save-systems

# Save versioning, migration, and robustness

Saves outlive the code that wrote them. A player who saved in v1.0 expects to
load in v1.4 after an update. Plan for schema evolution from the first release.

## The version field is non-negotiable

Stamp every save with an integer `version` (or a structured `{major, minor}` if
you need it). On load:

1. Read `version`.
2. If it is **newer** than the running build → refuse with a clear message
   ("This save is from a newer version"). Do not guess at fields you don't know.
3. If it is **older** → run migrations in sequence until it matches the current
   version.
4. Validate, then instantiate.

Prefer a single monotonic integer. Semantic two-part versions are only worth it
if you genuinely branch save compatibility across release lines.

## The migration chain

Write each migration as a **pure function** `vN -> vN+1` and apply them in order.
This keeps any version reachable to current with small, testable steps — never a
giant "convert anything to latest" function.

```python
SAVE_VERSION = 4

MIGRATIONS = {
    1: migrate_1_to_2,   # added "flags" map
    2: migrate_2_to_3,   # split "name" into first/last
    3: migrate_3_to_4,   # inventory items gained a "durability" field
}

def upgrade(data):
    v = data.get("version", 1)
    while v < SAVE_VERSION:
        data = MIGRATIONS[v](data)
        v += 1
        data["version"] = v
    return data
```

Guidelines:

- **Additive changes are easy**: new field with a default. The migration just
  inserts the default for old saves.
- **Renames/restructures**: move/transform in the migration; never read both old
  and new names in game code.
- **Removed fields**: drop them in the migration so later steps see a clean shape.
- **Keep old migrations forever.** Deleting `migrate_1_to_2` strands every v1
  save. They are cheap to keep and must remain a faithful record.
- **Test the chain**: keep a corpus of real saves from each shipped version and
  assert they load after every change (this is your regression net).

## Backups and rollback

- Before overwriting, copy the current save to `*.bak`. On a failed load of the
  primary, try the backup automatically.
- For autosaves, rotate a small ring (autosave_0..2) so a bad autosave doesn't
  destroy the only recent state.
- Consider a checksum/hash of the payload to detect truncation or corruption
  (distinct from anti-tamper, which a determined local player can always defeat).

## Format trade-offs

| Format | Pros | Cons | Use when |
|---|---|---|---|
| JSON / text | Human-readable, debuggable, diff-able, easy migration | Larger, slower, trivially editable | Default; most games |
| Binary (engine) | Compact, fast, mildly opaque | Hard to debug/migrate, version-fragile | Large state, perf-critical |
| Key/value store | Simple, robust per-key | No structure, manual schema | Settings, small flag sets |
| Cloud KV (e.g. DataStore) | Cross-device | Quotas, latency, conflicts | Online/cross-device saves |

Whatever the on-disk format, keep an **in-memory dictionary/struct** as the
canonical shape and (de)serialize at the boundary. Migrations then operate on
that neutral shape, not on format-specific bytes.

## Load-time validation checklist

- Required keys present; types correct (string vs number vs bool).
- Numbers within sane ranges (no NaN/Infinity; hp ≥ 0; indices in bounds).
- Enum/id values still exist in current content (an item id removed in a patch
  must be handled — drop it or substitute, don't crash).
- References resolve (a quest id points to a known quest).
- On any failure: log, fall back to backup, and if all else fails, fail *loudly*
  to the player rather than silently loading a broken state.

## What to store vs recompute

- **Store** authoritative, non-derivable state: choices, unlocked flags, RNG
  seed, inventory, positions.
- **Recompute** derived data on load (full stat tables from base + modifiers,
  procedural maps from the seed). Saving derived data invites desync when the
  formula changes in a patch — and bloats the file.

---
name: save-systems
description: >
  Design save/load for game state — choosing what to serialize, file formats,
  save slots, atomic crash-safe writes, schema versioning and migration, and
  autosave. Engine-neutral. Use when the user mentions save system, save/load,
  game state persistence, save slots, autosave, save file corruption, or
  migrating old saves to a new version.
---

# Save systems

A save file is a **serialized snapshot of game state** that survives restarts.
The hard parts aren't writing bytes — they're choosing *what* to save, writing it
so a crash mid-save can't corrupt it, and reading *old* saves after you ship a
patch. Get those three right and the rest is plumbing.

## When to use

- Use to persist progress: player stats, inventory, world flags, settings,
  positions — across sessions and game updates.
- Use to design save slots, quicksave/autosave, and crash-safe writes.
- Use when old save files break after a content/code change (versioning &
  migration).

**When *not* to use:** for Roblox cloud persistence specifics, use
`roblox-datastores`. For the data model the save serializes (resources/SOs), use
`godot-resources` / `unity-scriptableobjects`. For Godot's `FileAccess`/
`ResourceSaver` and `user://` paths, defer to the Godot engine skill while
applying the patterns here.

