googleCloudStorageR

Mark Edmondson

2021-12-15

An R library for interacting with the Google Cloud Storage JSON API (api docs).

Setup

Google Cloud Storage charges you for storage (prices here).

You can use your own Google Project with a credit card added to create buckets, where the charges will apply. This can be done in the Google API Console

Configuring your own Google Project

You will first need to create a Google Cloud Project and make sure that the Cloud Storage API is turned on. It is by default for new projects.

The recommended way is to use gcs_setup() which will help you create and download an authentication JSON key, and set your default bucket.

library(googleCloudStorageR)
gcs_setup()

#ℹ ==Welcome to googleCloudStorageR v0.6.0 setup==
#This wizard will scan your system for setup options and help you with any that are missing. 
#Hit 0 or ESC to cancel. 
#
#1: Create and download JSON service account key
#2: Setup auto-authentication (JSON service account key)
#3: Setup default bucket
#
#Selection: |

It uses googleAuthR::gar_setup_menu() to create the wizard. You will need to have owner access to the project you are using.

After each menu option has completed, restart R and rerun gcs_setup() function to continue to the next step.

Upon successful set-up, you should see a message similar to below:

library(googleCloudStorageR)
#Setting scopes to https://www.googleapis.com/auth/devstorage.full_control and https://www.googleapis.com/auth/cloud-platform
#Successfully auto-authenticated via /Users/xxxx/googlecloudstorager-auth-key.json
#Set default bucket name to 'xxxxxx'

Manual setup

gcs_setup() works through the steps detailed below.

The instructions below are for when you visit the Google API console (https://console.developers.google.com/apis/)

Activate API

  1. Click on “APIs”
  2. Select and activate the Cloud Storage JSON API if not already active

Set environment variables

By default, all cloudyr packages look for the access key ID and secret access key in environment variables. You can also use this to specify a default bucket, and auto-authentication upon attaching the library. For example:

Sys.setenv("GCS_DEFAULT_BUCKET" = "my-default-bucket",
           "GCS_AUTH_FILE" = "/fullpath/to/service-auth.json")

These can alternatively be set on the command line or via an Renviron.site or .Renviron file (https://cran.r-project.org/web/packages/httr/vignettes/api-packages.html).

e.g.

In your .Renviron:

GCS_AUTH_FILE="/fullpath/to/service-auth.json"
GCS_DEFAULT_BUCKET=my-default-bucket

Auto-authentication

The best method for authentication is to use your own Google Cloud Project. You can specify the location of a service account JSON file taken from your Google Project:

    Sys.setenv("GCS_AUTH_FILE" = "/fullpath/to/auth.json")

This file will then used for authentication via gcs_auth() when you load the library:

## GCS_AUTH_FILE set so auto-authentication
library(googleCloudStorageR)

gcs_get_bucket("your-bucket")

Token-authentication

You can also seamlessly authenticate as the service account your compute resource (i.e. Cloud-Function, AI Notebook etc…) is using by requesting a token and then passing that token to gcs_auth with the gargle library

## Load googleCloudStorageR and gargle
library(googleCloudStorageR)
library(gargle)

## Fetch token. See: https://developers.google.com/identity/protocols/oauth2/scopes
scope <-c("https://www.googleapis.com/auth/cloud-platform")
token <- token_fetch(scopes = scope)

## Pass your token to gcs_auth
gcs_auth(token = token)

## Perform gcs operations as normal
gcs_list_objects(bucket = "my-bucket")

Setting a default Bucket

To avoid specifying the bucket in the functions below, you can set the name of your default bucket via environmental variables or via the function gcs_global_bucket(). See the Setting environment variables section for more details.

## set bucket via environment
Sys.setenv("GCS_DEFAULT_BUCKET" = "my-default-bucket")

library(googleCloudStorageR)

## check what the default bucket is
gcs_get_global_bucket()
[1] "my-default-bucket"

## you can also set a default bucket after loading the library for that session
gcs_global_bucket("your-default-bucket-2")
gcs_get_global_bucket()
[1] "my-default-bucket-2"

Downloading objects from Google Cloud storage

Once you have a Google project and created a bucket with an object in it, you can download it as below:

library(googleCloudStorageR)

## get your project name from the API console
proj <- "your-project"

## get bucket info
buckets <- gcs_list_buckets(proj)
bucket <- "your-bucket"
bucket_info <- gcs_get_bucket(bucket)
bucket_info

