AT using its migrations:
+## 1. Compare an earlier version with the current record
 
-```ruby
-gem "paper_trail-association_tracking"
-```
+This example assumes an `Article` model with a `title` column and `has_paper_trail`:
 
+<!-- executable:recipe-compare -->
 ```ruby
-PaperTrail.config.track_associations = true
-```
+article = Article.create!(title: "Draft")
+article.update!(title: "Published")
 
-All associated models whose historical state is compared must use
-`has_paper_trail`.
+# PaperTrail stores the values from BEFORE the update.
+before = article.versions.last
+diff = PaperTrailDiff.compare(before, article)
 
-## How PaperTrail records state
-
-Almost everything else in this README follows from one property of PaperTrail,
-so it is worth being precise about it first: **a version stores the state that
-existed _before_ the event that created it.** A version is a record of what was
-overwritten, not of what was written.
-
-Take an article created as `"Draft"`, then updated to `"Published"`, then to
-`"Final"`:
-
-```text
-        v1              v2                v3            (no version)
-     "create"        "update"          "update"
-         |               |                 |                 |
-   ──────●───────────────●─────────────────●─────────────────●──────▶ time
-         |               |                 |                 |
-         └─── "Draft" ───┘                 |                 |
-              stored in v2                 |                 |
-                         └─ "Published" ───┘                 |
-                              stored in v3                   |
-                                           └──── "Final" ────┘
-                                             only in the table
+diff.attributes.fetch("title").to_h
+# => { from: "Draft", to: "Published" }
 ```
 
-Each state is stored by the version at the *end* of the interval it was live
-for. Three consequences run through the rest of this document:
-
-- **A `create` version reifies to `nil`.** Nothing preceded it. Comparing it
-  with a later state reports a structured `record_presence_change` rather than
-  inventing scalar changes.
-- **The newest state has no version at all.** It exists only in the table. A
-  fully historical result therefore needs a version *later* than the last change
-  it should reveal, which is why a `within:` window can raise
-  `IncompleteTimeRangeError`, and why `activity_timeline(..., to: article)`
-  exists for ending at current state instead.
-- **A change is visible between two boundaries**, never "at" one. This is why
-  both timeline APIs return steps rather than events.
-
-The [Quickstart](QUICKSTART.md) walks through the same idea against a real
-console session, creating the models it uses as it goes.
-
-## Which models and schemas work
+PaperTrail's `version.reify` returns the record as it was before that version's
+change. Passing a version to `compare` uses that same earlier state. Passing
+`article` uses its current saved values. Pass two versions to compare two
+historical states.
 
-Anything `has_paper_trail` tracks, including these, which are covered by the
-test suite rather than assumed:
+PaperTrail's `version.changeset` describes one recorded change. `compare`
+describes the net difference between the two states you choose. If a field
+changes and then changes back, there is no net difference for that field. To
+see every change instead, use `PaperTrailDiff.timeline(article)`, which covers
+the article's whole history through its current saved values.
 
-- **Single-table inheritance.** Results are keyed by the *base* class, matching
-  what PaperTrail writes to `item_type`. Analyzing a `Book < Publication`
-  produces the key `["Publication", "12"]`, so `results.fetch(["Book", ...])`
-  raises `KeyError`. Build keys with `PaperTrailDiff::Endpoint.identity(record)`
-  rather than by hand.
-- **Non-integer primary keys** on the tracked model, such as a UUID column.
-- **A custom version class**, via `has_paper_trail versions: { class_name: }`.
-  Nothing here references `PaperTrail::Version` directly.
+An attribute appears in `diff.attributes` only when it changed. `diff.empty?`
+tells you whether anything changed. A PaperTrail `create` version has no earlier
+record (`reify` returns `nil`), so comparing it with the current record reports
+`diff.record_presence_change`, meaning the record was added.
 
-### Versions that cannot be ordered
+To include comments, add `associations: [:comments]`. If comments have replies,
+`associations: ["comments.replies"]` includes those too. Associated models must
+also use `has_paper_trail`, and association tracking must be enabled.
 
-Versions are ordered by `created_at`, falling back to the id when timestamps
-tie. That fallback recovers the real sequence only while ids increase with
-insertion. An autoincrement id does; a **UUID version id does not**.
-
-So two conditions together are unsafe: version ids that are not sequential,
-*and* versions sharing a timestamp — which is ordinary on a MySQL `datetime`
-column, since it stores whole seconds. The order is then unrecoverable, and
-rather than report a plausible-looking history in the wrong order the gem
-raises `PaperTrailDiff::AmbiguousVersionOrderError` naming both versions.
-`diagnose` reports the same condition as an error, so a caller can check before
-running rather than after a surprise.
-
-Either half alone is fine. Sequential ids order tied timestamps correctly, and
-distinct timestamps never reach the fallback.
-
-### Association changes across a shared timestamp
-
-Sequential ids recover the *order* of tied versions, but not their association
-state. PT-AT records association membership per version and resolves it by
-timestamp, so two versions sharing one are indistinguishable for that purpose:
-**an association change between them is invisible.** It is not reported as
-absent — the association simply does not appear in the diff.
-
-Unlike an unorderable sequence this is not always wrong. If nothing associated
-changed between the two versions, the result is correct, and the gem cannot
-tell the two cases apart, because not seeing the change is the symptom. So it
-does not raise. Instead:
-
-- `diagnose` reports `:tied_version_timestamps` as a **warning** whenever
-  `associations:` are selected and versions in range share a timestamp.
-- a comparison in that state emits
-  `ambiguous_association_boundary.paper_trail_diff` through ActiveSupport
-  notifications, carrying the two version ids and the shared timestamp. Nothing
-  is logged; subscribe if you want it surfaced.
-
-Writing several versions inside one clock tick is easy to do — a service object
-that touches a record and its children in quick succession will manage it — so
-this is worth checking before trusting an association history. Recording
-versions at sub-second precision separates them.
-
-### ActiveStorage attachments
-
-`has_one_attached` and `has_many_attached` point at `ActiveStorage::Attachment`
-and `ActiveStorage::Blob`, which Rails owns and PaperTrail does not version.
-There is no history behind them, so a historical comparison over those paths
-raises `PaperTrailDiff::UnversionedAssociationError` rather than reporting that
-nothing changed. Adding `has_paper_trail` to Rails' own models does not rescue
-it either: `has_one_attached` is a scoped `has_one`, and reifying it from
-history is ambiguous.
-
-Audit the attachment through a model you own instead. Give it the facts worth
-auditing as ordinary columns, and version it:
+For nested results, `each_change` lets you read changes without walking every
+association yourself:
 
 ```ruby
-class DocumentRevision < ApplicationRecord
-  belongs_to :attachable, polymorphic: true
-  has_one_attached :file
-  has_paper_trail
-
-  # Attaching a file writes to ActiveStorage's tables, not to this record, so
-  # nothing would be versioned without copying the facts across.
-  after_save :record_file_metadata, if: -> { file.attached? }
-
-  def record_file_metadata
-    blob = file.blob
-    return if filename == blob.filename.to_s && checksum == blob.checksum
-
-    update_columns(
-      filename: blob.filename.to_s, content_type: blob.content_type,
-      byte_size: blob.byte_size, checksum: blob.checksum
-    )
-  end
-end
-```
-
-```ruby
-class Article < ApplicationRecord
-  has_many :document_revisions, as: :attachable
-  has_paper_trail
+diff.each_change do |entry|
+  entry.kind              # e.g. :attribute_changed or :record_added
+  entry.association_path  # e.g. ["comments", "replies"]
+  entry.value             # the changed values or the added/removed record's data
 end
-
-PaperTrailDiff.compare(before, after, associations: [:document_revisions])
-```
-
-The bytes stay in ActiveStorage; the *auditable facts* live where PaperTrail
-can see them. Nothing about the column names matters to this gem — it reports
-whichever columns changed. What matters is that replacing a file writes to a
-versioned record, because attaching one on its own does not.
-
-## Choosing an entry point
-
-| You need | Call |
-| --- | --- |
-| The net difference between two endpoints | `compare` |
-| The same, for many records in one pass | `compare_many` |
-| One step per version of the root record | `timeline` |
-| One step per version of the root *or a selected child* | `activity_timeline` |
-| A net difference and a timeline from one history pass | `analyze` |
-
-`timeline` and `activity_timeline` differ only in which recorded versions
-become boundaries. Given an article with two comment edits between two article
-versions:
-
-```text
-recorded versions   A1        C1      C2        A2      A = Article version
-                    |         |       |         |       C = Comment version
-                ────●─────────●───────●─────────●────▶ time
-                    |         |       |         |
-timeline            └───────── 1 step ──────────┘
-                    both comment changes land inside that one step
-
-activity_timeline   └── 1 ────┴── 2 ──┴─── 3 ───┘
-                    each recorded version becomes its own boundary
 ```
 
