This vignette provides a quick tour of the osfr package.

library(osfr)

What is OSF?

OSF is a free and open source web application that provides a space for researchers to collaboratively store, manage, and share their research materials (e.g. data, code, protocols).

Most work on OSF is organized around projects, which include a cloud-based storage bucket where files can be stored and organized into directories. Note there is no storage limit on the size of projects but individual files must be < 5Gb. Projects can be kept private, shared with a specific group of collaborators, or made publicly available with citable DOIs so you can get credit for their work.

If you’d like to learn more about OSF the Center for Open Science has published an excellent series of guides to help get you started. We’ll provide links to specific guides throughout this vignette. Here are a few relevant topics:

Accessing OSF projects

Let’s check out an example project containing materials for an analysis of the 2012 American National Election Survey (ANES). You can access the OSF project in your browser by navigating to its URL: https://osf.io/jgyxm/.

Let’s load this project into R with osfr::osf_retrieve_node():

anes_project <- osf_retrieve_node("https://osf.io/jgyxm")
anes_project
#> # A tibble: 1 x 3
#>   name                                id    meta            
#>   <chr>                               <chr> <list>          
#> 1 Political identification and gender jgyxm <named list [3]>

This returns an osf_tbl object, which is the data.frame-like class osfr uses to represent items retrieved from OSF. You can now use anes_project to perform a variety of project related tasks by passing it to different osfr functions.

Downloading files

Let’s list all of the files that have been uploaded to the project:

anes_files <- osf_ls_files(anes_project)
anes_files
#> # A tibble: 5 x 3
#>   name                 id                       meta            
#>   <chr>                <chr>                    <list>          
#> 1 cleaning.R           5e20d22bedceab002d82e0f1 <named list [3]>
#> 2 Questionnaire.docx   5e20d22bedceab002b82dc3f <named list [3]>
#> 3 raw_data.csv         5e20d22c675e0e00096b4de8 <named list [3]>
#> 4 Data Dictionary.docx 5e20d22c675e0e000e6b4b18 <named list [3]>
#> 5 analyses.R           5e20d22c675e0e000a6b4bd3 <named list [3]>

This returns another osf_tbl but this one contains 5 rows; one for each of the project files stored on OSF. A nice feature of OSF is it provides rendered views for a wide variety of file formats, so it’s not necessary to actually download and open a file if you just want to quickly examine it. Let’s open the Word Document containing the project’s data dictionary by extracting the relevant row from anes_tbl and passing it to osf_open():

osf_open(anes_files[4, ])

Because osf_tbls are just specialized data.frames, we could also subset() or dplyr::filter() to achieve the same result.

Note: If an osf_tbl with multiple entities is passed to an non-vectorized osfr function like osf_open(), the default behavior is to use the entity in the first row and warn that all other entities are ignored.

We can also download local copies of these files by passing anes_files to osf_download().

osf_download(anes_files)
#> # A tibble: 5 x 4
#>   name                id                     local_path           meta          
#>   <chr>               <chr>                  <chr>                <list>        
#> 1 cleaning.R          5e20d22bedceab002d82e… ./cleaning.R         <named list […
#> 2 Questionnaire.docx  5e20d22bedceab002b82d… ./Questionnaire.docx <named list […
#> 3 raw_data.csv        5e20d22c675e0e00096b4… ./raw_data.csv       <named list […
#> 4 Data Dictionary.do… 5e20d22c675e0e000e6b4… ./Data Dictionary.d… <named list […
#> 5 analyses.R          5e20d22c675e0e000a6b4… ./analyses.R         <named list […

We’ll use these files in the next section for creating a new project.

Pipes

As you’ve likely noticed, osf_tbl objects are central to osfr’s functionality. Indeed, nearly all of its functions both expect an osf_tbl as input and return an osf_tbl as output. As such, osfr functions can be chained together using the pipe operator (%>%), allowing for the creation of pipelines to automate OSF-based tasks.

Here is a short example that consolidates all of the steps we’ve performed so far:

Project management

Now let’s see how to use osfr to create and manage your own projects. The goal for this section is to create your own version of the Political Identification and Gender project but with a better organizational structure. To follow along with this section you’ll need to authenticate osfr using a personal access token (PAT). See the ?osf_auth() function documentation or the auth vignette for more information.

Creating a project

First you will need to create a new private project on OSF to store all the files related to the project. Here, we’re giving the new project a title (required) and description (optional).

my_project <- osf_create_project(
  title = "Political Identification and Gender: Re-examined",
  description = "A re-analysis of the original study's results."
)
my_project
#> # A tibble: 1 x 3
#>   name                                             id    meta            
#>   <chr>                                            <chr> <list>          
#> 1 Political Identification and Gender: Re-examined kqy2p <named list [3]>

The GUID for this new project is kqy2p, but yours will be something different. You can check out the project on OSF by opening it’s URL (https://www.osf.io/<GUID>), or, more conveniently: osf_open(my_project).