==Google Cloud Storage Bucket==
Bucket:          your-bucket 
Project Number:  1123123123 
Location:        EU 
Class:           STANDARD 
Created:         2016-04-28 11:39:06 
Updated:         2016-04-28 11:39:06 
Meta-generation: 1 
eTag:            Cxx=


## get object info in the default bucket
objects <- gcs_list_objects()

## save directly to an R object (warning, don't run out of RAM if its a big object)
## the download type is guessed into an appropriate R object
parsed_download <- gcs_get_object(objects$name[[1]])

## if you want to do your own parsing, set parseObject to FALSE
## use httr::content() to parse afterwards
raw_download <- gcs_get_object(objects$name[[1]], 
                               parseObject = FALSE)

## save directly to a file in your working directory
## parseObject has no effect, it is a httr::content(req, "raw") download
gcs_get_object(objects$name[[1]], saveToDisk = "csv_downloaded.csv")

Uploading objects - simple uploads

Objects can be uploaded via files saved to disk, or passed in directly if they are data frames or list type R objects. By default, data frames will be converted to CSV via write.csv(), lists to JSON via jsonlite::toJSON.

If you want to use other functions for transforming R objects, for example setting row.names = FALSE or using write.csv2, pass the function through object_function

## upload a file - type will be guessed from file extension or supply type  
write.csv(mtcars, file = filename)
gcs_upload(filename)

## upload an R data.frame directly - will be converted to csv via write.csv
gcs_upload(mtcars)

## upload an R list - will be converted to json via jsonlite::toJSON
gcs_upload(list(a = 1, b = 3, c = list(d = 2, e = 5)))

## upload an R data.frame directly, with a custom function
## function should have arguments 'input' and 'output'
## safest to supply type too
f <- function(input, output) write.csv(input, row.names = FALSE, file = output)

gcs_upload(mtcars, 
           object_function = f,
           type = "text/csv")

Since 2019 you can also set bucket level access permissions. To upload to those buckets, specify the defaultAcl="bucketLevel"

gcs_upload(mtcars, 
           bucket = "mark-bucketlevel-acl",
           predefinedAcl = "bucketLevel")

Upload metadata

You can pass metadata with an object via the function gcs_metadata_object().

the name you pass to the metadata object will override the name if it is also set elsewhere.

meta <- gcs_metadata_object("mtcars.csv",
                             metadata = list(custom1 = 2,
                                             custom_key = 'dfsdfsdfsfs))
                                             
gcs_upload(mtcars, object_metadata = meta)

Resumable uploads for files > 5MB up to 5TB

If the file/object is small, simple uploads are used. You can modify this limit using option(googleCloudStorageR.upload_limit) or gcs_upload_set_limit() - default is 5000000L or 5MB (#120)

For files greater than the upload limit, resumable uploads are used. This allows you to upload up to 5TB.

If you get an interrupted connection when uploading, gcs_upload will retry 3 times, if it fails it will return a Retry object, that you can try again later from where the upload stopped. Call this via gcs_retry_upload

## write a big object to a file
big_file <- "big_filename.csv"
write.csv(big_object, file = big_file)

## attempt upload
upload_try <- gcs_upload(big_file)

## if successful, upload_try is an object metadata object
upload_try
==Google Cloud Storage Object==
Name:            "big_filename.csv" 
Size:            8.5 Gb 
Media URL        https://www.googleapis.com/download/storage/v1/b/xxxx 
Bucket:          your-bucket 
ID:              your-bucket/"test.pdf"/xxxx
MD5 Hash:        rshao1nxxxxxY68JZQ== 
Class:           STANDARD 
Created:         2016-08-12 17:33:05 
Updated:         2016-08-12 17:33:05 
Generation:      1471023185977000 
Meta Generation: 1 
eTag:            CKi90xxxxxEAE= 
crc32c:          j4i1sQ== 


## if unsuccessful after 3 retries, upload_try is a Retry object
==Google Cloud Storage Upload Retry Object==
File Location:     big_filename.csv
Retry Upload URL:  http://xxxx
Created:           2016-08-12 17:33:05 
Type:              csv
File Size:        8.5 Gb
Upload Byte:      4343
Upload remaining: 8.1 Gb

## you can retry to upload the remaining data using gcs_retry_upload()
try2 <- gcs_retry_upload(upload_try)

Updating user access to objects

You can change who can access objects via gcs_update_acl to one of READER or OWNER, on a user, group, domain, project or public for all users or authenticated users.

By default you are “OWNER” of all the objects and buckets you upload and create.