-Both report the same underlying data: a `timeline` step still contains every
-selected association, because it has to describe what changed beneath the root.
-They differ in how finely that change is split, and in what the reconstruction
-costs. That cost trade-off is covered under
-[choosing a reconstruction strategy](#choosing-between-timeline-and-analyzeactivity-true).
+## 2. Show who changed a record and its comments
 
-## Compare two endpoints
-
-`compare` accepts two explicit endpoints: each may be a PaperTrail version or a
-clean, persisted model instance representing current database state. It reports
-only their net difference:
+This example requires the association setup in the quickstart. PaperTrail's
+`whodunnit` records who made a change; here we set it with `PaperTrail.request`.
+The timeline covers the article's whole history, including comment changes made
+after the article's last version.
 
+<!-- executable:recipe-activity -->
 ```ruby
-diff = PaperTrailDiff.compare(article.versions[1], article.versions[4])
-
-diff.attributes["title"].from # => "Draft"
-diff.attributes["title"].to   # => "Published"
-diff.empty?                    # => false
-diff.to_h
-# => {
-#   record_presence_change: nil,
-#   attributes: {
-#     "title" => { from: "Draft", to: "Published" }
-#   },
-#   associations: {}
-# }
-```
-
-Intermediate edits do not affect `compare`. If a title changes and later
-returns to its original value, endpoint comparison reports no title change.
-
-`record_presence_change` reports whether the root record *existed* at each
-endpoint, and is `nil` for the ordinary case where it existed at both. It is
-populated when one endpoint reifies to `nil` — most often a `create` version,
-whose pre-change state is the absence of the record — and it then carries whole
-`RecordSnapshot` values rather than fake scalar changes from `nil`.
-
-Destroying the root record is not reported this way, and the reason follows
-from the [pre-change model](#how-papertrail-records-state): the state *at* a
-`destroy` version is the state immediately before the deletion, so the record
-is still present there. A comparison ending at a `destroy` version reports that
-last edit, not the deletion. Deleting a *selected child* is reported normally,
-as a `removed` member of its parent's collection, and `activity_timeline`
-reports a destroyed root through a
-[closing removal step](#closing-a-destroyed-root).
-
-Pass the record explicitly when the desired endpoint is current state:
-
-```ruby
-diff = PaperTrailDiff.compare(
-  article.versions.last,
-  article,
-  associations: ["comments.replies"]
-)
-```
-
-Two version endpoints must be given in chronological order. A transposed pair
-produces the inverse diff, which is easy to do by accident and impossible to
-detect afterwards because a result carries no direction of its own, so it
-raises `PaperTrailDiff::ReversedEndpointsError` instead. Nothing is lost by
-this: the two orders differ only in which side of each change is `from`.
-
-A current-record endpoint is exempt and may appear on either side, because it
-is self-evidently the live state and placing it first is a deliberate reverse
-comparison. Current state is never inferred. The record must be persisted, not
-destroyed, and free of unsaved attribute changes. The gem reloads it unscoped
-before normalization, so stale association caches and in-memory edits are not
-compared. Use a database
-transaction with an appropriate isolation level when several live association
-queries must represent one atomic application snapshot.
-
-For collection-level reports, use `compare_many` to reload current roots and
-preload each explicitly selected live association path across the batch. Each
-entry has the same endpoints and options as `compare`; results are returned in
-input order as a frozen hash keyed by `[item_type, item_id]` strings:
-
-Endpoints are still supplied by the caller, so look them up in bulk too —
-otherwise the per-root queries this API removes come straight back. Two
-queries resolve the earliest version of every root, whatever the batch size:
-
-```ruby
-orders = Order.where(id: order_ids).to_a
-
-earliest_ids = PaperTrail::Version
-  .where(item_type: "Order", item_id: orders.map(&:id))
-  .group(:item_id)
-  .minimum(:id)
-first_versions = PaperTrail::Version
-  .where(id: earliest_ids.values)
-  .index_by(&:item_id)
-```
-
-```ruby
-diffs = PaperTrailDiff.compare_many(
-  orders.map { |order| { from: first_versions.fetch(order.id), to: order } },
-  associations: [:line_items],
-  ignore: []
-)
-
-diffs.fetch(["Order", orders.first.id.to_s]) # => PaperTrailDiff::Diff
-```
-
-Endpoints may also be given as `:first` or `:last`, resolved against the record
-the pair's other endpoint names. That replaces the lookup above entirely, and
-resolves every root in two queries per model class rather than one per root:
-
-```ruby
-diffs = PaperTrailDiff.compare_many(
-  orders.map { |order| { from: :first, to: order } },
-  associations: [:line_items]
-)
-```
-
-A symbol carries no identity of its own, so `{ from: :first, to: :last }` raises
-rather than guessing. A root with no recorded history resolves to an empty
-`Diff`, matching how the timeline APIs answer the same question.
-
-Root identities must be unique within one call. Historical reconstruction for
-ordinary versioned, unscoped association paths is also prepared across the
-collection. Paths that require the existing point-in-time PT-AT fallback retain
-their per-endpoint behavior and all historical reconstruction retains the same
-PaperTrail Association Tracking requirements as `compare`. Live collection
-scopes with per-owner semantics, such as `limit`, `offset`, or an owner
-argument, are loaded per root; other selected branches remain batched. Callers
-should still use an appropriate database transaction when all live queries
-must observe one atomic snapshot.
-
-### Look inside a JSON column
-
-A `json` or `jsonb` column reifies to one Hash, so an attribute diff can only
-say the blob changed. `nested_changes` says which keys changed:
-
-```ruby
-diff = PaperTrailDiff.compare(from_version, to_version)
-
-PaperTrailDiff.nested_changes(diff.attributes.fetch("config"))
-# => { ["theme"]          => <from "dark" to "light">,
-#      ["limits", "max"]  => <from 10 to 20> }
-```
-
-It reads a column that stores JSON as text as readily as a native one, and also
-takes a bare pair: `PaperTrailDiff.nested_changes(from_value, to_value)`.
-
-Nothing about the attribute diff itself changes. A pair that is not two readable
-structures — text on one side and JSON on the other, or a value that does not
-parse — reports nothing, and the caller still has the whole-value change it
-already had.
-
-Three things about the shape are worth knowing:
-
-**Paths are arrays, not dotted strings.** A JSON key may itself contain a dot, so
-`"a.b"` would be ambiguous between one key and two:
-
-```ruby
-PaperTrailDiff.nested_changes({ "a.b" => 1, "a" => { "b" => 1 } },
-                              { "a.b" => 2, "a" => { "b" => 3 } })
-# => { ["a.b"] => <from 1 to 2>, ["a", "b"] => <from 1 to 3> }
-```
-
-**Arrays are reported by membership, not by position.** Elements carry no
-identity, so an insertion at the front would make every later index look
-changed — one insertion described as several edits. An `ArrayChange` says what
-was added and removed, matching by value and respecting duplicates:
-
-```ruby
-change = PaperTrailDiff.nested_changes(
-  { "keywords" => %w[apollo nasa] },
-  { "keywords" => %w[saturn apollo nasa] }
-).fetch(["keywords"])
-
-change.added      # => ["saturn"]
-change.removed    # => []
-change.reordered? # => false
-change.from       # => ["apollo", "nasa"]   (the whole values remain)
-```
-
-Membership cannot see one thing, so it is named rather than passed over: the
-same elements in a new order report `reordered?` as true with nothing added or
-removed.
-
-An element that is itself an object is reported whole. Saying which field of
-which object changed would need a pairing of before-elements with
-after-elements, and nothing in the value licenses that pairing.
-
-**An absent key is not a null one.** `{"a": null}` and `{}` mean different things
-in JSON, and an audit trail that showed them alike would be lying about one of
-them, so a missing key reads as `PaperTrailDiff::NestedComparator::ABSENT`
-rather than `nil`.
-
-### Reuse already-preloaded current endpoints
-
-`compare`, `compare_many`, and `activity_timeline(..., to: record)` reload
-current endpoints by default. A caller that already owns a consistent, fully
-preloaded graph may opt out. Each endpoint is loaded once per call, so a batch
-that resolves every root up front and then reads one of them again pays
-nothing the second time, and every root in one result reflects the same instant.
-
-The option is not accepted by `timeline`, `analyze`, or `analyze_many`. Those
-are bounded by versions and read live state only under
-[`close_on: :current`](#reporting-up-to-now), which always reloads:
-
-```ruby
-orders = Order.where(id: order_ids).preload(line_items: :product).to_a
-# `first_versions` is the same bulk endpoint lookup shown above.
-
-diffs = PaperTrailDiff.compare_many(
-  orders.map do |order|
-    { from: first_versions.fetch(order.id), to: order }
-  end,
-  associations: ["line_items.product"],
-  reload_live_endpoints: false
-)
-```
-
-With `reload_live_endpoints: false`, scalar values and association targets come
-from the supplied instances. Every requested association path must already be
-loaded; otherwise `PaperTrailDiff::UnloadedAssociationError` is raised before
-normalization instead of silently issuing an N+1 query. The caller is
-responsible for the consistency and freshness of that in-memory graph.
-
-### Runtime performance diagnostics
-
-The gem never prints or logs automatically. It emits these
-`ActiveSupport::Notifications` events so applications can choose their logger
-and formatting:
-
-- `compare.paper_trail_diff`
-- `compare_many.paper_trail_diff`
-- `activity_timeline.paper_trail_diff`
-- `load_live_endpoints.paper_trail_diff`
-- `prepare_history.paper_trail_diff`
-
-```ruby
-ActiveSupport::Notifications.subscribe(/\.paper_trail_diff\z/) do |name, start, finish, _id, payload|
-  Rails.logger.debug(
-    event: name,
-    duration_ms: ((finish - start) * 1_000).round(1),
-    **payload
-  )
-end
-```
-
-The payloads contain counts, model names, association paths, and reload mode,
-not endpoint objects or record attributes. The activity-timeline event wraps
-the complete call and includes `step_count`, so its notification duration is
-the user-visible runtime. To inspect the exact SQL generated by one report,
-scope a standard `sql.active_record` subscriber around it:
-
-```ruby
-callback = proc do |_name, _start, _finish, _id, payload|
-  Rails.logger.debug(payload[:sql]) unless payload[:cached] || payload[:name] == "SCHEMA"
-end
-
-ActiveSupport::Notifications.subscribed(callback, "sql.active_record") do
-  PaperTrailDiff.compare_many(comparisons, associations: [:line_items])
+article = Article.create!(title: "Draft")
+PaperTrail.request(whodunnit: "alice") do
+  Article.transaction do
+    article.update!(title: "Published")
+    article.comments.create!(body: "Please review")
+  end
 end
-```
-
-## Analyze many records over one window
-
