513 lines
18 KiB
Markdown
513 lines
18 KiB
Markdown
|
# Machine Actor: 设备状态机 goroutine 化
|
|||
|
|
|
|||
|
|
> **目标:** 每台设备一个 goroutine，自治管理槽位状态和 PLC 信号监听，替换当前散落在 EventLoop/Station/SignalWatcher 中的设备逻辑。
|
|||
|
|
|
|||
|
|
**架构:** Actor 模型 — goroutine + channel，私有状态，外部只能发消息/读快照。全局调度器聚合快照做跨设备决策，EventLoop 保持 DB 单线程写。
|
|||
|
|
|
|||
|
|
**实现方式:** 重写。不迁移旧 Station 代码，新 Actor 替换旧 Station 后删除旧代码。
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 1. 动机
|
|||
|
|
|
|||
|
|
当前问题：
|
|||
|
|
|
|||
|
|
1. **设备状态更新散落两处** — `handleMachineDone` 先写 DB equipment_slot，再调 `updateStationDone` 写 Station 内存，两个操作不在同一事务
|
|||
|
|
2. **调度器读两个源** — `buildSystemState` 从 Station 内存读 `MachineBusy`，从 DB 读 `MachineHasJob`，语义重叠但不同步
|
|||
|
|
3. **handleMachineDone 职责过重** — 批量/单信号设备分支、slot 查找、DB 更新、Station 更新、Job 推进混在一个方法
|
|||
|
|
4. **SignalWatcher 集中轮询** — 所有设备走一个 ticker，设备数量增加时轮询粒度不可控
|
|||
|
|
|
|||
|
|
核心矛盾：设备状态（槽位 Empty/Occupied/Done + 工站 Idle/Processing/Full/Waiting）天然属于设备自身，但当前被 EventLoop 和 Station 分而治之。
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 2. 架构概览
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
PLC 硬件信号
|
|||
|
|
↓ SignalRouter 集中轮询，路由到各 Actor channel
|
|||
|
|
┌──────────────────────────────────────────────────────────────┐
|
|||
|
|
│ MachineActor-1 MachineActor-2 TempStoreActor │
|
|||
|
|
│ (CNC #1, goroutine) (清洗机, goroutine) (暂存台, goroutine) │
|
|||
|
|
│ │
|
|||
|
|
│ 私有: slots[],status 私有: slots[],status 私有: slots[] │
|
|||
|
|
│ 外部只能: Send/Snapshot │
|
|||
|
|
│ │
|
|||
|
|
│ 自写 DB: equipment_slot 自写 DB: job │
|
|||
|
|
│ 发布事件: EvtMachineDone 无 PLC 信号 │
|
|||
|
|
└──────┬───────────────────────────────────────────────────────┘
|
|||
|
|
│ 发布事件到 EventBus
|
|||
|
|
▼
|
|||
|
|
GlobalScheduler ──── 聚合所有 Actor 快照
|
|||
|
|
│ 换料配对、优先级排序
|
|||
|
|
│ 输出 CandidateTask
|
|||
|
|
▼
|
|||
|
|
EventLoop ────────── Job 步骤推进 (DB)
|
|||
|
|
│ AdvanceStep / FinishJob
|
|||
|
|
│ 不碰 equipment_slot
|
|||
|
|
▼
|
|||
|
|
PostgreSQL
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
**职责边界：**
|
|||
|
|
|
|||
|
|
| 组件 | 职责 | 不负责 |
|
|||
|
|
|------|------|--------|
|
|||
|
|
| MachineActor | 槽位状态机、PLC 信号监听、完成判定、DB equipment_slot 写入 | Job 步骤推进 |
|
|||
|
|
| TempStoreActor | 暂存台槽位分配/释放、DB job.temp_slot_no 写入 | Job 步骤推进 |
|
|||
|
|
| GlobalScheduler | 跨设备决策（换料配对、优先级） | 设备内部状态 |
|
|||
|
|
| EventLoop | Job 步骤推进、DB job 表写入 | 设备/暂存台状态管理 |
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 3. MachineActor 设计
|
|||
|
|
|
|||
|
|
### 3.1 接口
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
// MachineActor 设备状态机
|
|||
|
|
type MachineActor interface {
