The hardest modal in the app: one wizard that both creates and edits invoices,
with a conditional middle step (basic-details → [accounts] → next-steps, where
the expense-accounts step is skipped on the default-accounts path). Migrated off
mm/* + form-cursor + the EDN snapshot onto the session-backed engine (wizard2).
Finding: the OLD basic-details "Save" was broken. It hx-puts /invoice/new/navigate,
whose `[:to {:optional true} …]` query-schema 500s on empty query-params — Ring's
wrap-params yields {} for a no-query PUT, and main-transformer's parse-empty-as-nil
decodes {} → nil, which the bare [:map] rejects. Production uses the identical
wrap-params, so it was broken there too. So e2e/invoice-new.spec.ts is an ACCEPTANCE
gate (red on the old code, green on the engine, whose submit is a POST with no
query-schema): the migration fixes a latent bug. Create semantics (default → vendor
default account, location-spread; customize → posted grid; edit → prefill + updated
row) were pinned at the REPL.
What changed:
- defrecord 4 → 0 (NewWizard2 / BasicDetailsStep / AccountsStep / NextSteps), mm/ 0,
fc/ cursor refs 0, step-params[…] field names 0.
- Conditional `:next` `(if (= :customize …) :accounts :done)` replaces mm/CustomNext +
the broken 308-to-submit. Dual-purpose new+edit = one :init-fn branching on a route
:db/id; create-wizard! seeds :init-data as per-step step-data so edit opens populated.
- The broken new-wizard-navigate route is deleted; the genuine async helpers
(account-prediction, due/scheduled-payment-date, location-select, expense total/balance,
add-row) remain but read the posted flat form (+ ws/get-all for the cross-step total).
- next-steps becomes the done-fn's returned modal (Pay now / Add another / Close).
- Dates ride as java.util.Date (#inst) in step-data so it's EDN-safe across the
non-terminal step (clj-time DateTimes break the cookie store).
Verification: full e2e suite 61/61 (58 prior + 3 new); maybe-spread-locations unit
test 6/6; create semantics + edit prefill confirmed at the REPL. Skill fed
(scorecard Phase 8, gotchas {}→nil 500 + #inst dates, form-vs-wizard conditional
:next + dual-purpose).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
12 KiB
Forms vs. wizards (and the data-driven wizard engine)
Classify first
| Signal | Classification |
|---|---|
One logical step — even with a ?mode= toggle, $/% radio, or add/remove rows |
plain form |
| The user genuinely advances through ordered steps, each validated before the next | wizard |
| In doubt | form |
Most "wizards" in this codebase are single-step forms wearing wizard costumes: they
implement the multi-step protocol (mm/ModalWizardStep + friends), serialize an EDN
snapshot into hidden fields, and register 10–20 stacked-middleware routes — all for one
step. That is pure overhead to delete.
Done — Transaction Edit is now a plain form.
LinksStep/EditWizardand allmm/*usage were deleted fromtransaction/edit.clj; the worked example below is realized, not aspirational. See "Single-step → plain form (realized)".
The machinery being replaced
The old shape (kept here as the "before"):
(defrecord LinksStep [linear-wizard]
mm/ModalWizardStep
(step-name [_] "Transaction Actions")
(step-key [_] :links)
(edit-path [_ _] [])
(step-schema [_] (mm/form-schema linear-wizard))
(render-step [this {{:keys [snapshot step-params]} :multi-form-state :as request}] ...))
…plus the snapshot round-trip: the whole accumulating form state is serialized to hidden
fields (custom EDN readers), then rebuilt every request by merging the posted pieces back
into the snapshot (:multi-form-state :snapshot is read ~75× in edit.clj). The
serialization needs custom readers, the merge is error-prone, and the payload grows each
step.
Single-step → plain form
Two routes: GET (render) and POST (validate + save). State is plain form fields + an
entity id. No snapshot, no server state, no protocol.
{::route/edit (fn [req] (html-response (render-edit-form {:entity (get-entity req)})))
::route/edit-submit (fn [req] (validate-and-save req))}
A ?mode= toggle is just the GET re-rendering with a different query param — still a
plain form. An add-row interaction is one extra POST that appends a fresh row and
re-renders (the +1 route).
Single-step → plain form (realized: Transaction Edit)
What replacing the wizard actually looked like, end to end:
- Delete the records + middleware.
