In the r database, it is easy to extract the names of columns (variables) from the data frame
> testdf <- data.frame(a1 = rnorm(1e5), a2 = rnorm(1e5), a3 = rnorm(1e5), a4 = rnorm(1e5), a5 = rnorm(1e5), a6 = rnorm(1e5))
> names(testdf)
[1] "a1" "a2" "a3" "a4" "a5" "a6"
but when using sparklyr things get complicated. After copying the data frame to the spark window,
> testdf_tbl <- copy_to(sc, testdf, overwrite = TRUE)
> names(testdf_tbl)
[1] "src" "ops"
variable names are actually inside "ops"
> testdf_tbl$ops$vars
[1] "a1" "a2" "a3" "a4" "a5" "a6"
and if all this were, there would be no problem (and no need to ask this question). But each time the operation is performed on testdf_tbl, the column / variable names change their position, as shown below.
> testdf_tbl <- testdf_tbl %>% select(-a1)
> testdf_tbl$ops$vars
NULL
> testdf_tbl$ops$x$vars
[1] "a1" "a2" "a3" "a4" "a5" "a6"
other operations add another $ x path to the path, etc.
> testdf_tbl <- testdf_tbl %>% select(-a2)
> testdf_tbl$ops$x$vars
NULL
> testdf_tbl$ops$x$x$vars
[1] "a1" "a2" "a3" "a4" "a5" "a6"
To make matters worse, the list of variables does not reflect the selection operations that we did, they still list a1, a2 as column names. where as
> head(testdf_tbl)
Source: query [?? x 4]
Database: spark connection master=local[24] app=sparklyr local=TRUE
a3 a4 a5 a6
dbl dbl dbl dbl
1 -1.146368875 1.691698406 0.43231629 1.3349111
2 0.664928710 -1.332242020 0.05380729 1.0139253
3 1.158095695 -0.097098980 -0.61885204 0.1504693
4 0.001595841 -0.003765908 0.27935192 -0.3039085
5 -0.133446040 0.269329076 1.57210274 1.7762602
6 0.006468698 -1.300439537 0.74057307 0.1320428
it’s clear that the selection operations affected the conditions for using the frame block.
SURELY, / sparklyr, a la names() r.