|
|||
|
|
ID() int // 设备 ID
|
|||
|
|
Type() string // 设备类型编码
|
|||
|
|
Run(ctx context.Context) // 启动 goroutine，阻塞
|
|||
|
|
Send(msg MachineMsg) // 外部发消息（非阻塞，chan 容量 8）
|
|||
|
|
Snapshot() MachineSnapshot // 读快照（atomic.Value，无锁）
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
// MachineMsg 外部发给 Actor 的命令
|
|||
|
|
type MachineMsg struct {
|
|||
|
|
Type string // LOAD_COMPLETE / UNLOAD_COMPLETE / RELEASE / RESTORE
|
|||
|
|
JobID int
|
|||
|
|
SlotNo int
|
|||
|
|
Params map[string]any
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
// MachineSnapshot 只读快照，供调度器聚合
|
|||
|
|
type MachineSnapshot struct {
|
|||
|
|
ID int
|
|||
|
|
Type string
|
|||
|
|
Status ActorStatus // Idle/Processing/Full/Waiting/Fault
|
|||
|
|
Slots []SlotSnapshot
|
|||
|
|
UpdatedAt time.Time
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
type SlotSnapshot struct {
|
|||
|
|
SlotNo int
|
|||
|
|
Status SlotStatus // Empty/Occupied/Done
|
|||
|
|
JobID int
|
|||
|
|
OccupiedAt time.Time
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 3.2 工站状态机
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
┌──────────────────────────┐
|
|||
|
|
│ FAULT │←── PLC 通信异常 / 硬件故障
|
|||
|
|
│ (需人工介入恢复) │
|
|||
|
|
└──────────────────────────┘
|
|||
|
|
↑ 异常
|
|||
|
|
│
|
|||
|
|
┌──────────┐ Load完成 ┌──────────────┐ 满 ┌──────────┐
|
|||
|
|
│ IDLE │──────────→ │ PROCESSING │───→ │ FULL │
|
|||
|
|
│ (全Empty)│←────────── │ (有空+有占) │←─── │ (全占满) │
|
|||
|
|
└──────────┘ Unload/全部 └──────────────┘ └──────────┘
|
|||
|
|
槽位释放 │ ↑ ↑ │
|
|||
|
|
│ │ │ │
|
|||
|
|
│ │ └─ Load(有空位) │
|
|||
|
|
│ │ │ PLC 完成信号
|
|||
|
|
│ └── PLC Done ────────┘
|
|||
|
|
│
|
|||
|
|
│ PLC 完成信号（非批量设备）
|
|||
|
|
▼
|
|||
|
|
┌──────────────┐
|
|||
|
|
│ WAITING │ ← 有 Done 槽位，等待卸料/换料
|
|||
|
|
│ (有Done槽位) │
|
|||
|
|
└──────────────┘
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
状态规则：
|
|||
|
|
- **IDLE**: 所有槽位 Empty
|
|||
|
|
- **PROCESSING**: 有空 + 有占。批量设备单信号完成时短暂停留
|
|||
|
|
- **FULL**: 全部 Occupied 或 Occupied+Done 之和=capacity，不接受上料
|
|||
|
|
- **WAITING**: 有至少一个 Done 槽位，调度器优先处理
|
|||
|
|
- **FAULT**: PLC 通信异常，需人工恢复
|
|||
|
|
|
|||
|
|
### 3.3 槽位状态机
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
Empty ──Load完成──→ Occupied ──PLC Done──→ Done ──Unload完成──→ Empty
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 3.4 批量设备 vs 单信号设备
|
|||
|
|
|
|||
|
|
两类设备完成行为不同，在 Actor 内部通过 `batch` 配置区分：
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
type machineActor struct {
|
|||
|
|
id int
|
|||
|
|
batch bool // true: 批量完成（所有 Occupied→Done）
|
|||
|
|
// ...