## Core workflow

1. **Decide what state is authoritative.** Save the *data* (hp, position, seed,
   unlocked flags), not engine objects or scene nodes. You will reconstruct
   objects from data on load — never serialize live node references.
2. **Define a versioned schema.** Every save embeds a `version` integer. This is
   the single most important field for a game you intend to patch.
3. **Pick a format.** JSON/text for readability and debuggability; a binary
   format for size/speed or mild tamper-resistance. Start with JSON.
4. **Write atomically.** Serialize to a temp file, flush, then rename over the
   real file. A crash leaves either the old save or the new one — never a
   half-written one.
5. **Load defensively.** Read version → migrate up to current → validate →
   instantiate. Keep a backup of the last good save and fall back on parse error.
6. **Autosave on safe boundaries** (level change, checkpoint), throttled, and to a
   separate slot so it can't clobber a manual save.
7. **Verify**: save, fully quit, relaunch, load — and confirm by inspection that
   state matches. Test loading a save from the previous version.

## Patterns

### 1. Serialize state as plain data (not engine objects)

```gdscript
# Build a dictionary of pure data. Each savable object reports its own state.
func capture_state() -> Dictionary:
    return {
        "version": SAVE_VERSION,                 # ALWAYS stamp the schema version
        "player": { "hp": player.hp, "pos": [player.position.x, player.position.y] },
        "inventory": player.inventory.to_array(),  # ids + counts, not Item nodes
        "flags": world.flags,                    # e.g. {"met_guard": true}
        "seed": world.seed,                      # regenerate procedural content
    }

# On load, RECONSTRUCT objects from the data — do not expect live references back.
func apply_state(data: Dictionary) -> void:
    player.hp = data["player"]["hp"]
    player.position = Vector2(data["player"]["pos"][0], data["player"]["pos"][1])
    player.inventory.from_array(data["inventory"])
    world.flags = data["flags"]
```

### 2. Atomic, crash-safe write (temp + rename)

```gdscript
# RIGHT: write to a temp file, then atomically rename over the target.
func save_atomic(path: String, data: Dictionary) -> void:
    var tmp := path + ".tmp"
    var f := FileAccess.open(tmp, FileAccess.WRITE)
    f.store_string(JSON.stringify(data))
    f.flush()                                    # ensure bytes hit disk
    f.close()
    DirAccess.rename_absolute(tmp, path)         # replaces the target; atomic on POSIX
# WRONG: opening `path` directly and writing in place — a crash mid-write leaves a
# truncated, unloadable save and destroys the player's progress.
```

Rename-over-target is atomic on POSIX (same volume); on Windows a replace-by-rename
isn't guaranteed atomic, so keep the previous file as `path + ".bak"` before the
rename — that backup is what actually guarantees you can recover from a bad write.

### 3. Versioned load with migration

```python
SAVE_VERSION = 3

def load_save(raw_bytes):
    data = parse(raw_bytes)                  # JSON/binary -> dict
    v = data.get("version", 0)
    if v > SAVE_VERSION:
        raise NewerSaveError(v)              # save is from a newer build; refuse
    while v < SAVE_VERSION:                   # apply migrations in order, v -> v+1
        data = MIGRATIONS[v](data)
        v += 1
        data["version"] = v
    validate(data)                            # check required keys / ranges
    return data

# Each migration is a pure function from one version's shape to the next.
def migrate_1_to_2(d):
    d["flags"] = {k: True for k in d.pop("completed_quests", [])}  # list -> set-map
    return d
MIGRATIONS = {1: migrate_1_to_2, 2: migrate_2_to_3}
```

### 4. Save slots + throttled autosave

```gdscript
const SLOT_PATH := "user://save_%d.json"      # manual slots 0..N
const AUTOSAVE_PATH := "user://autosave.json"  # separate file: never clobbers a slot
var _autosave_cooldown := 0.0

func autosave_if_due(dt: float) -> void:
    _autosave_cooldown -= dt
    if _autosave_cooldown <= 0.0:
        save_atomic(AUTOSAVE_PATH, capture_state())
        _autosave_cooldown = 60.0             # throttle: at most once a minute
# Trigger an immediate autosave on checkpoints/level transitions, not mid-combat.
```

## Pitfalls

- **Serializing engine objects/node paths** ties saves to scene structure;
  renaming a node breaks every old save. Save data, rebuild objects on load.
- **No version field.** The day you ship a patch, every existing save is a
  guessing game. Stamp `version` from version 1.
- **In-place writes** corrupt saves on crash/power loss. Always temp-write then
  rename; keep a `.bak`.
- **Trusting the file blindly.** Saves get truncated, hand-edited, or
  cloud-synced stale. Validate on load and fall back to backup on failure.
- **Floats and locale.** Text serializers can drop precision or use comma
  decimal separators in some locales. Use a locale-invariant serializer.
- **Autosave clobbering manual saves**, or firing mid-action and saving an
  inconsistent state. Use a dedicated autosave slot and save on safe boundaries.
- **Storing secrets or trusting client saves in multiplayer.** A local save is
  player-controlled; never treat it as authoritative for online state. For cloud,
  handle the device's data limits and conflicts (`roblox-datastores`).

## References

- `references/versioning-and-migration.md` — schema evolution strategies, the
  migration chain, backups/rollback, format trade-offs (JSON vs binary), and a
  load-time validation checklist.

## Related skills

- `roblox-datastores` — cloud persistence, request limits, session locking.
- `godot-resources`, `unity-scriptableobjects` — the data model you serialize.
- `procedural-gen` — store the seed to regenerate worlds instead of saving them.
- `rpg`, `survival-crafting`, `visual-novel` — genres that compose this skill.