## update access of object to READER for all public
gcs_update_object_acl("your-object.csv", entity_type = "allUsers")

## update access of object for user joe@blogs.com to OWNER
gcs_update_acl("your-object.csv", 
               entity = "joe@blogs.com", 
               role = "OWNER")

## update access of object for googlegroup users to READER
gcs_update_object_acl("your-object.csv", 
                      entity = "my-group@googlegroups.com", 
                      entity_type = "group")

## update access of object for all users to OWNER on your Google Apps domain
gcs_update_object_acl("your-object.csv", 
                      entity = "yourdomain.com", 
                      entity_type = "domain", 
                      role = OWNER)

Since 2019 you can also set bucket level access permissions. To upload to those buckets, specify the defaultAcl="bucketLevel"

gcs_upload(mtcars, 
           bucket = "mark-bucketlevel-acl",
           predefinedAcl = "bucketLevel")

Deleting an object

Delete an object by passing its name (and bucket if not default)

## returns TRUE is successful, a 404 error if not found
gcs_delete_object("your-object.csv")

Viewing current access level to objects

Use gcs_get_object_acl() to see what the current access is for an entity + entity_type.

## default entity_type is user
acl <- gcs_get_object_acl("your-object.csv", 
                         entity = "joe@blogs.com")
acl$role 
[1] "OWNER"

## for allUsers and allAuthenticated users, you don't need to supply entity
acl <- gcs_get_object_acl("your-object.csv", 
                          entity_type = "allUsers")
acl$role 
[1] "READER"

R Session helpers

Versions of save.image(), save() and load() are implemented called gcs_save_image(), gcs_save() and gcs_load(). These functions save and load the global R session to the cloud.

## save the current R session including all objects
gcs_save_image()

### wipe environment
rm(list = ls())

## load up environment again
gcs_load()

Save specific objects:

cc <- 3
d <- "test1"
gcs_save("cc","d", file = "gcs_save_test.RData")

## remove the objects saved in cloud from local environment
rm(cc,d)

## load them back in from GCS
gcs_load(file = "gcs_save_test.RData")
cc == 3
[1] TRUE
d == "test1"
[1] TRUE

You can also upload .R code files and source them directly using gcs_source:

## make a R source file and upload it
cat("x <- 'hello world!'\nx", file = "example.R")
gcs_upload("example.R", name = "example.R")

## source the file to run its code
gcs_source("example.R")

## the code from the upload file has run
x
[1] "hello world!"

Uploading via a Shiny app

The library is also compatible with Shiny authentication flows, so you can create Shiny apps that lets users log in and upload their own data.

An example of that is shown below:

library("shiny")
library("googleAuthR")
library("googleCloudStorageR")

## you need to start Shiny app on port 1221
## as thats what the default googleAuthR project expects for OAuth2 authentication

## options(shiny.port = 1221)
## print(source('shiny_test.R')$value) or push the "Run App" button in RStudio

shinyApp(
  ui = shinyUI(
      fluidPage(
        googleAuthR::googleAuthUI("login"),
        fileInput("picture", "picture"),
        textInput("filename", label = "Name on Google Cloud Storage",value = "myObject"),
        actionButton("submit", "submit"),
        textOutput("meta_file")
      )
  ),
  server = shinyServer(function(input, output, session){

    access_token <- shiny::callModule(googleAuth, "login")

    meta <- eventReactive(input$submit, {

      message("Uploading to Google Cloud Storage")
      
      # from googleCloudStorageR
      with_shiny(gcs_upload,  
                 file = input$picture$datapath,
                 # enter your bucket name here
                 bucket = "gogauth-test",  
                 type = input$picture$type,
                 name = input$filename,
                 shiny_access_token = access_token())

    })

    output$meta_file <- renderText({
      
      req(meta())

      str(meta())

      paste("Uploaded: ", meta()$name)

    })

  })
)

Bucket administration

There are various functions to manipulate Buckets:

Object administration

You can get meta data about an object by passing meta=TRUE to gcs_get_object

gcs_get_object("your-object", "your-bucket", meta = TRUE)

Explanation of Google Project access

googleCloudStorageR has its own Google project which is used to call the Google Cloud Storage API, but does not have access to the objects or buckets in your Google Project unless you give permission for the library to access your own buckets during the OAuth2 authentication process.

No other user, including the owner of the Google Cloud Storage API project has access unless you have given them access, but you may want to change to use your own Google Project (that could or could not be the same as the one that holds your buckets).