-`analyze_many` answers "what changed for these records during this period" for a
-whole collection, selecting every root's versions and preparing their selected
-association history once for the batch:
-
-```ruby
-results = PaperTrailDiff.analyze_many(
-  Order.where(status: "open").to_a,
-  within: Time.zone.parse("2026-08-01")...Time.zone.parse("2026-09-01"),
-  associations: [:line_items]
-)
-
-results.fetch(["Order", order.id.to_s]).diff      # net change across the window
-results.fetch(["Order", order.id.to_s]).timeline  # its checkpoint steps
-```
-
-Results are a frozen hash keyed by `[item_type, item_id]` strings, and each
-value is the same `Analysis` that `analyze` returns for one record. A root with
-no versions inside the window gets an empty `Analysis` rather than raising, so a
-listing page needs no special case. Root identities must be unique.
-
-### Letting a relation choose the roots
-
-Assembling that array means querying `versions` for which roots moved, filtering,
-and loading the records — the same selection this gem already performs.
-`analyze_scope` does it for you:
-
-```ruby
-analyses, unreachable = PaperTrailDiff.analyze_scope(
-  Order.where(status: "open"),
-  within: Time.zone.parse("2026-08-01")...Time.zone.parse("2026-09-01"),
-  associations: [:line_items],
-  limit: 500
-)
-```
-
-`analyses` is the same frozen hash `analyze_many` returns. It is also available
-as `PaperTrailDiff.analyze_many(scope: ..., limit: ...)`, and it accepts a model
-class (`Order`) as readily as a relation. Every other option — `within:`,
-`version_scope:`, `associations:`, `ignore:`, `activity:`, `close_on:` — behaves
-exactly as it does for `analyze_many`, and the query cost stays flat in the
-number of roots.
-
-Two things to know before using it.
-
-**`limit:` is required, and exceeding it raises.** Selection moves into the gem
-here, so the bound on how much work one page can ask for has to move with it. It
-refuses rather than truncating, because an audit report that is quietly short is
-worse than one that fails.
-
-**`unreachable` names roots the relation could not reach.** A relation's
-conditions are evaluated against the live table, so a root destroyed during the
-window cannot be tested against them at all — its history is intact, and the
-state it held when it was destroyed may well have matched:
-
-```ruby
-unreachable # => [["Order", "412"]]
-```
-
-These are reported rather than dropped so that a page auditing deletions is told
-where to look instead of silently coming up short. `analyze_many` requires live
-records by design, so to analyze those roots reach for `activity_timeline`, which
-reads history without needing current state.
-
-Note also that a relation selects on **current** state, not on state during the
-window. `where(status: "open")` means open *now*, which is a different set from
-what was open while the window was open. If you need the historical population,
-select on the version history instead and pass the records.
-
-### Reporting on a subset of root mutations
-
-`version_scope:` narrows which root versions count as *selected mutations*, so a
-checkpoint timeline reports only some of them:
-
-```ruby
-user_edits = ->(scope) { scope.where.not(whodunnit: nil) }
-
-PaperTrailDiff.analyze_many(articles, within: window, version_scope: user_edits)
-PaperTrailDiff.timeline(article, within: window, version_scope: user_edits)
-```
-
-The hook receives the version relation for the range and returns a narrowed one.
-It is accepted by `timeline`, `activity_timeline`, `analyze`, and
-`analyze_many`, with any range form.
-
-**It sees root versions only.** Someone who edited only comments or tags has no
-root versions to select, so filtering to them yields an empty result even though
-they changed plenty. For "what did this person change?", reach for
-[one person's changes](#reporting-on-one-persons-changes) instead — this hook
-answers the different question of which *root checkpoints* a timeline reports.
-
-It filters *selected mutations only*. Versions the filter excludes are still
-loaded, because a version records the state before its own event: without the
-one that follows a selected change, whatever that change produced cannot be
-shown at all. Those extra versions are reconstruction context, so their own
-changes are never attributed to a selected mutation.
-
-Each selected mutation is bounded by the version that immediately followed it,
-not by the next selected one. So each step's diff is exactly what that mutation
-did, and it reads the same however many excluded changes happen to follow it:
-
-```
-versions   system → alice → system → bob → system
-reported   alice: what alice changed    bob: what bob changed
-```
-
-Given a user edit followed by a system edit, filtering to user changes yields
-one step running from the user version to the system version, whose diff is
-exactly the user's change. A selected mutation that nothing follows yet is not
-reported, since no version records the state it produced — the same blind spot
-an unfiltered timeline has at its `to:` boundary. A selected *destruction* is
-the exception: nothing can follow one, but the absence it leaves is what it
-produced, so `activity_timeline` still closes on it. A root with no selected
-mutation reports an empty `Analysis` rather than raising.
-
-The hook applies to root versions only. Under `activity: true` the filter
-decides where the span starts and ends, and `timeline` within that span reports
-only selected mutations, but `activity_timeline` still lists every boundary
-inside it — dropping one would fold the change it carried into a neighbouring
-step and credit it to whoever made that one.
-
-Omit `within:` to analyze each root's whole recorded history instead, which is
-the batched equivalent of `analyze(record, from: :first, to: :last)`. Explicit
-version endpoints are not accepted, because a single pair cannot mean the same
-thing for every root.
-
-Query cost is flat in the number of roots for the diff and timeline views:
-selecting versions and preparing history are both shared across the batch. On a
-twenty-root batch that is 7 queries against 120 for the same work done one
-record at a time. Passing `activity: true` also works and returns the same
-results, but discovering descendant events is inherently per-root, so that view
-batches far less.
-
-Roots are supplied as live records, so a root deleted inside the window cannot
-be included; use `activity_timeline` for a history that ends in a deletion.
-
-### Reporting on one person's changes
-
-An activity timeline already carries this. Every boundary in the span becomes a
-step, and a step's diff is exactly what the event at its `from_boundary` did, so
-that boundary's `whodunnit` is who made the change:
-
-<!-- executable:readme-person-changes -->
-```ruby
-steps = PaperTrailDiff.activity_timeline(
-  article, within: window, associations: %i[comments tags]
-)
-
-attributed = steps.reject(&:empty?)
-by_priya = attributed.select { |step| step.from_boundary.whodunnit == "Priya Shah" }
-
-by_priya.first.from_boundary.whodunnit  # "Priya Shah"
-by_priya.first.diff                     # exactly what that one event changed
-
-authors = attributed.group_by { |step| step.from_boundary.whodunnit }
-                    .transform_values(&:length)
-```
-
-This covers descendants, which is what `version_scope:` cannot do. On one demo
-history a contributor who only ever touched comments and authorships has four
-attributed activity steps and *zero* root steps.
-
-Filter in Ruby rather than asking the gem to select fewer versions. Attribution
-is only correct because every boundary is reconstructed: skip one and the
-snapshot carried into the next step is a state the record had already moved
-past, so the change lands on the wrong person. There is nothing to save by
-filtering earlier — the work is the reconstruction, not the comparison — and a
-predicate can say things a version relation cannot, such as several people at
-once or "anyone but the importer".
-
-`reject(&:empty?)` first, because a boundary that changed nothing you selected
-still produces a step. The final event in the range has no boundary after it, so
-what it produced is not shown — the same blind spot `to:` has everywhere else.
-
-For "which of these records did someone touch?" across a listing, note that
-`analyze_many` batches the diff and timeline views but not descendant discovery.
-Querying versions by `whodunnit` and mapping them back to roots yourself will be
-much cheaper than an activity pass per record.
-
-## Build a root-checkpoint timeline
-
-`timeline` accepts two version objects from the supplied record's history, or
-the symbols `:first` and `:last` when the range is simply the record's whole
-recorded history:
-
-```ruby
-steps = PaperTrailDiff.timeline(article, from: :first, to: :last)
-```
-
-`:first` and `:last` are resolved by the gem, independently of the order the
-`versions` association happens to use, so a caller never has to know it is
-sorted. They work anywhere a version does, including mixed with an explicit
-one (`from: :first, to: some_version`) and on `activity_timeline` and
-`analyze`. A record with no versions has no boundaries to resolve; that is an
-empty history rather than a bad request, so the result is an empty timeline
-instead of an error — which is usually what an index page wants.
 
-Combine `from: :first` with `to: article` for the fullest activity view of a
-live record: the whole recorded history, ending at current state. See
-[activity timelines](#build-an-activity-timeline) for why that end differs from
-`to: :last`.
-
-The range is inclusive, must be chronological, and produces one `Step` for each
-adjacent pair:
-
-```ruby
-steps = PaperTrailDiff.timeline(
-  article,
-  from: article.versions[1],
-  to: article.versions[4]
-)
-
-steps.first.from_version # the original PaperTrail version
-steps.first.to_version   # the next PaperTrail version
-steps.first.from_boundary # immutable presentation metadata
-steps.first.to_boundary
-steps.first.diff         # a PaperTrailDiff::Diff
-steps.first.empty?        # delegates to the diff
-steps.first.to_h
-# => { from_version_id: 2, to_version_id: 3, diff: { ... } }
-```
-
-Both timeline types expose `from_boundary`, `to_boundary`, `diff`, and
-`empty?`. A historical boundary has `event`, `whodunnit`, `record`,
-`recorded_at`, `version?`, `current?`, and `destroyed?` readers. Checkpoint `Step` objects
-also retain their original `from_version` and `to_version` for callers that
-need custom PaperTrail metadata. Existing `Step#to_h` and
-`ActivityBoundary#to_h` shapes remain unchanged; use the readers for the new
-metadata.
-
-Every version boundary remains in the result, even when its diff is empty after
-ignored fields are removed. Equal boundaries return a frozen empty array.
-
-This is deliberately a root-record timeline. Several child changes between two
-root versions appear together in the same step.
-
-## Build an activity timeline
-
-`activity_timeline` merges versions for the root and explicitly selected
-descendants. Child changes can therefore form separate steps instead of waiting
-to be aggregated at the next root checkpoint:
-
-```ruby
 steps = PaperTrailDiff.activity_timeline(
-  article,
-  from: article.versions[1],
-  to: article.versions[4],
-  associations: ["comments.replies", :author]