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
**批量设备**（`batch=true`，如高压清洗机）：收到 PLC 完成信号 → 所有 Occupied 槽位标记 Done
|
|||
|
|
**单信号设备**（`batch=false`，如 CNC）：收到 PLC 完成信号 → 按 OccupiedAt FIFO 找最早占用槽位标记 Done
|
|||
|
|
|
|||
|
|
### 3.5 检测设备
|
|||
|
|
|
|||
|
|
检测设备（内窥镜检测线、抽检台）在 Actor 中维护 `hasNG` bool：
|
|||
|
|
|
|||
|
|
- `handleMachineWaitCandidate` 的跳过行为保留在 EventLoop，不迁移到 Actor
|
|||
|
|
- 检测结果判定逻辑在 Actor 中：
|
|||
|
|
- Done 信号（pass）→ 发布 `EvtInspectionResult(machineID, true)`
|
|||
|
|
- NG 信号 → 发布 `EvtInspectionResult(machineID, false)`
|
|||
|
|
- EventLoop 消费 `EvtInspectionResult`：pass→CompleteStep→SetWaitingUnload；fail→FinishJob(Scrapped)+Alarm
|
|||
|
|
|
|||
|
|
### 3.6 Run 主循环
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
func (m *machineActor) Run(ctx context.Context) {
|
|||
|
|
ticker := time.NewTicker(m.pollInterval)
|
|||
|
|
defer ticker.Stop()
|
|||
|
|
|
|||
|
|
for {
|
|||
|
|
select {
|
|||
|
|
case msg := <-m.msgCh:
|
|||
|
|
m.handleMessage(msg)
|
|||
|
|
case evt := <-m.signalCh:
|
|||
|
|
m.handleSignal(evt) // 非阻塞接收 SignalRouter 推送
|
|||
|
|
case <-ctx.Done():
|
|||
|
|
return
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 3.7 消息处理
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
func (m *machineActor) handleMessage(msg MachineMsg) {
|
|||
|
|
switch msg.Type {
|
|||
|
|
case "LOAD_COMPLETE":
|
|||
|
|
m.slots[msg.SlotNo-1] = SlotStatus_Occupied
|
|||
|
|
m.slotJobs[msg.SlotNo-1] = msg.JobID
|
|||
|
|
m.slotOccupiedAt[msg.SlotNo-1] = time.Now()
|
|||
|
|
m.updateStatus()
|
|||
|
|
m.db.SetEquipmentSlot(ctx, m.id, msg.SlotNo, SlotStatus_Empty, SlotStatus_Occupied, msg.JobID)
|
|||
|
|
case "UNLOAD_COMPLETE":
|
|||
|
|
m.db.SetEquipmentSlot(ctx, m.id, msg.SlotNo, SlotStatus_Done, SlotStatus_Empty, 0)
|
|||
|
|
m.slots[msg.SlotNo-1] = SlotStatus_Empty
|
|||
|
|
m.slotJobs[msg.SlotNo-1] = 0
|
|||
|
|
m.slotOccupiedAt[msg.SlotNo-1] = time.Time{}
|
|||
|
|
m.updateStatus()
|
|||
|
|
case "RESTORE":
|
|||
|
|
m.slots[msg.SlotNo-1] = parseSlotStatus(msg.Params["status"])
|
|||
|
|
m.slotJobs[msg.SlotNo-1] = msg.JobID
|
|||
|
|
m.updateStatus()
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 3.8 信号处理 + DB 持久化
|
|||
|
|
|
|||
|
|
Actor 收到完成信号后：更新私有状态 → 写 DB → 发布事件通知 EventLoop 推进 Job。内存与 DB 在同一 goroutine 内顺序执行，无一致性问题。
|
|||
|
|
|
|||
|
|
检测设备的 Done 信号表示检测合格（pass），走 EvtInspectionResult 而非 EvtMachineDone。
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
func (m *machineActor) handleSignal(evt SignalEvent) {
|
|||
|
|
// 检测设备 NG 信号
|
|||
|
|
if evt.Type == SignalNG && evt.Value {
|
|||
|
|
m.bus.Publish(EventInspectionResult, m.id, false)
|
|||
|
|
return
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
if evt.Type == SignalDone && evt.Value {
|
|||
|
|
// 检测设备：Done 信号 = 检测合格
|
|||
|
|
if m.inspection {
|
|||
|
|
// 标记所有 Occupied 槽位为 Done
|
|||
|
|
for i := range m.slots {
|
|||
|
|
if m.slots[i] == SlotStatus_Occupied {
|
|||
|
|
m.slots[i] = SlotStatus_Done
|
|||
|
|
m.db.SetEquipmentSlot(ctx, m.id, i+1, SlotStatus_Occupied, SlotStatus_Done, m.slotJobs[i])
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