EditWizard/LinksStep,mm/open-wizard-handler,mm/next-handler,mm/submit-handler,mm/wrap-wizard,mm/wrap-decode-multi-form-state, and theedit-wizard-navigateroute all go.render-stepbecomes a plainrender-form. - Rename the fields off
step-params[...]. Field names are now the schema path directly ((path->name2 :transaction/accounts 0 :transaction-account/account)→transaction/accounts[0][transaction-account/account]). They decode straight into the form schema via the unchangedwrap-nested-form-params+mc/decode— no two-key snapshot/step-params decode. Strip stray keys after decode (select-keysto the schema's keys) or a non-schema input like the tab group'smethodhidden 500s the save (seegotchas.md). - Flat state.
wrap-derive-statebuilds a plain{:snapshot :edit-path :step-params}map (not theMultiStepFormStaterecord):entity-onlyfields from the entity, editable fields from the live posted form (absent = cleared). The ~34:snapshotreads keep working. - Validation/error flow without
wrap-ensure-step. Reuse the genericwrap-form-4xx-2directly:(-> submit-edit (wrap-form-4xx-2 render-form-response) …).submit-editrunsassert-schemathen dispatches the save; on a throw,wrap-form-4xx-2re-renders the whole form with:form-errorskeyed by schema paths. A*errors*dynamic var (bound byrender-form) replaces the form-cursor's*form-errors*for field lookups. - Routes shrink to
edit-wizard(GET open),edit-submit(POST),edit-form-changed(POST whole-form re-render for dependent changes),location-select(GET),unlink-payment(POST).
Genuinely multi-step → data-driven engine with session-stored step state
Inspiration — Django
formtoolsWizardView. Django's wizard does not round-trip a serialized blob of the whole form through the page. Each step's validated data is written to a storage backend (the user session by default) under that step's key, and the steps are combined only at the very end viaget_all_cleaned_data(). We adopt the same model: replace the EDN snapshot + piecewise merging with per-step form state stored in the Ring session. A step writes its own data under its own key; nothing is merged into a snapshot and nothing about other steps rides through the form. Refs:formtools.wizard.views.WizardView,SessionStorage,get_all_cleaned_data()(https://django-formtools.readthedocs.io/en/latest/wizard.html).
A wizard is data:
(def vendor-wizard-config
{:steps [{:key :info :schema info-schema :fields [...] :render render-info-step
:next (fn [data] :terms)}
{:key :terms :schema terms-schema :fields [...] :render render-terms-step
:next (fn [data] :done)}]
:init-fn (fn [req] {...})
:submit-route "/admin/vendor/wizard/submit"
:done-fn (fn [all-data req] (save! all-data) (html-response "Saved"))})
with a tiny engine (no protocols) whose state lives in the session, keyed by a wizard
instance id, each step's data under its own step key — the formtools SessionStorage
model. No snapshot, no custom EDN readers, no merge-into-snapshot:
;; Storage backed by the Ring session. Path: [:wizards <wizard-id> :step-data <step-key>]
(defn create-wizard! [session config]
(let [id (str (java.util.UUID/randomUUID))]
[id (assoc-in session [:wizards id]
{:current-step (-> config :steps first :key) :step-data {}})]))
(defn put-step [session id k data] (assoc-in session [:wizards id :step-data k] data)) ; replace, not merge
(defn set-step [session id k] (assoc-in session [:wizards id :current-step] k))
(defn get-all [session id] (->> (get-in session [:wizards id :step-data]) vals (apply merge)))
(defn forget [session id] (update session :wizards dissoc id))
The render emits only a reference token (wizard-id, current-step) in the form —
never the form's state. The submit handler validates the posted step, put-steps it,
computes :next, and either advances (set-step) or finishes (get-all + :done-fn +
forget). Every fn returns the updated session for the handler to thread into the Ring
response ((assoc resp :session session')).
Two routes per wizard: open (partial open-wizard config) and submit
(partial handle-step-submit config). State is namespaced by wizard-id inside the
session, so multiple in-flight wizards (and browser tabs) don't collide, and it is
discarded on completion (forget).
Storage lifetime (Open decision 1)
State lives in the Ring session, scoped to true multi-step wizards (plain forms hold
none). Lifetime follows the session; forget on completion prevents session bloat. For
long-lived wizards, confirm the session backend (in-memory vs. durable) is acceptable or
pick a durable store.
