Disassembly Order Sync from JikeYun to Kingdee Flagship Edition: A Single-Strategy Implementation Guide
What This Strategy Solves
After a retail company ran an e-commerce mid-platform (JikeYun) alongside a financial/inventory back office (Kingdee Flagship Edition), a familiar pain point emerged: warehouses performed disassembly orders in JikeYun, breaking a bundle into sub-components and adjusting stock, but Kingdee's inventory and costs stayed frozen. A few days later, finance closed the books and found the books did not match reality.
This disassembly order sync strategy solves exactly that: push JikeYun disassembly orders into Kingdee Flagship Edition using the same business semantics, generate the corresponding disassembly documents, and update inventory, so that quantities, costs, and document trails stay aligned on both sides. On the customer site we used the Qeasy data integration platform to host it as an independent, monitored, replayable sync strategy.
Data Flow and Field Mapping
The flow is unidirectional: JikeYun (source) → Qeasy middle layer → Kingdee Flagship Edition (target). The middle layer handles extraction, cleansing, mapping, and push.
Key field mapping (semantics only, after desensitization):
| Business Meaning | JikeYun Source Field | Middle Layer | Kingdee Target Field |
|---|---|---|---|
| Document number | billNo | Pass-through | FBillNo |
| Business date | billDate | yyyy-MM-dd normalization | FDate |
| Warehouse | stockName | Code mapping lookup | FStockID |
| Bundle material | materialCode + skuName | Code mapping | FMaterialID (header) |
| Sub-component lines | details[].sku / qty / unit | Expanded to multiple lines | FEntity (FMaterialID/FQty/FUnitID) |
| Remark | remark | Pass-through | FNote |
| Document status | status | State machine conversion | FStatus |
Worth highlighting: warehouses and materials use different identifiers on each side — JikeYun uses names ("Hangzhou Main Warehouse"), Kingdee uses internal codes (FStockID). The two systems are completely incompatible, so code mappings must be maintained centrally rather than scattered across every strategy. This is one of the common patterns we see among Qeasy customers: lift the mapping tables into a shared "base data" layer referenced by all strategies.
How to Configure It on Qeasy
In the Qeasy console, this strategy is typically configured around these points:
- Datasource registration: register JikeYun and Kingdee Flagship Edition separately (private deployment uses an internal address; tokens are stored securely by the platform).
- Source extraction: configure the JikeYun disassembly order query endpoint, pulling increments where
last_modified >= last_success_time; keep a full-reconciliation fallback channel. - Field mapping: in Qeasy's mapping canvas, drag source fields to target fields and attach transformation functions (date formatting, numeric precision, string trimming) directly on cells. Warehouses and materials reference mapping tables, never hard-coded in scripts.
- Target write: call Kingdee's disassembly order save endpoint, submitting header and lines in two stages — this avoids payload truncation and lets retries target only the lines.
- Exception handling: configure retry count, failure alerts (DingTalk / WeCom), and a dead-letter queue.
- Logging: enable end-to-end tracing by source document number so a single failed sub-component can be located in one click.
Implementation Steps
We recommend a three-phase rollout rather than going live at high frequency immediately:
- Phase 1 — Incremental starting point. Manually trigger a few sample documents by bill number to confirm mapping correctness and inventory update direction. No schedule yet.
- Phase 2 — Full reconciliation. Manually run a one-time full backfill for historical disassembly orders. After completion, reconcile inventory on both sides, and only then enable scheduling.
- Phase 3 — Scheduling. Disassembly volume is usually low; we recommend a dual-track pattern: incremental every 15 minutes plus a nightly full validation. The "incremental + full dual-track" pattern is common among Qeasy customers — incremental guarantees freshness, full reconciliation catches anything missed.
Pitfall Recap
- Pitfall 1: warehouse code embedded in scripts. On one project, warehouse mapping was hard-coded in a transformation script; when JikeYun renamed a warehouse, every dependent strategy broke overnight. The safe practice is centralized code mapping — a shared mapping table referenced by every strategy.
- Pitfall 2: header and lines in one request. Kingdee's endpoint is sensitive to payload size; too many sub-component lines cause a 500. Submit header and lines in stages and retry only the lines on failure.
- Pitfall 3: state machines misaligned. JikeYun's statuses (e.g., "Audited") don't map one-to-one to Kingdee's (Draft / Audited / Closed); passing them through verbatim can leave target documents in an incorrect state. A state-machine conversion in the middle layer is required.
- Pitfall 4: wrong field for the incremental cursor. Projects that use
created_timeas the incremental cursor almost always miss backdated entries and reverse modifications. The safe choice islast_modified_timeplus a bill-number deduplication. - Pitfall 5: private-deployment network jitter. When source and target sit in different rooms, large payloads occasionally time out. Qeasy's experience is to enable resumable transfer + asynchronous callbacks, avoiding a full rerun after a single failure.
Where It Fits and Where It Doesn't
Fits: retail/distribution scenarios with frequent bundle splitting, where the e-commerce platform performs physical disassembly and inventory must reflect immediately in the financial book; the dual-system architecture where JikeYun is the business front end and Kingdee Flagship Edition is the inventory and cost back office.
Doesn't fit: pure assembly business (use assembly orders instead); single-system closed loops; high-frequency automated production lines that need millisecond-level real-time inventory (those should go WMS-to-WMS directly rather than through an integration middleware).