+  article, associations: [:comments], group: :transaction, snapshots: true
 )
 
-steps.reject(&:empty?)
-```
-
-Pass the record explicitly as `to:` to include current state without creating a
-final root checkpoint. This is not the same as `to: :last`: descendant
-mutations recorded after the record's final root version fall outside a
-version-bounded range, so only a live end reports them.
-
-```ruby
-steps = PaperTrailDiff.activity_timeline(
-  article,
-  from: article.versions[1],
-  to: article,
-  associations: ["comments.replies", :author]
-)
-```
-
-The result is a frozen array of `ActivityStep` objects. Each step has immutable
-`from_boundary`, `to_boundary`, and `diff` readers. Historical boundaries are
-`kind: :version`; an explicitly requested live endpoint is `kind: :current`:
-
-```ruby
-steps.last.to_boundary.to_h
-# => {
-#   kind: :current,
-#   version_id: nil,
-#   item_type: "Article",
-#   item_id: 42,
-#   recorded_at: 2026-08-08 12:00:00 UTC
-# }
-```
-
-A historical boundary may belong to the root or a selected descendant model.
-PaperTrail versions describe pre-change state, so a mutation becomes visible
-between its version boundary and the next historical or explicit current
-boundary. Passing `to: article` is what removes the need to touch the parent
-after an ordinary versioned child mutation; current state is still never
-implicit.
-
-### Reporting one save as one step
-
-A parent and its children saved together produce a version each, so one
-deliberate action arrives as several steps. `group: :transaction` reports it as
-one:
-
-```ruby
-PaperTrailDiff.activity_timeline(
-  article, from: :first, to: article,
-  associations: [:comments], group: :transaction
-)
-```
-
-Given a single transaction that retitled an article, revised one comment, and
-added another, the ungrouped timeline reports three steps — the title change,
-then a step that reads as empty, then the comment changes. Grouped, it reports
-one step carrying the title *and* the comment changes, credited to whoever made
-the save. Off by default; also accepted by `analyze(activity: true)`.
-
-The middle step reads as empty for a reason worth knowing, because it is the
-same rule the rest of this document turns on. A version records the state
-*before* its own event, so a child's new value is revealed only by something
-later: a further version of that child, its destroy version, or the live row.
-Inside a transaction there is usually none of those yet, so a change made at
-that boundary surfaces a step or more afterwards, folded in with whatever that
-step carried. Nothing is lost — but "when" and "with what" both read wrong.
-
-Two properties this relies on:
-
-**A step belongs to the transaction of the event that opens it.** That is the
-event whose change the step reports, so grouping on the closing boundary instead
-would credit each change to whoever made the next one.
-
-**Merging compares the group's outer states**, rather than combining the diffs
-between them. A field set and then restored inside one transaction has not
-changed, and only comparing the endpoints can say so.
-
-A boundary that records no transaction groups with nothing. PaperTrail leaves
-`transaction_id` nil outside a transaction, and a custom version class need not
-carry the column at all; treating those as one shared transaction would merge
-unrelated history into a single step. `ActivityBoundary#transaction_id` exposes
-what a boundary recorded.
-
-### Reading the state behind a step
-
-A step's diff carries what changed. A renderer often needs what did *not* — to
-say whose comment was edited, it needs the comment's author, which the diff has
-no reason to mention. `snapshots: true` retains the reconstructed states each
-step was compared between:
-
-```ruby
-steps = PaperTrailDiff.activity_timeline(
-  article, from: :first, to: article, associations: [:comments], snapshots: true
-)
-
-step = steps.reject(&:empty?).last
-step.from_snapshot.associations["comments"].records.first.attributes["author"]
-```
-
-The gem already builds these states to compute each diff and otherwise discards
-them, so asking for them costs no extra queries. Reconstructing them yourself
-from the version table means reimplementing boundary ordering, and getting it
-wrong is quiet rather than loud.
-
-They are off by default because each one holds the whole selected graph, and
-retaining one per step keeps the entire timeline's graph alive. A step whose
-`from_boundary` is a `create` version has a `nil` `from_snapshot`: the record
-did not exist yet, which is the answer rather than a missing one.
-
-`analyze(activity: true, snapshots: true)` populates the same fields on the
-`activity_timeline` it returns.
-
-### Closing a destroyed root
-
-A `destroy` version is the one boundary whose following state needs no later
-version: the event itself says the record is gone. When an activity timeline
-ends at the root's own `destroy` version, it therefore closes with one more
-step, from that version to a `kind: :destroyed` boundary, whose diff is a
-`record_presence_change` from the record's final state to `nil`:
-
-```ruby
-steps = PaperTrailDiff.activity_timeline(
-  article,
-  from: article.versions.first,
-  to: article.versions.last # a destroy version
-)
-
-removal = steps.last
-removal.to_boundary.destroyed?                  # => true
-removal.to_boundary.kind                        # => :destroyed
-removal.diff.record_presence_change.from        # the state it was deleted in
-removal.diff.record_presence_change.to          # => nil
-```
-
-The removal step's `from_boundary` is the ordinary `kind: :version` boundary
-for the same destroy version, since that boundary still holds the record.
-Boundaries therefore have three kinds — `:version`, `:current`, and
-`:destroyed` — so a consumer that branches on `version?` alone should also
-handle `destroyed?`.
-
-`analyze(activity: true)` reports the same closing step in its
-`activity_timeline`. Its `diff` and `timeline` keep their `compare` and
-`timeline` semantics and do not report the deletion.
-
-A `within:` window behaves the same way. A window whose last selected mutation
-is the root's destruction needs no later root version, because none can ever
-exist, so it closes on the removal instead of raising
-`IncompleteTimeRangeError`. When the destruction falls *outside* the window it
-remains ordinary reconstruction context and is not reported as a selected
-mutation. The same relaxation lets the plain `timeline` accept such a window,
-though it still reports only the edits.
-
-Live-ended HABTM activity is rejected with
-`PaperTrailDiff::UnsupportedLiveActivityError`. HABTM join rows have no model
-versions, so the gem cannot reliably split their membership mutations into
-intermediate events without transaction-backed owner checkpoints. Historical
-HABTM activity and `compare(version, article, associations: [:tags])` remain
-supported where the historical endpoint has usable PT-AT metadata.
-
-Both timeline APIs preserve empty boundaries. A display may use
-`steps.reject(&:empty?)`, while audit-oriented callers can retain every
-recorded boundary.
-
-## Select mutations by time
-
-`timeline`, `activity_timeline`, and `analyze` also accept a finite time range
-through `within:` instead of explicit `from:` and `to:` versions. Half-open
-ranges are recommended for adjacent reporting windows:
-
-```ruby
-window = Time.zone.parse("2026-08-01")...Time.zone.parse("2026-09-01")
-
-steps = PaperTrailDiff.timeline(article, within: window)
-
-activity_steps = PaperTrailDiff.activity_timeline(
-  article,
-  within: window,
-  associations: ["comments.replies", :author]
-)
-
-analysis = PaperTrailDiff.analyze(
-  article,
-  within: window,
-  associations: ["comments.replies"],
-  activity: true
-)
-```
-
-The range selects mutations by the timestamp of their source PaperTrail
-version. Ruby's inclusive (`..`) and exclusive (`...`) end semantics are
-honored. A returned step's `to_boundary` may be after the range: because a
-PaperTrail version is a pre-change snapshot, the gem needs one later root
-version to reveal the final selected mutation. That version is reconstruction
-context, not an additional selected mutation.
-
-If the window contains a relevant mutation but no later root version exists,
-the call raises `PaperTrailDiff::IncompleteTimeRangeError`. The gem does not
-silently substitute current database state; pass
-[`close_on: :current`](#reporting-up-to-now) to ask for it, or create a root
-checkpoint after the reporting window. A root-only window with no selected
-mutation returns a frozen empty timeline. The one exception is a window that
-closes on the root's own destruction, which no later version can ever follow;
-see [closing a destroyed root](#closing-a-destroyed-root).
-
-### Reporting up to now
-
-A window whose end is the present has nothing recorded after its final
-mutation, so by default it raises. `close_on: :current` closes it on the live
-record instead:
-
-```ruby
-PaperTrailDiff.analyze_many(
-  articles, within: month_start..Time.current, close_on: :current
-)
-```
-
-It is accepted by `timeline`, `activity_timeline`, `analyze`, and
-`analyze_many`, and only alongside `within:` — a range whose endpoints you gave
-explicitly already says where it ends. The closing step's `to_boundary.kind` is
-`:current` and its `to_version` is `nil`, so read the boundary rather than the
-version on the last step. Under `activity: true` the window runs to the instant
-state is captured rather than to the last root version, so a descendant that
-moved after that version is still reported.
-
-Two things follow from closing on live state. Current state includes changes
-PaperTrail never recorded — an `update_columns` write shows up in that final
-step, attributed to the version that precedes it, because nothing else records
-who made it. And a destroyed root has no current state to close on, so it keeps
-closing on its own destruction as it already did.
-
-Time ranges and explicit `from:`/`to:` endpoints are mutually exclusive.
-Malformed, open-ended, or reversed ranges raise
-`PaperTrailDiff::InvalidTimeRangeError`. With `activity_timeline`, selected
-descendant versions are included even when no root mutation occurred inside
-the window; the later root checkpoint is still required so PT-AT can
-reconstruct the enclosing graph and determine whether descendant activity is
-present. The ordinary `timeline` remains a root-only checkpoint timeline.
-
-## Build an endpoint and timeline together
-
-When a caller needs the net endpoint diff and root timeline, `analyze`
-normalizes each selected version once. Pass `activity: true` when the same view
-also needs descendant-aware activity. The endpoint diff and root timeline are
-then derived from the same activity snapshot pass instead of reconstructing the
-root history separately:
-
-```ruby
-analysis = PaperTrailDiff.analyze(
-  article,
-  from: article.versions[1],
-  to: article.versions[4],
-  associations: ["comments.replies"],
-  activity: true
-)
-
-analysis.diff
-analysis.timeline
-analysis.activity_timeline
-```
-
-Without `activity: true`, `analysis.activity_timeline` is `nil` and no activity
-work is performed. `analyze` accepts explicit historical versions or
-`within:`, but not a current-record endpoint; use the standalone
