如何在Gherkin+Cucumber+Cypress中合理规划用户清理测试流程

问题背景 #

我正在为专注于用户操作的WebUI应用（FreeIPA）编写测试，需测试多用户添加及删除功能。由于本地开发时后端持续运行不重启，需确保清理前次测试残留的用户。现有测试代码如下：

Background:
    Given I am logged in as "admin"
    Given I am on "active-users" page

  Scenario Outline: Add a new user
    When I click on "Add" button
    * I type in the field "User login" text "<userLogin>"
    * I type in the field "First name" text "<firstName>"
    * I type in the field "Last name" text "<lastName>"
    * I type in the field "New Password" text "<password>"
    * I type in the field "Verify password" text "<password>"
    * in the modal dialog I click on "Add" button
    Then I should see "<userLogin>" entry in the data table
    Examples:
      | userLogin | firstName | lastName | password             |
      | testuser1 | Arthur    | Dent     | ILoveKlingonPoetry42 |
      | testuser2 | Banana    | Bread    | FishAndChips097      |
      | testuser3 | Cypress   | Gateway  | TestingIsFun73       |


  Scenario: Delete a user
    Given I should see "testuser1" entry in the data table
    When I select entry "testuser1" in the data table
    And I click on "Delete" button
    Then I see "Remove active users" modal

    When in the modal dialog I check "Delete" radio selector
    And in the modal dialog I click on "Delete" button
    Then I should not see "testuser1" entry in the data table

我尝试过三种方案但均存在问题：

• 使用Before钩子：该钩子在Background前执行（管理员登录前），导致测试失败；
• 在Background块中加入"清理状态"步骤：该步骤会在每个场景前执行，导致Delete a user场景失败；
• 在Add a new user场景中添加"清理状态"步骤作为第一步：这会破坏多用户可见的测试目的。

请问应如何重构测试流程，让Add a new user场景从空用户表开始，同时保留添加的用户供Delete场景使用？若此方式违反耦合原则，是否可行在场景间复用场景？

解决方案 #

思路1：标签+针对性清理钩子（保留场景依赖） #

通过给场景打标签，让清理逻辑只在指定场景前触发，既保证添加用户场景从空表启动，又不干扰删除场景依赖已添加的用户。

1. 给添加用户场景添加标签：

@cleanup-before
Scenario Outline: Add a new user
    # 原有步骤不变

2. 调整Before钩子逻辑，确保在Background之后执行清理（不同测试框架实现略有差异）：
   • 如果框架支持钩子优先级，设置更高优先级让钩子在Background后运行：

     # 以Cucumber Ruby为例
     Before('@cleanup-before', :order => 2) do
       step("I delete all users with login starting with 'testuser'")
     end
     

   • 如果不支持优先级，改用API完成登录和清理（跳过WebUI操作，效率更高）：

     Before('@cleanup-before') do
       # 调用后台API完成管理员认证
       api_login("admin", "admin-password")
       # 调用API批量删除测试用户
       api_delete_all_test_users
     end
     

思路2：独立场景设计（避免耦合，推荐） #

场景间依赖会导致测试脆弱——一旦添加场景失败，删除场景也会跟着失效。更稳定的方案是让每个场景独立处理前置条件：

1. 重构添加用户场景：

Scenario Outline: Add multiple users and verify presence
    Given I clean up all test users
    When I add a user with login "<userLogin>", first name "<firstName>", last name "<lastName>" and password "<password>"
    Then I should see "<userLogin>" entry in the data table
    Examples:
      | userLogin | firstName | lastName | password             |
      | testuser1 | Arthur    | Dent     | ILoveKlingonPoetry42 |
      | testuser2 | Banana    | Bread    | FishAndChips097      |
      | testuser3 | Cypress   | Gateway  | TestingIsFun73       |

这里把添加用户的重复步骤封装成复合步骤I add a user with...，简化场景结构。

2. 重构删除用户场景：

Scenario: Delete an existing user
    Given I add a user with login "testuser1", first name "Arthur", last name "Dent" and password "ILoveKlingonPoetry42"
    And I should see "testuser1" entry in the data table
    When I select entry "testuser1" in the data table
    And I click on "Delete" button
    Then I see "Remove active users" modal

    When in the modal dialog I check "Delete" radio selector
    And in the modal dialog I click on "Delete" button
    Then I should not see "testuser1" entry in the data table

删除场景不再依赖其他场景，自行创建测试用户，测试独立性和稳定性大幅提升。

思路3：场景大纲复用逻辑（耦合性高，不推荐） #

如果必须保留场景间关联，可以在Background中先批量添加用户，再执行删除场景，但这种方式耦合性强，容易引发连锁失败：

Background:
    Given I am logged in as "admin"
    Given I am on "active-users" page
    Given I clean up all test users
    And I add a user with login "testuser1", first name "Arthur", last name "Dent" and password "ILoveKlingonPoetry42"
    And I add a user with login "testuser2", first name "Banana", last name "Bread" and password "FishAndChips097"

Scenario: Delete testuser1
    Given I should see "testuser1" entry in the data table
    # 原有删除步骤

Scenario: Delete testuser2
    Given I should see "testuser2" entry in the data table
    # 原有删除步骤

总结 #

• 若暂时需要保留场景依赖，优先用标签+针对性清理钩子的方案，满足需求的同时降低耦合；
• 从长期测试稳定性考虑，强烈推荐独立场景设计，每个场景自行处理前置条件；
• 场景间复用逻辑更适合通过封装复合步骤实现，而非直接依赖其他场景的执行结果。

内容的提问来源于stack exchange，提问作者Michal Polovka