Data not exported from namespace in R

Question

I've set up and been regularly updating my R package to GitHub following Hadley's extensive documentation about Devtools, Roxygen2 etc., on my laptop. Yesterday I decided to use my main PC instead and am now looking to push changes up to GitHub. I got the following error after entering document() :

Error: 'Adult_Females' is not an exported object from 'namespace:gbm.auto'

Adult_Females is the name of the first data file in /Data. According to this (scroll down to 'Data')

"files that live in data/ don’t use the usual namespace mechanism and don’t need to be exported."

So... what's a guy to do? I've not edited Adult_Females in any way and the R script I edited doesn't reference it. My suspicion is that this error will pop up for all the data files and it just happened that this is the first of them, but that's conjecture at this point.

Thanks in advance. install_github("SimonDedman/gbm.auto") if you want to have a look.

2020/01/25 edit: looks like I've fixed it. A commit on 26/11/19 saw /Data and all RData files added, with subsequent commit deleting the identical /data folder and files. Not sure if I did that myself, can't think why I would suddenly decide to, but such mysteries are now lost to the sands of time. This change and document() and commit caused the .R files to be removed as export()s from NAMESPACE and the RData files to no longer trigger the titular "data not exported" problem EVEN THOUGH this isn't noted anywhere in DESCRIPTION, NAMESPACE, nor the files themselves. May this weirdness be a lighthouse that warns others of the rock I've spent the last 3 years trapped on!

Anacoluthia

It looks like you are using gbm.auto::Adult_Females in a couple of places. Did you try dropping the gbm.auto::? – Rhymester

Hi, thanks. Yes in @examples in gbm.cons.R I have "Adult_Females <- gbm.auto::Adult_Females"; will the data load for users if I don't specify the "::" since data shouldn't be explicitly exported/named? This is defo an area where my understanding of how everything github/roxygen/namespace/description fits together, falls down! – Anacoluthia

I'm not positive, but just using Adult_Females should work fine. :: is intended to be used with exported objects specifically, so I believe prepending gbm.auto:: is what is causing the error. – Rhymester

hmm. I've just removed all gbm.auto::Adult_Females references in gbm.cons.R and gbm.todoP2.R (a working doc) and no change. Adult_Females only appears within files in those two plus data.R. Could it be that since it's named in data.R as "#' @name Adult_Females" it's expected to be exported to the namespace? That would seem to fly in the face of what Hadley's said in his documentation... – Anacoluthia

:: doesn't work for unattached packages and lazy-loaded data. – Emery

Answer 1

I encountered a similar problem when writing an R package which contains a dataset. I guess you must have saved the dataset in a different name.

For example, you may write:

devtools:::use_data(YourDataSetName, pkg = "Path_to_Pkg/data", internal = FALSE)

but in your data.R file, you specified a dataset name at the very end other than YourDataSetName (suppose you followed Hadley's instructions here: http://r-pkgs.had.co.nz/data.html).

Make sure the data object you saved to the "data" folder is the same as you specified in your data.R file.

---

Note: use_data is now part of usethis package.

Scrubby

This must have worked at the time however I've now got the same issue again, out of nowhere. File is /Data/Adult_Females.RData. load() object result name is Adult_Females. /R/data.R object name is "Adult_Females", @name Adult_Females. I have 6 data files sequentially included in data.R; Hadley & you & others all discuss data.R as 1 file but you say there'll be trouble if I specify a different dataset name at the end, implying I should have 6 data.R files? – Anacoluthia

Answer 2

for data objects, the names must match in four (4) places, so check them all:

1. name of data/foo.rda file
2. name of object in data/foo.rda file
3. name of R/foo.R file
4. string at end of R/foo.R file

all four items must match--in this case 'foo'. If you change the name of the foo.rda and foo.R files, say to bar.rda and bar.R, it's easy to forget to rename the object in the .rda file from 'foo' to 'bar.' It's usually easiest to load the file, rename the object, and save the file under the new name:

load('data/foo.rda')
bar <- foo
save(bar, file='data/bar.rda')

If you don't do this, you get an unhelpful error about the object not loaded from namespace. You DON'T need to @export data objects, so instead ensure the names match in all places.

Renewal

Thanks. Re: 3 & 4: I don't have my individual data files as .R files in /R folder, but instead as one file, data.R, with details of the 6 datasets in my /data folder. I must have done this based on instructions on Hadley's site somewhere, is it wrong? The naming of that dataset is correct in data.R – Anacoluthia

I guess that's fine too... In that case I guess 3 isn't necessary, but I know 1 & 2 are, and I suppose 4 is only necessary for the docs. – Renewal

Thanks for this 0mn1. Am I missing something? I have this for 'Adult_Females' in my package. 1: /Data/Adult_Females.RData 2: load() object result is named Adult_Females 3&4: /R/data.R object name "Adult_Females", @name Adult_Females. Could it be they now need to be RDa not RData? It used to work fine... Cheers! – Anacoluthia

I believe RData is correct; at least according to the R packages book. Possibly that Data should be data (not capitalized.) Did you generate the package folder with devtools? Doing so would ensure the package structure is correct. – Renewal

I believe so but it was so long ago that I can't recall. I'll have a check tonight in any case. – Anacoluthia

Coming back to this as I never fixed it; 1 & 2 are correct (albeit as RData files in /Data), but you (point 4) and alittleboy allude to the dataset name specified AT THE END of data.R. I have 6 datasets specified in data.R and the last one isn't Adult_Females... but only 1 can be the last, so presumably all others will fail? How does one include >1 datasets in this case? – Anacoluthia

Thank you so much for this answer! The save() behavior which also stores the name of the R-object is just such a huge obstacle!! – Unrestraint

@Anacoluthia - at the end of data-raw/datasets.R (if you're using internal data), you list all your datasets in one line with a single usethis::use_data call. `usethis::use_data(dataset_a, dataset_b, ..., internal = TRUE) – Misprint

Answer 3

I used save() to create the .rda in the /data folder and added the oxygen headers in the R/data.R file. I was trying all the solutions above but was still getting the error.

What solved it for me was using usethis::use_data(MY_DATA) as suggested here: https://r-pkgs.org/data.html

Halidom

Likewise. It's as if use_data is doing something else we are unaware of. – Tyeshatyg