-`activity_timeline(..., to: article)` API when the final boundary must be live.
-
-### Choosing between `timeline` and `analyze(activity: true)`
-
-`analysis.timeline` and `timeline` return the same root-checkpoint steps for the
-same range. They are two reconstruction strategies for one result, not two
-levels of detail: a `timeline` step covers only root version boundaries, but
-each of its snapshots still contains every selected association, because a step
-must report what changed underneath the root as well.
-
-The one behavioural difference is at the edge of a `within:` window. Selected
-descendants can move inside a window that contains no root version at all, so
-the activity form requires a root boundary it can reconstruct from and raises
-`IncompleteTimeRangeError` when there is none. `timeline` has no activity view
-to anchor and returns no steps for that window.
-
-The strategies differ in what that costs:
-
-- `timeline` reconstructs the whole selected graph independently at every root
-  boundary, so it costs roughly *root versions x selected graph size*. It is
-  insensitive to how much descendant activity happened in between.
-- `analyze(activity: true)` reconstructs once and then advances that snapshot
-  through each recorded mutation, so it costs roughly *one reconstruction +
-  total events*. It is insensitive to how wide the selected graph is.
-
-Neither dominates. Reconstructing once per checkpoint wins when a few root
-versions span very heavy descendant churn; advancing incrementally wins when
-the selected graph is wide and descendant activity is comparable to root
-activity. As a rule of thumb, prefer `analyze(activity: true)` when selected
-associations are wide, and `timeline` when descendant events greatly outnumber
-root versions. Measure with `ActiveSupport::Notifications` on a representative
-history rather than a seeded example if the choice matters.
-
-`timeline`, `activity_timeline`, and both forms of `analyze` prepare the selected
-historical range once. The loader walks only the explicit association paths and
-builds an immutable temporal index of scalar states, relationship candidates,
-HABTM transaction snapshots, and live fallbacks. Direct relationships,
-`has_many :through` with a `belongs_to` source, and transaction-backed HABTM
-membership are resolved from that index. Endpoint-only `compare` retains the
-lighter two-point reconstruction path.
-
-Activity reconstruction then carries immutable snapshots forward between
-boundaries. Isolated root and descendant events with usable serialized changes
-advance only the affected immutable nodes. Events sharing a PT-AT transaction
-refresh their combined branches atomically. Scoped associations, unversioned
-targets, composite relationship keys, collection-source through associations,
-and other unsupported shapes fall back to the ordinary PT-AT point reifier on
-a per-reflection basis. This hybrid path preserves existing reconstruction
-behavior while avoiding repeated association queries for supported paths.
-
-For supported collection events, adjacent snapshots share an immutable
-identity-position index and carry a one-step transition hint. Diffing therefore
-visits only the changed member instead of hashing the whole collection again.
-Prepared scalar history also exposes the predecessor and successor attributes
-for isolated update events, allowing activity reconstruction to update the
-immutable snapshot directly instead of reifying and deserializing the same
-PaperTrail event again. Compatible scalar payloads are decoded into prepared
-state without first constructing a disposable Active Record object. Encrypted,
-schema-mismatched, missing, ambiguous, or identity-changing states fall back to
-the ordinary event reifier.
-For a direct nested `has_many` such as `comments.replies`, the child's foreign
-key also locates its parent snapshot directly rather than walking every
-comment. Membership changes and ambiguous or unsupported relationship shapes
-retain the general comparator and traversal fallback.
-
-Activity event loading is bounded to the selected range. Association identity
-discovery retains one indexed checkpoint for members present at the starting
-boundary, then considers later association activity and current members; it no
-longer materializes every pre-range association row. Because a PaperTrail
-version is a pre-change snapshot, the state at a range's final boundary can
-live only in the next version after it, so prepared scalar state retains one
-trailing version per selected identity. It does not retain the rest of the
-history recorded after the range, so a short range early in a long history
-costs the same as the same range in a short one. Keep requested paths and
-ranges intentional. An activity timeline must also emit a diff for every selected
-event. Repeated events within one wide collection still copy the frozen records
-array when producing each immutable snapshot, so they can do pointer-copying
-work proportional to the number of events times the collection width even when
-SQL and Ruby-level comparison work stay linear. Bound or paginate unusually
-wide activity ranges in latency-sensitive requests.
-
-For large histories, applications should give the database a matching
-composite index. A typical PaperTrail installation can add one without making
-it a requirement of this gem:
-
-```ruby
-add_index :versions, %i[item_type item_id created_at id],
-          name: "index_versions_on_item_and_timeline"
-```
-
-This index is especially useful for repeated `within:` queries because root
-selection is constrained by `item_type`, `item_id`, and `created_at`, then
-ordered deterministically by timestamp and version ID.
-
-PT-AT's normal index beginning with `foreign_key_name`, `foreign_key_id`, and
-`foreign_type` should also be retained on `version_associations`.
-
-## Ignore noise fields
-
-The comparison APIs ignore `updated_at` by default. `ignore:` replaces that default and
-accepts string or symbol names. The array form applies to the root and every
-selected association:
-
-```ruby
-PaperTrailDiff.compare(from, to, ignore: %i[updated_at lock_version])
-PaperTrailDiff.compare(from, to, ignore: []) # compare every scalar field
-```
-
-For exact path control, pass `all:` plus `paths:`. `$` identifies the root:
-
-```ruby
-PaperTrailDiff.compare(
-  from,
-  to,
-  associations: ["comments.replies"],
-  ignore: {
-    all: [:updated_at],
-    paths: {
-      "$" => [:lock_version],
-      "comments.replies" => [:delivery_state]
-    }
+feed = steps.reject(&:empty?).map do |step|
+  {
+    actor: step.source_boundary.whodunnit,
+    recorded_at: step.source_boundary.recorded_at,
+    changes: step.diff.to_h
   }
-)
-```
-
-Hash rules replace the default just like an array. `all:` applies everywhere;
-each path entry applies only at that exact path, not its descendants. Ignore
-paths must be `$` or one of the selected or implicitly selected association
-paths.
-
-Primary keys are always represented as record identity rather than scalar
-attributes.
-
-## Compare associations
-
-Pass explicit names or dot-separated paths. Paths are finite traversal plans,
-not a request to recursively inspect the whole object graph. Ancestors are
-selected implicitly:
-
-```ruby
-diff = PaperTrailDiff.compare(
-  from,
-  to,
-  associations: [:author, :profile, "comments.replies.author"]
-)
-
-diff.associations["author"].relationship
-# => #<PaperTrailDiff::ValueChange ...>  # belongs_to replacement
-
-comments = diff.associations["comments"]
-comments.added   # full RecordSnapshot objects
-comments.removed # full RecordSnapshot objects
-comments.changed # RecordChange objects for stable identities
-
-replies = comments.changed.first.associations["replies"]
-replies.added
-replies.removed
-replies.changed
-```
-
-For `belongs_to` and `has_one`, replacing the related identity is a
-`relationship` change. Updating attributes on the same related identity is a
-`changed` record. For `has_many` and `has_and_belongs_to_many`, membership is
-split into `added` and `removed`; shared identities with scalar changes appear
-in `changed`. The result preserves the reflected macro in `kind`.
-
-Both Rails many-to-many forms are supported. `has_many :through` uses PT-AT's
-history for the versioned join model. HABTM reports membership and target
-attribute changes, but it cannot report join attributes because HABTM has no
-join model.
-
-The same structure repeats at every selected depth. `RecordChange#associations`
-contains nested changes for a stable parent identity. Added and removed record
-snapshots retain their selected subtrees so nested state is not discarded.
-
-Selecting a `belongs_to` removes its foreign-key (and polymorphic type) column
-from scalar changes at that path. Without association selection, that column
-remains a normal scalar attribute. A direct `has_one` or `has_many` edge also
-removes its incoming foreign key from the child snapshot because membership is
-already represented structurally. Join-model attributes for `has_many :through`
-remain available. Cyclic model relationships are safe because only the finite
-paths supplied by the caller are traversed.
-
-Requesting associations for any historical endpoint without loaded and enabled
-PT-AT raises `PaperTrailDiff::AssociationTrackingUnavailableError`. A comparison
-whose endpoints are both live records can normalize associations without PT-AT,
-though it will normally be empty because both records reload the same current
-database state. Unknown names and unsupported macros raise explicit
-`PaperTrailDiff::Error` subclasses.
-Malformed public options raise `PaperTrailDiff::ConfigurationError`, also under
-that base error.
-
-## Traverse result trees
-
-The nested tree remains the canonical, lossless result. For renderers,
-counters, exports, and notifications, `Diff#each_change` provides a
-deterministic depth-first stream of semantic changes:
-
-<!-- executable:readme-traversal-changes -->
-```ruby
-diff.each_change do |entry|
-  entry.kind              # :attribute_changed, :record_added, ...
-  entry.association_path  # ["comments", "replies"]
-  entry.record_path       # frozen RecordReference objects
-  entry.association_kind  # :has_many, :belongs_to, ...
-  entry.attribute         # "body" for an attribute entry
-  entry.value             # ValueChange, RecordChange, or RecordSnapshot
 end
 
-counts = diff.each_change.map(&:kind).tally
-```
-
-`record_changed` entries are emitted before that record's attribute and nested
-association changes, allowing an application to count both changed records and
-changed fields. Singular relationship operations are classified separately as
-`relationship_added`, `relationship_removed`, or `relationship_replaced`.
-Root create/delete transitions remain `record_presence_changed`.
-
-Added, removed, and replaced records carry complete bounded snapshots. Use
-`each_entry` when a renderer also needs that nested state without writing a
-second snapshot walker:
-
-<!-- executable:readme-traversal-entries -->
-```ruby
-diff.each_entry do |entry|
-  next unless entry.included_state?
-
-  entry.kind     # :record_included, :attribute_included, :association_included
-  entry.state    # :before or :after
-  entry.context  # :included_state
-end
+payload = steps.map { |step| step.to_h(metadata: true) }
 ```
 