m.updateStatus()
|
|||
|
|
// 发布 EvtInspectionResult(pass=true)，EventLoop 处理
|
|||
|
|
m.bus.Publish(EventInspectionResult, m.id, true)
|
|||
|
|
return
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
// 非检测设备：正常加工完成
|
|||
|
|
var doneSlots []int
|
|||
|
|
if m.batch {
|
|||
|
|
for i := range m.slots {
|
|||
|
|
if m.slots[i] == SlotStatus_Occupied {
|
|||
|
|
m.slots[i] = SlotStatus_Done
|
|||
|
|
doneSlots = append(doneSlots, i+1)
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
} else {
|
|||
|
|
slotNo := m.findFirstOccupied()
|
|||
|
|
if slotNo > 0 {
|
|||
|
|
m.slots[slotNo-1] = SlotStatus_Done
|
|||
|
|
doneSlots = append(doneSlots, slotNo)
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
m.updateStatus()
|
|||
|
|
|
|||
|
|
for _, slotNo := range doneSlots {
|
|||
|
|
m.db.SetEquipmentSlot(ctx, m.id, slotNo, SlotStatus_Occupied, SlotStatus_Done, m.slotJobs[slotNo-1])
|
|||
|
|
}
|
|||
|
|
for _, slotNo := range doneSlots {
|
|||
|
|
m.bus.Publish(EventMachineDone, m.id, slotNo, m.slotJobs[slotNo-1])
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 4. SignalRouter
|
|||
|
|
|
|||
|
|
PLC 连接集中管理（S7 单连接），一个 goroutine 轮询，路由到各 Actor channel：
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
type SignalRouter struct {
|
|||
|
|
plc goplc.Client
|
|||
|
|
mu sync.RWMutex
|
|||
|
|
watches map[int]*SignalWatch
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
type SignalWatch struct {
|
|||
|
|
DoneAddr string // 完成信号 PLC 地址
|
|||
|
|
NGAddr string // NG 信号地址（检测设备才有）
|
|||
|
|
Ch chan SignalEvent
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
type SignalEvent struct {
|
|||
|
|
Type SignalType // Done / NG
|
|||
|
|
Value bool
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
- `Ch` 容量为 1，SignalRouter 写不阻塞
|
|||
|
|
- Actor 从 `Ch` 非阻塞读取（`select default`），积压时丢弃旧值
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 5. TempStoreActor（暂存台）
|
|||
|
|
|
|||
|
|
暂存台没有 PLC 信号，但需要管理槽位占用。与 MachineActor 的关键区别：**槽位从分配到释放贯穿 Job 整个生命周期**。
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
补料(分配slot3) → 上料到CNC(slot3仍属此job) → 下料回暂存台(slot3) → ... → 完成(释放slot3)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
中间 job 在机器上加工时，暂存台槽位是"虚占"状态 — 物理上工件不在，但槽位仍属此 job，不能被其他 job 使用。
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
type TempStoreActor struct {
|
|||
|
|
capacity int
|
|||
|
|
slots []int // slotNo → jobID (0 = 空)
|
|||
|
|
msgCh chan MachineMsg
|
|||
|
|
db *DBState
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
func (t *TempStoreActor) Run(ctx context.Context) {
|
|||
|
|
for {
|
|||
|
|
select {
|
|||
|
|
case msg := <-t.msgCh:
|
|||
|
|
switch msg.Type {
|
|||
|
|
case "ALLOCATE":
|
|||
|
|
t.allocate(ctx, msg.JobID)
|
|||
|
|
case "RELEASE":
|
|||
|
|
t.release(ctx, msg.JobID)
|
|||
|
|
case "RESTORE":
|
|||
|
|
t.restore(ctx, msg.SlotNo, msg.JobID)
|
|||
|
|
}
|
|||
|
|
case <-ctx.Done():
|
|||
|
|
return
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
func (t *TempStoreActor) allocate(ctx context.Context, jobID int) {
|
|||
|
|
for i := range t.slots {
|
|||
|
|
