coolbutuselessMar 23

#RStats Truth:

Nested data.frames which are *not* in list-columns are an abomination!

> df1 <- data.frame(x = 1:2)
> df2 <- data.frame(y = c('a', 'b'), z = c(1.1, 2.2))
> df1$nest <- df2

> names(df1)
[1] "x" "nest"

> df1
x nest.y nest.z
1 1 a 1.1
2 2 b 2.2
>

Show threadRyan McShane, Ph.D.Mar 23

@coolbutuseless preach

But also, {jsonlite} reads in .json that way 😬

Not sure there's a better way to read in .json files.

Show threadKonrad RudolphMar 23
@ryanmcshane @coolbutuseless … not guess the format, and preserve all objects as lists (unless the user explicitly specifies a fixed deserialisation format). Auto-guessing as the default is an antipattern.
Show threadRyan McShane, Ph.D.

@klmr @coolbutuseless aren't all data.frames lists? Just lists composed of vectors of equal length? Not sure of the disadvantage of having something as a data.frame column instead of a list column (temporarily!).

Between tidyr::unpack, purrr::unnest, purrr::pluck, and tidyr::hoist, there's something to handle every situation to reach a rectangled data.frame anyway (?)

Mar 23 at 1:17pmWeb
Show threadKonrad RudolphMar 23
@ryanmcshane @coolbutuseless The issue in this case is that lists in the JSON get deserialised as vectors, not unnamed lists, so you lose type information (scalars vs lists of length one), so you can’t losslessly round-trip your data.
Show threadJordi Rosell6d ago
@ryanmcshane data.frames are lists but not lists of lists :)