-An included record is historical context, not evidence that the nested record
-changed independently. `each_change` therefore omits included-state entries.
-Both methods return an `Enumerator` when no block is supplied and return the
-diff when called with a block. Entries and their paths are immutable and have
-deterministic `to_h` output. The root location uses empty association and record
-paths; descendant `record_path` values contain only the explicitly traversed
-descendant identities.
+Each step contains a `diff`. `source_boundary` describes the PaperTrail version
+responsible for that change, including its `whodunnit`, `event`, `item_type`,
+and `item_id`. Its `recorded_at` is the version's `created_at`.
+Use this source information for attribution: the version used to read the
+resulting state, exposed as `to_boundary`, may have been saved by someone else.
 
-Traversal is path-preserving and does not deduplicate. If the same record is
-reachable through two selected paths, it appears at both locations so the
-consumer can choose whether path identity or record identity controls counting.
-Walking a result never performs database access and never discovers additional
-associations.
+`group: :transaction` combines changes from one database transaction. Its source
+information comes from the first event in that group. Omit `group:` if you need
+to attribute each version separately.
 
-## Discover and diagnose associations
+`snapshots: true` also keeps copies of the compared records' data, including
+unchanged fields. These read-only copies are called snapshots. Access them as
+`step.from_snapshot` and `step.to_snapshot`, or include them in a hash with
+`step.to_h(snapshots: true)`. Keeping them uses more memory. Omit the option if
+changed values are enough. `metadata: true` adds event and author information
+to the hash.
 
