V49, plain components: 100 at familiarity 5
V48 reached 100 at familiarity 4. V49 asks whether the most familiar LiveView shape can get there too. That shape is the one phx.gen.live hints at: LiveComponents that hold their own state and tell their parent what happened, and a page that reacts in handle_info.
It starts from V41-max, which already had that shape but wired it through a route table with nine target kinds. Since 2026-10-01 the checklist reads wiring written as function clauses in one module as meeting R8, so the table was never required. V49 removes it.
The page is plain LiveView
Each port a panel sends gets one handle_info clause, and every call in it is LiveView’s own:
def handle_info({:cart, :removed, item}, socket) do
send_update(Undo.Banner, id: "undo", capture: item)
{:noreply, socket}
end
def handle_info({:undo, :restored, item}, socket) do
send_update(Cart.Panel, id: "cart", receive: item)
{:noreply, put_flash(socket, :info, "Item restored")}
end
The route table’s harder kinds existed so the page wouldn’t take one abstraction’s answer and hand it to another, which is Spray’s “handling the data” (§1.6.3). feed (ask the store, pass the answer to a panel) was one of them. V49 uses LiveView’s own task for it. The page starts the store work, and handle_async sends the answer on:
def handle_info({:wishlist, :taken, product}, socket) do
add_line = socket.assigns.instances.add_line
{:noreply, socket |> start_async(:add_line, fn -> AddLine.run(add_line, product) end) |> put_flash(:info, "Added to cart")}
end
def handle_async(:add_line, {:ok, line}, socket) do
send_update(Cart.Panel, id: "cart", receive: line)
{:noreply, socket}
end
That changes behaviour slightly: the added line appears a moment after the click, not in the same render. The tests’ settle/1 helper now waits for the page’s tasks with render_async/1, and no assertion changed.
Three things the walk found
- Panels named themselves. Each panel sent
{:cart, port, payload}with:cartwritten in its own code, and the page had to match it. Two modules agreeing on a literal is a silent contract (R5). V41 and V41-max both had it, and the earlier walks missed it. In V49 the page passes the name:<.live_component module={Cart.Panel} id="cart" name={:cart} />. - The order panel built an output from its own assigns. Its Review button sent
{:order, :review, summary}straight from the panel’s state, so itsOrderLinesfeature had noreviewport at all. V49’sOrderLineshasrequest_reviewandreview, and the panel runs it. The new coverage test caught this. It checks both ways: every port a panel sends has a clause, and every clause names a port something sends. V41-max’s test only checked the first direction. FormComponentandFeatures. V49 has a genericRecordFormLiveComponent, a widget whose own state is the form being typed, configured with the record, a changeset function, the fields and the words. The page saves through a task and sendsinvalidback. The coded modules are namedState.
What it scores
| V41-max | V49 | |
|---|---|---|
| Full checklist, by hand (2026-10-02 readings) | 98, now 97 (the panel names) | 100 (99.5) |
ala_lint default / --strict / --super-strict | 100 / 100 / 99 | 100 / 100 / 99 |
Rules met under --strict (of 10 checked) | 9 | 10 |
| Message hops per cross-feature effect | 1 | 1 |
| Tests | 85 | 86 |
| Mean / median time per test | 6.4 / 0.7 ms | 8.2 / 1.0 ms |
Non-blank lines in lib/ | 4,317 | 4,414 |
| Familiarity | 4 | 5 |
The half point is the hop: a cross-feature effect goes from a panel to the page and back to another panel, and the receiving panel stays mounted behind a hidden tab to receive it. It rounds to 100, as V41-max’s did.
One judgement call holds the score up, and it’s the one the checklist records as held loosely. The panels render their own markup. I count them as stateful domain UI abstractions, because they take their words, rules and data from the page. A reviewer who reads their markup as shaped for one page would call them contained sub-components (§2.2), and V49 would lose a point or two.
What it shows
With the route table gone, V49’s page uses nothing a Phoenix developer hasn’t seen. The rest of the vocabulary is V41’s: {state, outputs} feature modules, and an Instance module of two functions (run a step, send an output to the parent). I rated V41 a 5 with that vocabulary, so V49 is a 5 by the same judgement. That’s my reading, not a measurement.
