# Design: optional per-printing tracking

Status: **implemented.** The as-built decisions are in
[As built](#as-built) at the end, which takes precedence over the original
design below wherever the two differ.

## Context

`mtgc` currently identifies every owned/decked card purely by `oracle_id`
(one row + a quantity column in both `library_cards` and `deck_cards`). There
is no way to record *which printing* a physical copy is, or to own multiple
copies of the same card across different printings (e.g. one non-foil `2X2`
copy and one foil `PRNA` promo copy of the same card).

Notably, the schema already anticipated this: `library_card_copies`
([0001_init.sql](../crates/mtgc-core/src/sql/library/0001_init.sql)) was
defined at the very start of the project as "optional per-physical-copy
printing detail (the 'hybrid' identity)" — but it has never been read or
written by any code path. This document designs activating it.

This was split out from the "normal default card images" fix (see the ingest
change in `crates/mtgc-core/src/ingest.rs` / `db.rs` that made the *default*
image for a card come from Scryfall's representative Oracle Cards printing).
That fix solves "the default preview looks weird"; this document solves "I
want to track/display a *specific* printing for a specific copy or deck."

**Decision from the requesting conversation:** a tracked copy should record
printing **+ finish** (nonfoil/foil/etched). Condition (NM/LP/etc.) is
deferred — the dormant table already has a `condition` column, so adding it
later is a compatible follow-up, not a redesign.

## Identity model — "hybrid"

- `library_cards` **stays authoritative** for totals: `oracle_id` +
  `quantity_owned`. Every existing oracle-level feature (legality checks, deck
  analysis, synergy suggestions, over-assignment warnings) keeps working
  completely untouched.
- `library_card_copies` becomes *optional* annotations: each row means "N of
  my copies of this card are this specific printing + finish." Copies are a
  subset of `quantity_owned` — a card can have some copies pinned to a
  printing and the rest left unspecified. The UI must surface that remainder
  (e.g. "3 owned · 1 tracked (2X2 foil) · 2 unspecified") rather than forcing
  every copy to be classified.
- Decks get a per-card **display preference** — which printing's art to show
  for that deck slot — independent of quantity/legality. This is deliberately
  *not* tied to a specific owned copy; it's just "when I look at this deck,
  show me this version."

## Schema / migrations

### library.db (append-only migrations, per the existing convention in `db.rs`)

- **`0005_copy_quantity_finish.sql`** — add `quantity INTEGER NOT NULL DEFAULT 1`
  to `library_card_copies`. Without this, the dormant table is one-row-per-copy,
  which is awkward for e.g. "4 copies, all the same 2X2 nonfoil printing." A
  quantity column lets one row represent a group of identical copies. (The
  `finish` column already exists in the original table; `condition` stays,
  unused for now.)
- **`0006_deck_card_printing.sql`** — add `preferred_scryfall_id TEXT` (nullable)
  to `deck_cards`. Pure display preference, no FK enforcement needed (printings
  live in the read-only, rebuildable reference.db, not library.db — a dangling
  reference after a re-ingest should just fall back to the default image, not
  error).

### reference.db (edit `reference_schema.sql` directly — rebuilt, not migrated)

- Add `lang TEXT NOT NULL DEFAULT ''` and `released_at TEXT NOT NULL DEFAULT ''`
  to `printings`, for picker labels ("English, 2X2, 2022-07-08") and stable
  chronological ordering. Both fields are already present on Scryfall printing
  objects, so `load_printings` just needs to read two more fields — no new
  ingest pass.

## Core (`mtgc-core`)

- New `PrintingRef` model: `scryfall_id, set_code, collector_number, lang,
  finishes, rarity, image_uri, back_image_uri, price_usd, released_at`.
- `Catalog::list_printings(oracle_id) -> Vec<PrintingRef>`, ordered by
  `(released_at DESC, rowid)`, for the picker UI.
- Library copy CRUD in `library.rs`:
  - `set_copy(oracle_id, scryfall_id, finish, quantity)` — upsert a copy group.
  - `list_copies(oracle_id) -> Vec<LibraryCardCopy>`.
  - `remove_copy(copy_id)`.
  - Validation: sum of copy quantities for an `oracle_id` must stay
    `<= quantity_owned`; reducing `quantity_owned` below the tracked sum should
    either clamp copies or surface a conflict — pick one and document it before
    implementing (leaning toward clamping the newest/most-recently-edited copy
    row, since silently deleting user-entered printing data is worse than
    losing precise quantity).
- Deck: `set_deck_card_printing(deck_id, oracle_id, scryfall_id | null)`.
- **Display resolution:** extend the image-fallback chain added for the
  default-image fix so a deck card with `preferred_scryfall_id` set resolves
  its image from that specific printing first:
  `COALESCE(<preferred printing image>, oc.image_uri, <old per-printing
  subquery>)`. No component changes needed — everything still just reads
  `card.image_uri` off the `CardRef`/`DeckEntry` it's given.