-Configuration UIs can use bounded public reflection instead of duplicating the
-gem's macro and cycle logic:
+## 3. Report on records changed since a given time
 
-```ruby
-PaperTrailDiff.supported_association_macros
-PaperTrailDiff.association_paths(Article, max_depth: 3).map(&:to_h)
-```
-
-Descriptors include `path`, `kind`, `target_type`, `through`, `polymorphic`, and
-`cycle`. Cycles are marked and not descended; PaperTrail's infrastructure
-`versions` association is excluded.
-
-Before relying on association history, inspect known setup hazards:
+Use `report` with a model or ActiveRecord relation. It includes activity from
+`since:` through the current saved state. Association tracking is only needed
+if you also pass `associations:`.
 
+<!-- executable:recipe-report -->
 ```ruby
-report = PaperTrailDiff.diagnose(
-  from_version,
-  to_version,
-  associations: [:tags, "comments.replies"]
+report = PaperTrailDiff.report(
+  Article.all, since: 1.day.ago, limit: 100, include_deleted: true
 )
 
-report.ok?
-report.errors.map(&:code)
-report.warnings.map(&:code)
-```
-
-`ok?` means no *errors* were found among the checks that ran, not that every
-possible hazard was ruled out. Association checks need `associations:` to have
-something to inspect; without it only the version-order check runs, which is
-why that one runs whether or not associations are selected — an unorderable
-history corrupts a scalar timeline just as thoroughly.
-
-Diagnostics are read-only guidance, not proof that arbitrary old data is
-complete. HABTM endpoints without transaction-backed association snapshots fail
-loudly with `PaperTrailDiff::IncompleteAssociationHistoryError` during normal
-comparison as well.
-
-### Recommended checkpoint recipe
-
-For PT-AT history—especially HABTM—use transaction-backed checkpoints, disable
-PaperTrail's version timestamp synchronization on the root, and checkpoint a
-fresh record instance:
-
-```ruby
-class Article < ApplicationRecord
-  has_paper_trail synchronize_version_creation_timestamp: false
+rows = report.analyses.map do |identity, analysis|
+  {
+    record: identity,                 # ["Article", "42"]: model name and ID
+    changes: analysis.diff.to_h,
+    activity: analysis.activity_timeline.reject(&:empty?).map do |step|
+      step.to_h(metadata: true)
+    end
+  }
 end
 
-def checkpoint_article(article_id)
-  Article.transaction do
-    Article.find(article_id).paper_trail.save_with_version
-  end
-end
+report.unreachable # Model names and IDs for records no longer in the database
+report.deleted     # Reports for those records, or nil when destroy history is missing
 ```
 
-A root checkpoint after a batch supplies an immutable historical final state.
-The ordinary `timeline` aggregates that batch; `activity_timeline` uses recorded
-descendant versions to split it into activity boundaries. For an interactive
-view, `activity_timeline(..., to: article)` can instead terminate explicitly at
-current state without touching the root. Prefer checkpoints for reproducible
-audits and HABTM, and prefer an explicit join model with `has_many :through`
-when join attributes or join mutations must appear as first-class history.
-
-## Result objects
-
-Every level of a result separates the same two kinds of change in the same way:
-
-| | a record appears or disappears | a record stays and its fields change |
-| --- | --- | --- |
-| the root record | `record_presence_change` | `attributes`, `associations` |
-| `has_many`, HABTM | `added`, `removed` | `changed` |
-| `belongs_to`, `has_one` | `relationship` | `changed` |
-
-The left column carries whole `RecordSnapshot` values; the right column carries
-field-level `ValueChange` deltas. That split is deliberate. A record that has
-just appeared has no previous value for any of its fields, so reporting one
-`nil` to value change per attribute would both add noise and blur the
-difference between "this field was edited" and "this record did not exist".
-
-The consequence is that reading a result tree directly means branching on which
-column applies: a created record's state is under
-`record_presence_change.to.attributes`, an edited record's is under
-`attributes`. Consumers that would rather not branch should use
-[`each_entry`](#traverse-result-trees), which flattens both into one stream —
-an edit arrives as `attribute_changed` carrying a `ValueChange`, and a created
-record's fields arrive as `attribute_included` entries carrying values with
-`state: :after`.
-
-The public result types are:
-
-- `PaperTrailDiff::Diff`
-- `PaperTrailDiff::Step`
-- `PaperTrailDiff::ActivityStep`
-- `PaperTrailDiff::ActivityBoundary`
-- `PaperTrailDiff::Analysis`
-- `PaperTrailDiff::ValueChange`
-- `PaperTrailDiff::TraversalEntry`
-- `PaperTrailDiff::RecordReference`
-- `PaperTrailDiff::RecordSnapshot`
-- `PaperTrailDiff::RecordChange`
-- `PaperTrailDiff::ToOneAssociationDiff`
-- `PaperTrailDiff::CollectionAssociationDiff`
-- `PaperTrailDiff::AssociationDescriptor`
-- `PaperTrailDiff::DiagnosticReport`
-- `PaperTrailDiff::DiagnosticIssue`
+`report.analyses` is a hash of reports for records that still exist. Each key is
+`[model_name, id_as_a_string]`. Each value has the net `diff` and an
+`activity_timeline`. `report.roots` returns the selected ActiveRecord objects.
 
