Console list: the 「导入」 button is gated on allowCreate although the wizard can update, and 「导出」 emits a file the wizard refuses — the export→edit→import round trip is unreachable #16345

Measured on a real app (ObjectStack 17.2.0, console list view of a business object) while walking the intended end-user journey export → edit one column → import back. Two independent console behaviours make that journey unreachable for the role it is meant for.

现象 A — the list toolbar gates 「导入」 on create permission, but the import can update

An object whose permission set grants allowEdit: true and allowCreate: false shows 「导出」 in the list toolbar and no 「导入」 button at all. Measured on three objects for the same user in one session:

The button tracks allowCreate exactly. But the wizard itself is not create-only — its step 3 offers 「当某行匹配到已有记录时」 with 始终新建 / 更新已有(无匹配则跳过) / 匹配则更新,否则新建, and the update mode works: with match fields selected the run reports 全部 4 行均有效 → 导入完成 · 更新 4 条, and the collection row count is unchanged.

So an update-only import — the exact shape of "bulk-edit one column in Excel and put it back" — is gated behind a permission it does not need. The role that is supposed to do the data entry is the one role that cannot reach it; only an admin can.

Related but not the same: objectstack-ai/objectui#10102 (wizard hardcoding writeMode:"insert") is about the mapping payload; on 17.2.0 the wizard clearly does send update modes. This is the entry point being gated on the wrong verb.

现象 B — the list export omits the object's required lookup columns, so its own output cannot be imported back

Exporting the same list (导出 → 导出为 XLSX) emits exactly the view's visible columns:

指标名称 | 计量单位 | 目标值 | 权重(%) | 实际值 | 完成率(%) | 得分率(%) | 最终得分 | 已调整 | 备注

No record id, and neither of the object's two required lookups. Feeding that file straight back into the wizard auto-maps all 10 columns and then hard-stops on step 2:

无法继续——以下必填字段未映射:所属填报单 (sheet), 来源下达 (plan_indicator)。
请在文件中添加对应的列,或返回上一步重新上传包含该列的文件。

The 「下载模板」 CSV confirms the asymmetry — it carries 28 columns including the two required lookups, but no record id column either, so even the template offers no id-keyed update path; matching is only ever by business columns.

Adding the two lookup columns by hand (display values, e.g. 2026 年第 3 季度考核 · 华南分公司) makes the whole loop work in 9 clicks, and the row-level validation is good: an unmatched row reports 第 4 行:来源下达: 找不到匹配 "…" 的记录, a missing row is simply left untouched, and the run ends 更新 3 条 · 已跳过 1 条 with no stray inserts. So the machinery is sound — the export just does not emit a re-importable file.

最小复现

1. Define an object with at least one required lookup, and a list view whose columns do not include it.
2. Grant a role allowRead / allowEdit / allowExport but not allowCreate.
3. As that role, open the list — toolbar shows 「导出」, no 「导入」 (现象 A).
4. As an admin, 导出 → 导出为 XLSX, then 导入 and upload that exact file — step 2 blocks on the unmapped required lookups (现象 B).

期望能力

• A: gate the import entry point on the permission the chosen write mode actually needs — allowEdit for update, allowCreate for insert, both for upsert. Offering the button and letting the wizard restrict its write modes would be enough; silently hiding it from the edit-capable role is what makes the feature undiscoverable.
• B: make the export round-trippable — emit a stable key (record id, ideally as a column the wizard recognises as a match field) or at least the object's required lookups, so the file the console produces is a file the console accepts. Today the two halves of the same toolbar disagree about what a row is.

平台版本

@objectstack/* 17.2.0