如何在RSpec中优雅实现带不同预期的循环测试？

问题 #

我有三个模型测试（model specs）代码逻辑相似，仅预期的数值结果略有不同。为避免重复代码，尝试遵循DRY原则实现：

tests = [
  {test_name: "this-year", arrival_date: 2.days.ago,    ytd: 1, ly: 0},
  {test_name: "last-year", arrival_date: 14.months.ago, ytd: 0, ly: 1},
  {test_name: "very old",  arrival_date: 3.years.ago,   ytd: 0, ly: 0},
]

tests.each do |t|
  it "increments #{t[:test_name]} order correctly" do
    expect{
      order = create(:order, buyer: @bs.buyer, seller: @bs.seller, arrival_date: t[:arrival_date])
      @bs.reload
    }.to change{ @bs.orders_count }.from(0).to(1)
    .and change{ @bs.orders_count_ytd }.from(0).to(t[:ytd])
    .and change{ @bs.orders_count_ly }.from(0).to(t[:ly])
  end
end

执行时抛出错误：

[Test 1]: expected `@bs.orders_count_ly` to have changed from 0 to 0, but did not change
[Test 2]: expected `@bs.orders_count_ytd` to have changed from 0 to 0, but did not change
[Test 3]: expected `@bs.orders_count_ytd` to have changed from 0 to 0, but did not change
            ...and:
          expected `@bs.orders_count_ly` to have changed from 0 to 0, but did not change

已知的解决方案需要添加冗余的条件判断来调整预期，但会破坏方案的简洁性与清晰度。请问是否有符合规范的优雅实现方式？

优雅解决方案 #

方法1：动态构建断言链 #

核心是只对实际需要变化的字段添加验证，跳过0→0的无效断言：

tests.each do |t|
  it "increments #{t[:test_name]} order correctly" do
    # 先定义必有的总数变化断言
    expectation = expect {
      create(:order, buyer: @bs.buyer, seller: @bs.seller, arrival_date: t[:arrival_date])
      @bs.reload
    }.to change{ @bs.orders_count }.from(0).to(1)

    # 仅当预期值非0时，添加对应字段的变化断言
    expectation = expectation.and(change{ @bs.orders_count_ytd }.from(0).to(t[:ytd])) if t[:ytd] != 0
    expectation = expectation.and(change{ @bs.orders_count_ly }.from(0).to(t[:ly])) if t[:ly] != 0
  end
end

这种方式既保留了DRY特性，又避免了无效验证，代码简洁性几乎不受影响。

方法2：自定义RSpec匹配器（复用性更强） #

如果这类场景频繁出现，封装自定义匹配器能让测试代码更直观：

RSpec::Matchers.define :update_order_counts_to do |expected|
  match do |block|
    block.call
    expect(subject.orders_count).to eq(expected[:total])
    expect(subject.orders_count_ytd).to eq(expected[:ytd])
    expect(subject.orders_count_ly).to eq(expected[:ly])
    true
  end

  chain :from do |original|
    @original = original
    # 可选：添加初始值验证
    expect(subject.orders_count).to eq(original[:total])
    expect(subject.orders_count_ytd).to eq(original[:ytd])
    expect(subject.orders_count_ly).to eq(original[:ly])
  end
end

# 测试代码简化为：
tests.each do |t|
  it "increments #{t[:test_name]} order correctly" do
    expect {
      create(:order, buyer: @bs.buyer, seller: @bs.seller, arrival_date: t[:arrival_date])
      @bs.reload
    }.to update_order_counts_to(total: 1, ytd: t[:ytd], ly: t[:ly]).from(total: 0, ytd: 0, ly: 0)
  end
end

自定义匹配器把复杂断言逻辑封装起来，测试代码可读性拉满，完全符合DRY原则，也不会出现无效验证问题。

方法3：拆分测试关注点（定位问题更精准） #

如果希望每个测试的职责更清晰，可以把通用断言和个性化断言分离：

shared_examples "order count updates" do |arrival_date, ytd, ly|
  it "increments total order count" do
    expect {
      create(:order, buyer: @bs.buyer, seller: @bs.seller, arrival_date: arrival_date)
      @bs.reload
    }.to change{ @bs.orders_count }.from(0).to(1)
  end

  it "sets correct ytd count" do
    create(:order, buyer: @bs.buyer, seller: @bs.seller, arrival_date: arrival_date)
    @bs.reload
    expect(@bs.orders_count_ytd).to eq(ytd)
  end

  it "sets correct ly count" do
    create(:order, buyer: @bs.buyer, seller: @bs.seller, arrival_date: arrival_date)
    @bs.reload
    expect(@bs.orders_count_ly).to eq(ly)
  end
end

tests.each do |t|
  describe "#{t[:test_name]} order" do
    include_examples "order count updates", t[:arrival_date], t[:ytd], t[:ly]
  end
end

这种方式把每个计数的验证拆成独立测试，虽然代码行数略多，但测试失败时能精准定位问题，可读性更强。

内容的提问来源于stack exchange，提问作者sscirrus