if t.slots[i] == 0 {
|
|||
|
|
t.slots[i] = jobID
|
|||
|
|
t.db.SetJobTempSlot(ctx, jobID, i+1) // 写 DB job.temp_slot_no
|
|||
|
|
return
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
func (t *TempStoreActor) release(ctx context.Context, jobID int) {
|
|||
|
|
for i := range t.slots {
|
|||
|
|
if t.slots[i] == jobID {
|
|||
|
|
t.slots[i] = 0
|
|||
|
|
t.db.ClearJobTempSlot(ctx, jobID) // 清 DB job.temp_slot_no
|
|||
|
|
return
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
func (t *TempStoreActor) Snapshot() TempStoreSnapshot {
|
|||
|
|
return TempStoreSnapshot{
|
|||
|
|
Capacity: t.capacity,
|
|||
|
|
Slots: append([]int(nil), t.slots...),
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
调度器通过 `Snapshot` 读暂存台占用，替代 `buildSystemState` 中的 `TempSlotJobs` 计算和 `TempSlotAllocator`。
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 6. 调度器集成
|
|||
|
|
|
|||
|
|
`buildSystemState` 从 Actor 单一源读（设备从 MachineActor，暂存台从 TempStoreActor）：
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
func (l *ProductionEventLoop) buildSystemState(ctx context.Context) scheduler.SystemState {
|
|||
|
|
state := scheduler.SystemState{
|
|||
|
|
MachineBusy: make(map[int]bool),
|
|||
|
|
MachineHasJob: make(map[int]int),
|
|||
|
|
TempSlotJobs: make(map[int]int),
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
for _, actor := range l.machineActors {
|
|||
|
|
snap := actor.Snapshot()
|
|||
|
|
state.MachineBusy[snap.ID] = snap.Status == ActorStatus_Full
|
|||
|
|
for _, slot := range snap.Slots {
|
|||
|
|
if slot.JobID > 0 && slot.Status != SlotStatus_Empty {
|
|||
|
|
state.MachineHasJob[snap.ID] = slot.JobID
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
// 暂存台
|
|||
|
|
tempSnap := l.tempStoreActor.Snapshot()
|
|||
|
|
occupied := 0
|
|||
|
|
for slotNo, jobID := range tempSnap.Slots {
|
|||
|
|
if jobID > 0 {
|
|||
|
|
state.TempSlotJobs[slotNo] = jobID
|
|||
|
|
occupied++
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
state.TempSlotFree = tempSnap.Capacity - occupied
|
|||
|
|
|
|||
|
|
return state
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
换料配对逻辑不变，仍在 `trySchedule` 阶段。
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 7. EventLoop 变更
|
|||
|
|
|
|||
|
|
**Job 状态持久化仍由 EventLoop 负责** — Job 是跨设备实体，不属于任何单一 Actor。
|
|||
|
|
|
|||
|
|
持久化职责分割：
|
|||
|
|
|
|||
|
|
| 表 | Owner | 时机 |
|
|||
|
|
|---|---|---|
|
|||
|
|
| `equipment_slot` | MachineActor | 上料/下料/完成信号时 |
|
|||
|
|
| `job`（状态/步骤/位置/上下文） | EventLoop | AdvanceStep/FinishJob/SetJobProcessing |
|
|||
|
|
| `work_order` | EventLoop | FinishJob 时累加 |
|
|||
|
|
| `alarm` | EventLoop | Actor 发布故障事件时 |
|
|||
|
|
| `job.temp_slot_no` | TempStoreActor | ALLOCATE/RELEASE 时 |
|
|||
|
|
|
|||
|
|
**新增：消费 Actor 事件，推进 Job**
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
case EventMachineDone:
|
|||
|
|
l.handleMachineDone(ctx, msg) // 精简后：只 AdvanceStep/SetWaitingUnload + trySchedule
|
|||
|
|
case EventInspectionResult:
|
|||
|
|
l.handleInspectionDone(ctx, machineID, pass)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
