pull() to collect a single column

[krlmlr]

For regular data sources, I can do the following to obtain a count as an atomic value:

> data_frame(1:3) %>% tally %>% as.numeric
[1] 3

For SQL sources, I have to insert a collect():

> memdb_frame(a=1:3) %>% tally %>% as.numeric
Error in magrittr(.) : (list) object cannot be coerced to type 'double'
3: traceback(1)
2: magrittr(.) at pipe.R#6
1: memdb_frame(a = 1:3) %>% tally %>% as.numeric
> memdb_frame(a=1:3) %>% tally %>% collect %>% as.numeric
[1] 3

Perhaps we can implement as.numeric.tbl_lazy() and others?

[hadley]

I think we should implement pull() from lplyr: https://github.com/paulponcet/lplyr. It could default to the first column.

[krlmlr]

What's special about the first column? Often you want to pull data after applying mutate(), which adds columns at the end.

The verb itself could also have more general mutate semantics: iris %>% pull(Sepal.Length * Sepal.Width) .

[hadley]

It's just a reasonable default because you often have a single column data frame.

I think I like the idea of allowing it to do inline mutation too.

[krlmlr]

If the last column is the default, it will still work for one-column data frames.

[hadley]

Hmmmm, that just somehow feels less predictable to me. It probably doesn't really matter if we support inline mutate.