## Commands + `src/api.ts`

- Tauri commands: `list_printings(oracle_id)`, `set_library_copy(...)`,
  `list_library_copies(oracle_id)`, `remove_library_copy(copy_id)`,
  `set_deck_card_printing(deck_id, oracle_id, scryfall_id)`.
- `src/api.ts`: matching invoke wrappers; new `PrintingRef` and
  `LibraryCardCopy` TS interfaces; extend `DeckEntry` with an optional
  `preferred_scryfall_id`.

## Frontend (`src/`)

- New `PrintingPicker` modal (in `components.tsx`): given an `oracle_id`, fetch
  printings and render a thumbnail grid with set / language / finish / price
  labels; `onPick(scryfallId)`.
- **DeckView** ([src/views/DeckView.tsx](../src/views/DeckView.tsx)): a
  per-card "Choose printing" affordance opens the picker and calls
  `set_deck_card_printing`; the deck's thumbnail/preview then reflects that
  printing.
- **LibraryBrowser** / **CardEntry**: a per-card "Versions" affordance to
  assign printing + finish + quantity to copies; show the
  specified-vs-unspecified breakdown next to the existing quantity stepper.

## Open questions to settle before implementing

1. Should the library *list/tile thumbnail* follow a pinned copy's printing
   (and if a card has multiple different tracked printings, which one wins?),
   or should the thumbnail always show the oracle default and per-copy art
   only appear in a detail/expanded view? Recommend the latter — simpler, and
   avoids an arbitrary tie-break in the common list view.
2. Should decklist import/export (`decklist.rs`) round-trip printing
   preferences (many decklist formats support a `(SET) 123` suffix)? Out of
   scope for the initial cut; note it as a natural follow-up once the core
   model exists.
3. Quantity-vs-copies conflict resolution (see `set_copy` validation above) —
   decide the exact clamping behavior before writing the command, and add a
   unit test that pins it down.

## Verification plan (once implemented)

- Rust unit tests: copy CRUD + quantity clamping; preferred-printing image
  resolution (deck card with `preferred_scryfall_id` set vs. unset vs.
  dangling after a re-ingest).
- Command round-trip smoke test through `src-tauri`.
- Manual: assign a printing to a deck card and confirm the deck preview/thumbnail
  changes to that art; record two copies of one card under two different
  printings and confirm the library view shows the specified/unspecified
  breakdown correctly.

## Critical files

`crates/mtgc-core/src/sql/library/0005_*.sql`, `0006_*.sql`,
`crates/mtgc-core/src/sql/reference_schema.sql`, `crates/mtgc-core/src/ingest.rs`,
`crates/mtgc-core/src/library.rs`, `crates/mtgc-core/src/decks.rs`,
`crates/mtgc-core/src/model.rs`, `src-tauri/src/commands.rs`, `src/api.ts`,
`src/components.tsx`, `src/views/DeckView.tsx`, `src/views/LibraryBrowser.tsx`,
`src/views/CardEntry.tsx`.

## As built

Where the implementation settled the open questions, or departed from the
design above:

- **Migration** is `0006_printings.sql` (0005 had gone to deck uuids). It adds
  `library_card_copies.quantity`, a unique index on
  `(library_card_id, scryfall_id, finish)`, and `deck_cards.preferred_scryfall_id`.
  Everything lives in `crates/mtgc-core/src/printings.rs`.
- **No `remove_copy(copy_id)`.** `set_library_copy(.., quantity)` with
  `quantity <= 0` removes a group, mirroring `set_deck_card`. Row ids are local
  and are rebuilt on every sync, so they would be a poor handle anyway.
- **Quantity rules (open question 3).**
  - *Grow:* setting a group above the unspecified remainder raises
    `quantity_owned` to fit.
  - *Clamp:* lowering `quantity_owned` below the tracked total trims the
    highest-id (most recently added) groups first.
  - Lowering a group never lowers `quantity_owned`; those copies become
    unspecified.
  - All of this is pinned by tests in `printings.rs`.
- **Library thumbnails (open question 1)** always show the oracle default.
  Only deck slots show a chosen printing.
- **Decklists (open question 2)** round-trip in this first cut, as
  `(SET) NUM`, optionally followed by `*F*` / `*E*`. See `decklist.rs`.
- **UI.** There is no per-row button. A *Printings* section in the
  click-preview info tile appears whenever the caller passes a
  `PreviewContext`, and it opens a `PrintingPicker`. Both are in
  `src/components.tsx`.
- **Sync, which postdates this design.**
  - `CardEntry.printings` is keyed `"<scryfall_id>/<finish>"`, and
    `DeckCardEntry.printing` holds the chosen printing.
  - Cards now merge field by field, and their groups merge per key, so edits
    to one card from two devices both survive. A merge result is clamped by
    trimming from the end of key order, which comes out the same on every
    device.
  - A document is written as version 2 only once it holds printing data, so
    older builds refuse it rather than dropping the data and syncing the loss.
- **Follow-up:** choosing a printing while entering cards on Add Cards.