spread_values cannot create column named "x"

[jucor]

Hi folks, thanks for a great package!

I noticed one bug though: spread_values crashes when creating a column named x:

'{ x: 3}' %>% spread_values(x = jnumber("x"))

returns:

Error in UseMethod("as.tbl_json") : 
  no applicable method for 'as.tbl_json' applied to an object of class "function"

whereas

'{ x: 3}' %>% spread_values(myx = jnumber("x"))

works seamlessly.

Is there aw workaround, please?
Thanks!

[colearendt]

The problem here is apparent if you look at the documentation: ?spread_values. The first parameter of spread_values is x. As a result, the pipe is getting overridden and you are, in essence calling spread_values(x=jnumber('x'),'{ x: 3}'). When using column name myx, the pipe is calling spread_values(x='{ x: 3}', myx=jnumber('x')).

To explicitly see the problem, try calling spread_values(x='{ x: 3}', x=jnumber('x')), as that is what you are trying to do. A simple workaround is to use dplyr::rename.

'{ "x": 3}' %>% 
spread_values(myx=jnumber('x')) %>% 
rename(x=myx)

[jucor]

Thanks @colearendt . I figured this was indeed the problem, and have worked around with such a rename [as my post suggested], but it feels very kludgy and un-clean, especially as "x" is quite a frequent field name. So I'll be more explicit in my feature request: is there a way to modify spread_values so that the name of its parameters is not mutually exclusive with the name of the fields?

[colearendt]

That makes sense. I think you may be interested in taking a look at the continued development of the package here. The changes have not yet made it to CRAN, but I have found the development version there to be superior to the CRAN version.

devtools::install_github('jeremystan/tidyjson')

In that version, spread_values is defined as spread_values(.x,...). Unfortunately, the first parameter still needs to be named, so that data can be passed to the function, but .x seems like a fair and much more reasonable solution, in my opinion. .x should be a less frequent name for a column, at least. The dplyr standard seems to be .tbl as the "column which cannot be named," so this is a similar (although different) convention.

[jucor]

That indeed sounds like a much less problematic name! Thanks for the type @colearendt .