R for Excel Users

Re:R for Excel Users by _Wintermute on Thursday February 16, 2017 @12:31PM (#13661805)

That's a nice tutorial for beginners, though sticking dplyr code in functions is skirting dangerously close to the world of hurt that is non-standard evaluation.

The example:

    join_and_summarize <- function(df, colour_df){
        left_join(df, colour_df, by = "cyl") %>%
        group_by(colour) %>%
        summarize(mean_displacement = mean(disp))
    }

Will go really badly wrong if someone following the tutorial simply replaces the `disp` with a function argument.

[ Parent | on HN ]

Re:R for Excel Users by gordon_shotwell on Thursday February 16, 2017 @03:00PM (#13663072)

That's a good point, I'm not exactly sure what the best practice is for using NSE vs. SE functions within functions. My practice is to just use the NSE version unless I have a good reason not to. Generally my goal with teaching beginners to use R is to just not be afraid of functions as the barrier tends to be feeling like they can use them at all rather than using them in the safest way.

[ Parent | on HN ]

Re:R for Excel Users by hadley on Thursday February 16, 2017 @03:33PM (#13663330)

Yeah, NSE sucks in dplyr, but it'll be fixed soon :)

[ Parent | on HN ]

Re:R for Excel Users by dandermotj on Thursday February 16, 2017 @12:51PM (#13661975)

The standard evaluation versions of all dplyr functions are available, just add an underscore to the end: filter_, select_, ...

[ Parent | on HN ]

Re:R for Excel Users by _Wintermute on Thursday February 16, 2017 @01:01PM (#13662054)

I'm of the opinion that non-standard evaluation probably isn't the best introduction for converting people from excel to R. As nice as dplyr's syntax is for beginners you can quickly get into the whole ugly side of R.

Also it's often not as simple as just prefixing an underscore, you end up reading the lazyeval vignette and messing about with quote() and formula syntax.

[ Parent | on HN ]

Re:R for Excel Users by minimaxir on Thursday February 16, 2017 @12:58PM (#13662032)

A cool feature of select_ I discovered accidently is that you can pass in a vector of column names to get the corresponding columns, which is much easier than abusing which() to lookup indices.

[ Parent | on HN ]

Re:R for Excel Users by _Wintermute on Thursday February 16, 2017 @01:11PM (#13662139)

Not sure if I'm misunderstanding, but you've always been able to slice on a vector of column names.

    cols <- c("colA", "colB")
    dataframe[, cols]

[ Parent | on HN ]

Re:R for Excel Users by minimaxir on Thursday February 16, 2017 @01:13PM (#13662154)

True, although not as a part of a dplyr chain, which is cleaner.

[ Parent | on HN ]

Re:R for Excel Users by disgruntledphd2 on Thursday February 16, 2017 @01:48PM (#13662444)

extract and extract2 from magrittr also work. I do normally use dplyr though.

[ Parent | on HN ]