jhudsl / matahari Goto Github PK

As mentioned in an issue that I raised in the repo for tidycode, it seems that matahari has a hard time spying on functional code. Any idea how to better spy when someone e.g. does purrrr::map(df, some_function) ? Currently, only the call to purrr::map is caught, meaning that the person could hide anything behind some_function...

I am trying to use dance_report() in RStudio Cloud, but it doesn't play nicely with the clipboard 😢. Would you be okay with this change? If so, I can submit a PR.

copy_base64 <- function(clip = TRUE) {
  enc_string <- base64_enc(serialize(dance_tbl(), NULL))
  if (clip) {
   write_clip(enc_string)
  } else {
   enc_string
 }
} 

I am sure this is something weird on my end, but when I run the code in the README, my output is different (in fact I can't seem to populate the value column with anything except sessionInfo for the first and last column, and the rest of the columns with the exception of expr are all NA). I am using RStudio, running this inside a project.

dance_start()

4 + 4
#> [1] 8
"wow!"
#> [1] "wow!"
mean(1:10)
#> [1] 5.5

# Pause logging
dance_stop()
dance_tbl()
#> # A tibble: 6 x 6
#>   expr       value             path      contents  selection dt                 
#>   <list>     <list>            <list>    <list>    <list>    <dttm>             
#> 1 <language> <S3: sessionInfo> <lgl [1]> <lgl [1]> <lgl [1]> 2018-06-15 13:55:55
#> 2 <language> <lgl [1]>         <lgl [1]> <lgl [1]> <lgl [1]> 2018-06-15 13:55:55
#> 3 <language> <lgl [1]>         <lgl [1]> <lgl [1]> <lgl [1]> 2018-06-15 13:55:55
#> 4 <chr [1]>  <lgl [1]>         <lgl [1]> <lgl [1]> <lgl [1]> 2018-06-15 13:55:55
#> 5 <language> <lgl [1]>         <lgl [1]> <lgl [1]> <lgl [1]> 2018-06-15 13:55:55
#> 6 <language> <S3: sessionInfo> <lgl [1]> <lgl [1]> <lgl [1]> 2018-06-15 13:55:55

👆 Notice here all of my value column are <lgl[1]> (which are actually NA) compared to the readme where they differ based on what the value output would have been.

Context: I am trying to use this for a project with Jeff so trying to figure out the 📦 🎉

Instead create a new environment inside your package:

env <- new.env(parent = emptyenv())

And then use that env instead of globalenv():

dance_remove <- function() {
  if (!exists("dance", envir = env)
    return(invisible(FALSE)
      
  rm("dance", envir = env)
  invisible(TRUE)
}

@seankross, I'm adding a few issues to beef this up before submitting the paper we've been working on 🎉

Some of these are based on this: https://github.com/jtleek/rpackages

If it is 👌 with you I can go ahead and make these changes and submit as PRs for you to 👀 at

data_frame() and as_data_frame() are now deprecated, changing to the tibble friends. (I LOVE THAT TIBBLE AUTOCORRECTS TO NIBBLE, it always makes me laugh 😂)

This code is creating an issue where you need to run dance_remove() in order to run dance_start() again (which I don't think we intended?) before you could just run

dance_start()
dance_stop()

@seankross, would it make you sad if I refactor this a bit to remove the dplyr dependency? I know dplyr isn't as heavy as it once was, but I've found that people get persnickety about picking up new packages that depend on it 🤷‍♀ BUT if you would rather it stays, I am fine with that too.

It looks like this changed it when evaluate = FALSE:

• cee3c77

but it is still result when evaluate = TRUE?

• matahari/R/dance.R

  Line 174 in 4c73ad4

  mutate(result = map(.data$result, ~ .x$result))

Thanks for a great package, I wanted to pass on that I don't find the function names super intuitive ("dance" is not the first thing that comes to mind when I think of automated code analysis). I get that this is related to the package name, but it might be worth changing the names or aliasing them to something more memorable.

Thanks for creating a great package.

I am currently using R Studio Cloud with middle grades-aged students. I am curious about whether I could use matahari to understand what code students write in a document. Apart from asking them to load matahari and to run the dance_start() and dance_stop() commands manually (assuming they have shared their project with me), is there any way to do this in an (even) more seemless way?

I am wondering about something like sourcing an R file that loads matahari and then runs dance_start(), but am not sure how to get the tracking to end without students manually running dance_end(). I know this question is a bit open-ended; thanks for considering this.

Thanks for creating a very useful package for logging executed expressions in RStudio!

I did come across a limitation I wanted to point out. If one only relied on dance_tbl(), the resulting table does not include code that produces an error, along with the error message. An example of executing on the console with error-producing code:

> dance_start()
> 2 + "foo"
Error in 2 + "foo" : non-numeric argument to binary operator
> dance_tbl()[['expr']]
[[1]]
sessionInfo()

[[2]]
dance_start()

The 2 + "foo" is not included.

However, this same ability is present if using dance_recital():

> dance_recital("2 + 'foo'")[['expr']]
[[1]]
2 + "foo"
> dance_recital("2 + 'foo'")[['error']]
[[1]]
<simpleError in 2 + "foo": non-numeric argument to binary operator>

I like the passive logging flow and was expecting it to also capture errors but it doesn't seem to be the case. Please let me know if there's a way to achieve it with dance_tbl() that I may have missed. Otherwise, it would be useful to include problematic code + error for dance_tbl().