Adding structure with components

A key organizational feature of OSF is the ability to augment a project’s structure with sub-projects, which are referred to as components on OSF. Like top-level projects, every component is assigned a unique URL and contains its own cloud-based storage bucket. They can also have different privacy settings from the parent project.

We are going to create two nested components, one for the raw data and one for the analysis scripts.

data_comp <- osf_create_component(my_project, title = "Raw Data")
script_comp <- osf_create_component(my_project, title = "Analysis Scripts")

# verify the components were created
# osf_open(my_project)

If you refresh the OSF project in your browser the Components widget should now contain two entries for each of our newly created components.

Uploading files

Now that our project components are in place we can start to populate them with files. Let’s start with the csv file containing our raw data.

data_file  <- osf_upload(my_project, path = "raw_data.csv")
data_file
#> # A tibble: 1 x 3
#>   name         id                       meta            
#>   <chr>        <chr>                    <list>          
#> 1 raw_data.csv 5e3c4af3032a4d00dee71f02 <named list [3]>

Oh no! Instead of uploading raw_data.csv to the Raw Data component, we uploaded it to the parent project instead.

Fear not. We can easily fix this contrived mistake by simply moving the file to its intended location.

data_file <- osf_mv(data_file, to = data_comp)

Crisis averted. Now if you open Raw Data on OSF (osf_open(data_comp)), it should contain the csv file.

Our next step is to upload the R scripts into the Analysis Scripts component. Rather than upload each file individually, we’ll take advantage of osf_upload()’s ability to handle multiple files/directories and use list.files() to identify all .R files in the working directory:

r_files <- osf_upload(script_comp, path = list.files(pattern = ".R$"))
r_files
#> # A tibble: 3 x 3
#>   name         id                       meta            
#>   <chr>        <chr>                    <list>          
#> 1 analyses.R   5e3c4af6f1369e01158aaf3b <named list [3]>
#> 2 cleaning.R   5e3c4af9032a4d00e8e70915 <named list [3]>
#> 3 precompile.R 5e3c4afaf1369e010e8acaf7 <named list [3]>

Putting it all together

Finally, let’s repeat the process for the 2 .docx file containing the survey and accompanying data dictionary. This time we’ll use a more succinct approach that leverages pipes to create and populate the component in one block:

my_project %>%
  osf_create_component("Research Materials") %>%
  osf_upload(path = list.files(pattern = "\\.docx$"))
#> # A tibble: 2 x 3
#>   name                 id                       meta            
#>   <chr>                <chr>                    <list>          
#> 1 Data Dictionary.docx 5e3c4afef1369e01108ad708 <named list [3]>
#> 2 Questionnaire.docx   5e3c4b00032a4d00ece6faab <named list [3]>

We can verify the project is now structured the way we wanted by listing the components we have under the main project.

osf_ls_nodes(my_project)
#> # A tibble: 3 x 3
#>   name               id    meta            
#>   <chr>              <chr> <list>          
#> 1 Research Materials hvmpr <named list [3]>
#> 2 Analysis Scripts   85tck <named list [3]>
#> 3 Raw Data           mzq49 <named list [3]>

which gives us an osf_tbl with one row for each of the project’s components.

Updating files

OSF provides automatic and unlimited file versioning. Let’s see how this works with osfr. Make a small edit to your local copy of cleaning.R and save. Now, if we attempt to upload this new version to the Analysis Scripts component, osfr will throw a conflict error:

osf_upload(script_comp, path = "cleaning.R")
Error: Can't upload file 'cleaning.R'.
  * A file with the same name already exists at the destination.
  * Use the `conflicts` argument to avoid this error in the future.

As the error indicates, we need to use the conflicts argument to instruct osf_upload() how to handle the conflict. In this case, we want to overwrite the original copy with our new version:

osf_upload(script_comp, path = "cleaning.R", conflicts = "overwrite")

Learn more about file versioning on OSF here.

Sharing

Remember, new OSF projects are always private by default. You can change this by opening the project’s settings page on OSF and making it public. See the following guides more information about OSF permissions and how to optionally generate a DOI so other can cite your project.

A few details about files on OSF

On OSF, files can exist in projects, components, and/or directories. Files can be stored on OSF’s Storage or in another service that is connected to an OSF project (e.g. GitHub, Dropbox, or Google Drive). However, osfr currently only supports interacting with files on OSF Storage.

We can download files from any public or private node that we have access to and can identify files to download in two different ways:

  1. If we know where the file is located, but don’t remember its GUID, you can use the osf_ls_files function to filter by filename within a specified node and then pipe the results to osf_download().

    anes_project %>%
      osf_ls_files(pattern = ) %>%
      osf_download(conflicts = "overwrite")
  2. For a public file that was referenced in a published article, you may already have the GUID, and so can retrieve the file directly before downloading it. For example, let’s download Daniel Laken’s helpful spreadsheet for calculating effect sizes (available from https://osf.io/vbdah/).

    osf_retrieve_file("vbdah") %>%
      osf_download(excel_file)

Additional resources

For more information on OSF and osfr check out: