R 两个 SummarizedExperiment 对象的比较

Question

我的单元测试在比较引用和预期的 SummarizedExperiment 对象时失败。错误信息:

> expect_identical(target, current)
Error: `target` not identical to `current`.
Attributes: < Component “assays”: Class definitions are not identical >

MWE:

LINK TO DATASET

代码:

load("se-comparison.Rdata")
library(SummarizedExperiment)
library(testthat)

expect_identical(target, current)
# expect_identical() uses attr.all.equal() to compare S4 objects so check this
attr.all.equal(target, current)
# ok, check the attributes
cur <- attributes(current)
tar <- attributes(target)

class(cur$assays)
class(tar$assays)

expect_identical(
  cur$assays,
  tar$assays
)

expect_identical(
  class(cur$assays),
  class(tar$assays)
)

输出:

> library(SummarizedExperiment)
> library(testthat)
> expect_identical(target, current)
Error: `target` not identical to `current`.
Attributes: < Component “assays”: Class definitions are not identical >
> # expect_identical() uses attr.all.equal() to compare S4 objects so check this
> attr.all.equal(target, current)
[1] "Attributes: < Component “assays”: Class definitions are not identical >"
> # ok, check the attributes
> cur <- attributes(current)
> tar <- attributes(target)
> class(cur$assays)
[1] "ShallowSimpleListAssays"
attr(,"package")
[1] "SummarizedExperiment"
> class(tar$assays)
[1] "ShallowSimpleListAssays"
attr(,"package")
[1] "SummarizedExperiment"
> cur$assays
Reference class object of class "ShallowSimpleListAssays"
Field "data":
List of length 1
names(1): counts
> tar$assays
Reference class object of class "ShallowSimpleListAssays"
Field "data":
List of length 1
names(1): counts
> expect_identical(
+   cur$assays,
+   tar$assays
+ )
Error: cur$assays not identical to tar$assays.
Class definitions are not identical
> expect_identical(
+   class(cur$assays),
+   class(tar$assays)
+ )
>

有什么想法为什么比较失败？

Answer 1

不确定为什么它通常会失败，但 attr.all.equal 会比较 getClass 方法返回的类定义(如果可用)。在这种情况下，每个 $assays 对象的 getClass 返回在 fieldPrototypes 和 refMethods 中不同的类定义(环境字符串不同)。这会导致identical失败。

比较:

str(class(cur$assays))
str(cur$assays$getClass())

解决方法是跳过检查属性:expect_equal(target, current, check.attributes = FALSE)。

作为旁注，这也可以正常工作(请注意，我在这里使用 SummarizedExperiment 类的 getter 方法而不是 $assays):

expect_equal(
  assays(current),
  assays(target)
)