The engine — REALIZED (Phase 6)
Built and REPL-proven in Phase 6 as two namespaces (no protocols, no defrecords):
-
auto-ap.ssr.components.wizard-state— the pure session-storage layer (the skeleton above, fleshed out):create-wizard!/instance/exists?/current-step/context/step-data/put-step(replace) /set-step/get-all/forget. Each issession -> session'(or a read); nothing mutates global state.:contextholds read-only data the steps need (e.g. an entity id) outside:step-data, so it never gets merged into the combined result. -
auto-ap.ssr.components.wizard2— the engine:open-wizard,render-wizard,handle-step-submit, and thewizard-formshell. A wizard is a config map:{:name :vendor :form-id "wizard-form" :submit-route "<resolved url>" :init-fn (fn [request] {:context {...} :init-data {step-key data}}) :done-fn (fn [all-data request] ring-response) :steps [{:key :info :decode (fn [request] -> data-map) ; parse this step's posted fields :validate (fn [data request] -> errors|nil) ; optional :render (fn [ctx] -> hiccup) ; step body; engine wraps the <form> :next (fn [data] -> next-step-key | :done)} ...]}The step's
:rendergets{:wizard-id :current-step :context :all-data :step-data :errors :request :config}.:all-data(every step combined so far) is exactly what a read-only summary/preview step consumes. Nav buttons post adirectionfield:"next"(validate+advance via:next),"back"(no validate),"submit"(== next, for the last step). Onlywizard-id+current-stepride in the form — no snapshot.
Two routes per wizard: (partial open-wizard config) (GET) and
(partial handle-step-submit config) (POST). No wrap-wizard / wrap-decode-multi-form-state
stack — the engine threads the session itself and (assoc resp :session session').
Proven via REPL (lifecycle, before any modal used it): open seeds session state and
renders step 1 with no accumulated data in the form; next stores {step-key data} and
advances; an invalid step re-renders itself with errors (no advance); the final step's
:done calls done-fn with the combined get-all data and forgets the instance; back
navigates without validating; an unknown/expired wizard-id re-opens fresh instead of
500-ing. See the lifecycle eval in the Phase 6 commit message.
Note (Phase 6 fit). Transaction Rule itself is edit + read-only preview of one
entity, not a true multi-data-step flow — so it exercises the engine's render/navigation/
preview path (:all-data feeds the test table) but not the cross-step merge. The merge
(get-all combining independent steps) gets its real workout in Phase 7+ (Invoice Pay,
New Invoice, Vendor, Client), where steps collect genuinely different fields.
Conditional :next + dual-purpose (new+edit) — New Invoice (Phase 8)
A step's :next is just (fn [data] -> next-step-key | :done), so branching the flow is a
one-liner — no CustomNext protocol, no 308-redirect-to-submit hack:
{:key :basic-details
:next (fn [data] (if (= :customize (:customize-accounts data)) :accounts :done))}
:default skips the expense-accounts step entirely (the done-fn uses the vendor's default
account); :customize routes through the grid. The old wizard expressed this with
mm/CustomNext returning either navigate-handler{:to :accounts} or a 308 to the submit
route — and the 308 path was broken (see gotchas.md, the {}→nil 500). The engine's
conditional :next is both simpler and correct.
Dual-purpose (create and edit) = one config, one :init-fn that branches on a route id:
(defn new-init-fn [request]
(if-let [id (->db-id (get-in request [:route-params :db/id]))]
{:init-data {:basic-details (… entity prefilled, :customize-accounts :customize)
:accounts {:invoice/expense-accounts (… existing rows)}}} ; edit
{:init-data {:basic-details {:invoice/date (coerce/to-date (time/now)) ; new
:customize-accounts :default}}}))
create-wizard! stores :init-data as the per-step :step-data map directly, so seeding
{:basic-details … :accounts …} opens both steps populated — the edit case repopulates the
grid without a separate hydrate. Two open routes (new-wizard, edit-wizard) both reduce to
(partial wizard2/open-wizard config); the done-fn branches on (:db/id all-data) to return
the next-steps modal (create) vs the swapped table row (edit).
Async step fragments read the posted form, not multi-form-state. The basic-details
fragments (account-prediction radio, due-date / scheduled-payment suggestions) and the
accounts totals all post the whole #wizard-form; in the engine that form carries the flat
invoice/* fields + the opaque wizard-id, so a fragment decodes what it needs straight from
form-params (and, for a cross-step value like the invoice total on the accounts step, reads
ws/get-all via the posted wizard-id). No mm/wrap-decode-multi-form-state stack survives.