r - PowerPC 和 SPARC(均为大端)上的 data.table 不正确行为

Question

有一个data.table dt，我在其中使用cut做了一些分类:

require(data.table)
set.seed(1)
dt <- data.table(x = rnorm(10))
dt[, y := cut(x, breaks = c(-Inf, 0, Inf), labels = 1:2)]

如果我将结果因子 y 转换为数值(使用基于 ?factor 的函数 as.Numeric)，二分查找不再起作用，尽管 z 是数字。

as.Numeric <- function(f){
  as.numeric(levels(f))[f]
}

dt[, z := as.Numeric(y)] # as.numeric(as.character(y))
                                    # is working ...
dt
##              x y z
##  1: -0.6264538 1 1
##  2:  0.1836433 2 2
##  3: -0.8356286 1 1
##  4:  1.5952808 2 2
##  5:  0.3295078 2 2
##  6: -0.8204684 1 1
##  7:  0.4874291 2 2
##  8:  0.7383247 2 2
##  9:  0.5757814 2 2
## 10: -0.3053884 1 1

setkey(dt, z)
dt
##              x y z
##  1:  0.1836433 2 2
##  2:  1.5952808 2 2
##  3:  0.3295078 2 2
##  4:  0.4874291 2 2
##  5:  0.7383247 2 2
##  6:  0.5757814 2 2
##  7: -0.6264538 1 1
##  8: -0.8356286 1 1
##  9: -0.8204684 1 1
## 10: -0.3053884 1 1

dt[J(1)] # doesn't work
##     x  y z
## 1: NA NA 1

dt[y == 1, ] # works fine
##             x y z
## 1: -0.6264538 1 1
## 2: -0.8356286 1 1
## 3: -0.8204684 1 1
## 4: -0.3053884 1 1

str(dt)
## Classes ‘data.table’ and 'data.frame':   10 obs. of  3 variables:
##  $ x: num  0.184 1.595 0.33 0.487 0.738 ...
##  $ y: Factor w/ 2 levels "1","2": 2 2 2 2 2 2 1 1 1 1
##  $ z: num  2 2 2 2 2 2 1 1 1 1
##  - attr(*, ".internal.selfref")=<externalptr> 
##  - attr(*, "sorted")= chr "z"

尝试再次设置 key ，没有帮助:

setkey(dt, z)
## Warning message:
## In setkeyv(x, cols, verbose = verbose) :
##   Already keyed by this key but had invalid row order, key rebuilt. If you didn't go under the hood please let datatable-help know so the root cause can be fixed.

dt
##              x y z
##  1:  0.1836433 2 2
##  2:  1.5952808 2 2
##  3:  0.3295078 2 2
##  4:  0.4874291 2 2
##  5:  0.7383247 2 2
##  6:  0.5757814 2 2
##  7: -0.6264538 1 1
##  8: -0.8356286 1 1
##  9: -0.8204684 1 1
## 10: -0.3053884 1 1

dt[J(1)] # doesn't work
##     x  y z
## 1: NA NA 1

矢量扫描 正在运行，因为不需要 key 。使用 as.numeric(as.character(y)) 也可以。也许 [-operator in as.Numeric 是个问题？使用与 dt 1.8.10 相同的代码，一切都像预期的那样。要找出代码在 1.9.3 中不再工作的原因并不是那么容易...

问题:

这是一个错误吗？

附:

sessionInfo()
## R version 3.1.0 (2014-04-10)
## Platform: powerpc64-unknown-linux-gnu (64-bit)

## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

## attached base packages:
## [1] splines   grid      stats     graphics  grDevices utils     datasets 
## [8] methods   base     

## other attached packages:
##  [1] plyr_1.8.1        reshape2_1.4      lubridate_1.3.3   HP14int_0.1-15   
##  [5] Hmisc_3.14-4      Formula_1.1-1     survival_2.37-7   lattice_0.20-29  
##  [9] HP14unidata_1.1-1 data.table_1.9.3 

## loaded via a namespace (and not attached):
## [1] cluster_1.15.2      digest_0.6.4        latticeExtra_0.6-26
## [4] memoise_0.2.1       RColorBrewer_1.0-5  Rcpp_0.11.1        
## [7] stringr_0.6.2       tools_3.1.0

writeLines(paste("Endianess:", .Platform$endian))
## Endianess: big

Answer 1

现已修复 v1.9.5 on GitHub .感谢您的报告。

Compatibility with big endian machines (e.g., SPARC and PowerPC) is restored. Most Windows, Linux and Mac systems are little endian; type .Platform$endian to confirm. Thanks to Gerhard Nachtmann for reporting and the QEMU project for their PowerPC emulator.