**handleMachineDone 精简**：Actor 已完成 DB equipment_slot 写入，EventLoop 仅做 Job 步骤推进。不再调 `updateStationDone`，不再查 slot 状态。仅从事件中拿 machineID/slotNo/jobID 参数。
|
|||
|
|
|
|||
|
|
**Worker 完成后通知 Actor**：上料/下料成功后发送 `LOAD_COMPLETE`/`UNLOAD_COMPLETE` 消息，Actor 更新内存槽位状态并写 DB。
|
|||
|
|
|
|||
|
|
**删除**：
|
|||
|
|
- `updateStationDone` 方法
|
|||
|
|
- `batchMachines` / `inspectionMachines` map — Actor 内部自管理
|
|||
|
|
- equipment_slot 的 DB 写入逻辑 — 移至 Actor
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 8. 启动与恢复
|
|||
|
|
|
|||
|
|
### 8.1 创建
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
func BuildMachineActors(ctx context.Context, entClient *ent.Client, signalRouter *SignalRouter, bus eventbus.Bus) map[int]MachineActor {
|
|||
|
|
equipments, _ := entClient.Equipment.Query().WithEquipmentType().All(ctx)
|
|||
|
|
actors := make(map[int]MachineActor)
|
|||
|
|
for _, eq := range equipments {
|
|||
|
|
cfg := MachineActorConfig{
|
|||
|
|
ID: eq.ID,
|
|||
|
|
Type: eq.Edges.EquipmentType.Code,
|
|||
|
|
Capacity: eq.SlotCount,
|
|||
|
|
Batch: eq.Batch,
|
|||
|
|
Inspection: eq.Edges.EquipmentType.Code == InspectionType, // 检测设备
|
|||
|
|
PollInterval: 1 * time.Second,
|
|||
|
|
SignalCh: signalRouter.Watch(eq.ID, doneAddr, ngAddr),
|
|||
|
|
Bus: bus,
|
|||
|
|
}
|
|||
|
|
actors[eq.ID] = NewMachineActor(cfg)
|
|||
|
|
}
|
|||
|
|
return actors
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 8.2 恢复
|
|||
|
|
|
|||
|
|
`RecoverOnStartup` 从 DB `equipment_slot` 表恢复初始状态：
|
|||
|
|
|
|||
|
|
```go
|
|||
|
|
func (d *DBState) RecoverMachineActors(ctx context.Context, actors map[int]MachineActor) {
|
|||
|
|
slots, _ := d.client.EquipmentSlot.Query().All(ctx)
|
|||
|
|
for _, s := range slots {
|
|||
|
|
actor, ok := actors[*s.EquipmentId]
|
|||
|
|
if !ok {
|
|||
|
|
continue
|
|||
|
|
}
|
|||
|
|
actor.Send(MachineMsg{
|
|||
|
|
Type: "RESTORE",
|
|||
|
|
JobID: slotJobID(s),
|
|||
|
|
SlotNo: s.SlotNo,
|
|||
|
|
Params: map[string]any{"status": string(s.Status)},
|
|||
|
|
})
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
RESTORE 直接设置槽位状态（跳过 Empty→Occupied 转换校验）。
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 9. 删除清单
|
|||
|
|
|
|||
|
|
新 Actor 上线后删除：
|
|||
|
|
- `internal/station/` — 整个 Station 包（接口、BaseStation、CNC/Cleaning/Washer/Inspection/Deburr/Sampling/Scanner/LaserMarker/Registry），保留 `HandheldTool` 接口
|
|||
|
|
- `internal/processor/temp_slot_allocator.go` — TempSlotAllocator（被 TempStoreActor 替代）
|
|||
|
|
- `internal/processor/signal_watcher.go` — SignalWatcher（被 SignalRouter 替代）
|
|||
|
|
- `internal/processor/step_timeout.go` — 工序超时
|
|||
|
|
- `internal/eventloop/loop.go` — `updateStationDone`、`batchMachines`、`inspectionMachines`
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 10. 测试策略
|
|||
|
|
|
|||
|
|
- **MachineActor 单元测试**: 注入模拟 signalCh，验证 Idle→Processing→Full→Waiting 全路径
|
|||
|
|
- **TempStoreActor 单元测试**: 验证 ALLOCATE/RELEASE 分配释放
|
|||
|
|
- **批量设备测试**: 验证 PLC Done → 所有 Occupied → Done
|
|||
|
|
- **检测设备测试**: 验证 pass/NG 信号 → 正确事件发布
|
|||
|
|
- **恢复测试**: RESTORE 消息恢复各状态槽位
|
|||
|
|
- **集成测试**: Actor + SignalRouter + EventLoop 联调完整链路
|