-`Analysis` also exposes the reconstructed states its diff was taken between, as
-`from_snapshot` and `to_snapshot`. A report that renders unchanged columns needs
-the whole final state, not only what moved, and these are the states the gem
-already reconstructed:
+`include_deleted: true` also builds reports from PaperTrail's destroy versions.
+Read them from `report.deleted` using the same keys:
 
+<!-- executable:recipe-deleted -->
 ```ruby
-analysis.to_snapshot.attributes    # every selected scalar, changed or not
-analysis.to_snapshot.associations  # the selected association tree
+unavailable = report.deleted.select { |_identity, analysis| analysis.nil? }.keys
+deleted_activity = report.deleted.transform_values do |analysis|
+  analysis&.activity_timeline
+end
 ```
 
-Either is `nil` when that endpoint has no reconstructable state — most often a
-`from_snapshot` at a `create` boundary, whose pre-change state is the absence of
-the record. Following the precedent set by `Step`, `Analysis#to_h` is unchanged;
-serialize `to_snapshot.to_h` when a serialized form is wanted.
-
-They expose readers, are frozen after construction, and provide deterministic
-`to_h` output. Collection results are ordered by record identity: by type, then
-naturally within one id type, so numeric ids sort `2` before `10`. Mixed or
-unusual id types still order totally rather than raising.
-
-Structural hash keys are symbols; attribute and association
-names are strings. Attribute values retain their Ruby types. `RecordChange#record`
-is a `RecordReference` with `type` and `id` readers. `TraversalEntry#record` and
-`#association` return the final components of their corresponding paths. `Step`
-itself is frozen but intentionally retains the original, potentially mutable
-PaperTrail version objects for metadata access; its new boundary readers provide
-an immutable presentation interface, while `Step#to_h` continues to emit only
-version IDs. `ActivityStep` retains only immutable boundary metadata and
-serializes as `{ from: ..., to: ..., diff: ... }`.
+The results are already computed; you do not need to query or reify each deleted
+record. Omit `include_deleted:` if you only need records that still exist.
 
-## Historical correctness and limitations
+Three selection rules matter:
 
-The output is only as correct and complete as the historical state PaperTrail
-and PT-AT can reconstruct. In particular:
+- The record itself needs a PaperTrail version in the time window. For example,
+  editing only a comment does not select its article. Comment activity is included
+  for selected articles when you request `associations: [:comments]`.
+- A relation such as `Article.where(status: "open")` checks today's values.
+  It cannot establish whether deleted articles matched that condition. Deleted
+  records are therefore listed separately.
+- `limit: 100` is a safety ceiling, not pagination. Selecting too many records
+  raises `BatchLimitExceededError`. With `include_deleted: true`, both existing
+  and deleted records count toward it. Use a narrower relation or time window,
+  or deliberately raise the ceiling.
 
-- attributes skipped by PaperTrail, deleted versions, and changes made while
-  versioning was disabled cannot be recovered;
-- model/schema or serializer changes can affect reification of old versions;
-- comparison and activity APIs add live state only when the caller explicitly
-  passes a persisted record endpoint; the adapter reloads that record, but
-  multiple queries are not automatically one database-isolated snapshot;
-- PT-AT requires its schema, callbacks, versioned child models, and transaction
-  metadata; callback-skipping writes may not be reconstructable;
-- HABTM membership is limited to the join snapshots PT-AT recorded in
-  `version_associations`; historical target attributes require versioned target
-  models, otherwise PT-AT may return live target state;
-- `compare`, `timeline`, and `analyze`'s endpoint diff do not report the root
-  record's destruction, because the state recorded at a `destroy` version is the
-  state before the deletion; `activity_timeline` closes on a `:destroyed`
-  boundary instead, and a selected child's removal is reported by its parent;
-- `timeline` and `analyze` are bounded by root versions; `activity_timeline`
-  adds recorded descendant boundaries and may terminate at an explicitly passed
-  current record, while a fully historical result still requires a later root
-  version because PaperTrail stores pre-change snapshots;
-- HABTM join-table mutations do not have their own model versions and therefore
-  cannot become standalone activity boundaries; live-ended HABTM activity is
-  rejected, and a versioned join model should be used when that activity matters;
-- activity discovery uses PT-AT association rows and cannot recover descendant
-  events whose relevant callbacks or association metadata were never recorded;
-- PT-AT has documented edge cases, especially around some `has_one` histories;
-  `paper_trail_diff` respects PT-AT's configured reification error behavior;
-- deeper paths require more historical reconstruction and database work, so
-  callers should select only the branches they need;
-- no implicit or unbounded recursive association traversal is performed.
+Use the relation's `order`, `offset`, and `limit` to choose a page, for example
+`Article.order(:id).limit(20)`. Joined rows can repeat a record; each record is
+analyzed once. See the [reporting reference](docs/reference.md#which-records-are-selected)
+for pagination, deleted records, and selecting by past values with
+`historical_filter:`.
 
-Review the [PaperTrail version semantics](https://github.com/paper-trail-gem/paper_trail#3-working-with-versions)
-and [PT-AT limitations](https://github.com/westonganger/paper_trail-association_tracking#limitations)
-before relying on reconstructed history for restoration or compliance work.
+For a fixed past window, use `analyze_scope(Article.all, within: start_time...end_time,
+limit: 100)`.
 
-## Development
+## Choose a method
 
-The project uses mise for its pinned development tools:
-
-```console
-mise install
-mise exec -- bundle install
-mise exec -- bundle exec rake
-mise exec -- bundle exec rake build
-mise exec -- act push -j quality --matrix ruby:4.0 --matrix paper_trail:17
-```
-
-The default Rake task runs core specs without PT-AT loaded, association specs in
-a separate process, RuboCop, generated-signature verification, and Steep.
-Selected Ruby blocks in this README and the quickstart are executed in those
-same isolated test sessions; the `<!-- executable:... -->` marker opts a block
-into the appropriate stateful session.
-Inline `#:` annotations in `lib/` generate the committed RBS files under
-`sig/generated/`.
-
-Release tags matching `v*` publish through RubyGems trusted publishing. Before
-tagging, replace `Unreleased` in the changelog with the release date, run the
-full CI workflow, and configure the repository's `release` environment as a
-trusted publisher on RubyGems. Do not store a long-lived RubyGems API key in
-the repository.
-
-## License
-
-MIT
+| What you need | Method |
+| --- | --- |
+| Net difference between two versions, or a version and the current record | `compare(from, to)` |
+| The same for a list of `{ from:, to: }` pairs, one per record | `compare_many(pairs)` |
+| Each change to one record, version by version | `timeline(record)` |
+| Each change to one record and its chosen associations, with who made it | `activity_timeline(record, associations: ...)` |
+| Net difference and timeline together | `analyze(record)` |
+| The same, plus the activity timeline | `analyze(record, activity: true)` |
+| Analyses for an array of records | `analyze_many(records)` |
+| Analyses for the records a relation selects, over a time window | `analyze_scope(scope, within:, limit:)` |
+| Recent activity through now, for the records a relation selects | `report(scope, since:, limit:)` |
+
+`timeline`, `activity_timeline`, and `analyze` cover the whole history through
+the current saved record by default. Narrow them with `from:` and `to:`
+(versions) or `within:` (a time range). See
+[choosing a range](docs/reference.md#choosing-a-range).
+
+## Performance and defaults
+
+Use the batch methods for reports instead of calling `reify` or a single-record
+method inside a loop. They share history loading across records. Ordinary
+record and direct-association histories can be read without creating a model
+instance for every historical state.
+
+Some associations still need per-record queries. Batch methods reject scoped
+associations and certain `has_many :through` histories by default with
+`UnsupportedBatchPathError`. The error names the association. Remove it from
+`associations:`, or pass `allow_expensive: true` if you accept that cost.
+Callbacks and callable historical filters can also issue queries. See the
+[performance guide](docs/performance.md) for measured behavior and limitations.
+
+- `updated_at` is ignored. Passing `ignore:` replaces the default list.
+- Records passed as current state must be saved and have no unsaved changes.
+  They are reloaded from the database.
+- Empty timeline steps are kept. Use `reject(&:empty?)` to hide them in a feed.
+- Only the associations you name are included.
+- Historical data cannot be recovered if PaperTrail did not record it.
+
+See the [API reference](docs/reference.md) for all options and result methods,
+[testing](docs/testing.md) for development checks, and
+[CHANGELOG.md](CHANGELOG.md) for changes. Licensed under the [MIT license](LICENSE).