So the strict rules don’t stop a familiar 100. The price at familiarity 5 is a message hop, an asynchronous add-to-cart, and a judgement call about panel markup. At familiarity 4 (V48) there’s no hop and no judgement call.
The cart page, drawn from its clauses
Paradigms.Drawing.mermaid/1 reads the page’s handle_info and handle_async clauses. A send_update is a solid arrow to the panel whose input it calls, labelled port → input, and a start_async is an arrow to a task, whose outcome clause carries on from it. Any other call is a dotted arrow to the page effect it lands on. As in V41-max, what each panel wires to itself isn’t drawn.
flowchart LR
cart -.->|"summary"| page_assign["page: assign"]
cart -.->|"changed"| call_apply_change["call Carts.apply_change"]
cart -->|"removed → capture"| undo
cart -->|"saved → stash"| saved
cart -.->|"saved"| page_put_flash["page: put_flash"]
cart -->|"line → toggle"| wishlist
cart -.->|"promo_applied"| page_put_flash["page: put_flash"]
cart -.->|"promo_rejected"| page_put_flash["page: put_flash"]
undo -->|"expire → expire"| undo
undo -->|"expired → confirm_removal"| cart
undo -->|"restored → receive"| cart
undo -.->|"restored"| page_put_flash["page: put_flash"]
saved -.->|"count"| page_assign["page: assign"]
saved -->|"moved → receive"| cart
saved -.->|"moved"| page_put_flash["page: put_flash"]
wishlist -.->|"count"| page_assign["page: assign"]
wishlist -.->|"ids"| page_assign["page: assign"]
wishlist -.->|"added"| page_put_flash["page: put_flash"]
wishlist -.->|"dropped"| page_put_flash["page: put_flash"]
wishlist -.->|"taken"| add_line["task AddLine.run"]
wishlist -.->|"taken"| page_put_flash["page: put_flash"]
checkout -.->|"blocked"| page_put_flash["page: put_flash"]
checkout -.->|"ready_to_pay"| payment["task StartPayment.call"]
checkout -.->|"done"| call_place["call PlaceOrder.place"]
checkout -.->|"done"| page_redirect["page: redirect"]
checkout -.->|"step"| page_push_patch["page: push_patch"]
stock_changed -->|"fact → set_stock"| cart
add_line -->|"ok → receive"| cart
payment -->|"ok → succeeded"| checkout
payment -->|"error → failed"| checkout
payment -->|"exit → failed"| checkout
Rescored 2026-10-03: 99
Spray’s §6.17.2 is now read strictly under R10. The small fix is in: checkout gets the cart’s id and stored lines, {cart_id, items}, not the cart, and owns the readiness rule. What remains is the shared entity: the portal’s order lines are built on the storefront’s cart, which carries promo, gift-wrap and shipping state, one use case’s data in a struct another use case shares (−1). The fix is to split the storefront’s concerns out of the struct the portal uses; no variant has done that yet. 99.5 becomes 98.5, which rounds to 99. Lint under --strict is 99, meeting 9 of 10 rules: the linter now scores the shared aggregate from --strict up.
Rescored 2026-10-03 (later): 100
The shared entity is gone. The portal’s OrderLines has its own struct and pricing (subtotal, volume tier, shipping), and builds, like the storefront’s cart, on a new Lines domain abstraction: plain functions over a list of stored order lines (find, add, change a quantity, remove, set stock), holding no prices, codes or flags of any use case. The cart lost set_discount/2, which its own doc said was “added for the portal consumer”. With checkout already getting only {cart_id, items}, R10 is met, and 98.5 becomes 99.5 (the hop), which rounds to 100. Lint under --strict is 100, meeting 10 of 10 rules. The linter needed one correction to read it: a story that only builds the cart (Cart.new) creates an instance, which a Spray feature does (§2.2), and doesn’t know its data, so it no longer counts as a reader.
The code is in ala_variants_elixir/v49-plain-components.