Skip to content

Latest commit

 

History

History
259 lines (193 loc) · 10.2 KB

README.md

File metadata and controls

259 lines (193 loc) · 10.2 KB

README

Introduction

Many (if not all) data analytical projects integrate raw input data from upstream producers with tools and scripts aiding in their interpretation. Where R is used as the platform for such analysis, the canonical R packaging infrastructure (as e.g. described here in detail) is well suited for this integration and the representation of a package tracking input data and its analysis parallely might (minimally) look something like this in the file system:

  • analysis.of.x.2017 (top level directory named as the resulting package)
    • DESCRIPTION (standardized file containing basic information about the package)
    • NAMESPACE (file describing what functions/objects the resulting package provides and requires from other packages; see here for details)
    • R (Directory containing R functions)
      • Generalizable_Function_A.R
      • Generalizable_Function_B.R
      • ...
    • data (directory containing R objects saved using save(...) and lazy loaded on attaching the resulting package)
      • raw_data.Rda
    • inst (directory holding other material installed into Rs infrastructure along with the package)
      • extdata (directory holding compressed external raw data)
      • raw_data.csv.zip
      • scripts (directory holding work flow-documenting scripts)
        • 00_Script_Documenting_Data_Import.R
        • 01_Script_Documenting_Exploratory_Preliminary_Analysis.R
        • ...

After building and loading, such a package makes readily available from within R a) raw data (inst/extdata/raw_data.csv.zip), b) its parsed, immediately R-accessible counterpart (data/raw_data.Rda, respectively a raw_data object), c) generalizable custom R functionality (R/...), as well as d) work flow documentation (inst/scripts/...) in a neat, integrated and easily distributed unit. The structure is easily extended and further customized to e.g. include a manuscript when using rmarkdown and/or bookdown (e.g. in inst/mansucript/manuscript.Rmd).

When using R's packaging infrastructure in this manner, two major challenges are

  1. The complexity of establishing the project-specific infrastructure and

  2. Obstacles in distributing the data at the basis of the documented project, which may, among other issues, be rooted in

    • Size restrictions of shipped data in a targeted package repository (CRAN, for example, as of 23.10.2017 states "... As a general rule, neither data nor documentation should exceed 5MB.").
    • The need to enforce data onwnership/licensing in the context of freely providing peer review of analytical methodology and strategies employed.

Building heavily on the toolkit provided by the excellent devtools package, datapackageR aims to aleviate these difficulties and provides a simple interface for the management of raw data and deriving objects within the packaging infrastructure, emphasizing cryptographically strong data integrity assurance (using digest) and mechanisms for separate distribution of data and analyzing code.

How to get datapackageR set up?

datapackageR is targeted to eventually be released on CRAN, but until that happens and for the latest features and fixes, the following installation procedure should be followed:

install.packages('devtools')
devtools::install_bitbucket('graumannlabtools/datapackageR')

Example usage

Comprehensive presentation of functionality

  1. Load the required tools

     library(datapackageR)
 library(magrittr)
 library(readxl)
 requireNamespace('devtools')

  2. Define a temporary working directory & packaging root

     pkg_root <- '~' %>%
   file.path('packagetest')

  3. Generate a (dummy) data file

     data.frame(
     x   = 1,
     y   = 1:10,
     fac = sample(LETTERS[1:3], 10, replace = TRUE)) %>%
   write.table(
     file      = file.path(dirname(pkg_root), 'data_dummy.tsv'),
     sep       = '\t',
     col.names = TRUE,
     row.names = FALSE)

  4. Create the packaging skeleton, at the same time inserting the dummy data file from above:

     data_catalogue <- init(
   objects_to_include = file.path(dirname(pkg_root), 'data_dummy.tsv'),
   root               = pkg_root,
   parsing_function   = 'read.csv',
   parsing_options    = list(sep = '\t', stringsAsFactors = FALSE))

    • Investigate the resulting data_catalogue

        str(data_catalogue)

    • (Crudely) investigate the result in the file system

        list.files(pkg_root, recursive = TRUE)

  5. Addition of other object classes:

    • Add a remote file that needs parsing (from Billing et al. (2016). Comprehensive transcriptomic and proteomic characterization of human mesenchymal stem cells reveals source specific cellular markers. Sci Rep 6, 21507. Licensed under the Creative Commons Attribution 4.0 International License. Note that the file is excluded from built packages (via .rBuildignore) and wouldn't be tracked by git (using .gitignore).

        data_catalogue <- include_data(
    object_to_include    = 'http://www.nature.com/article-assets/npg/srep/2016/160209/srep21507/extref/srep21507-s4.xls',
    root                 = pkg_root,
    parsing_function     = 'read_excel',
    parsing_options      = list(skip = 1),
    package_dependencies = 'readxl',
    distributable        = FALSE)

    • Add a local object

        local_data_object <-list(A = LETTERS, B = letters)
  data_catalogue <- include_data(
    object_to_include = 'local_data_object',
    root              = pkg_root)

    • Add a remote *.Rda

        data_catalogue <- include_data(
    object_to_include = 'https://bitbucket.org/graumannlabtools/datapackager/downloads/remote_rda.Rda',
    root              = pkg_root)

    • Add a remote *.Rds

        data_catalogue <- include_data(
    object_to_include = 'https://bitbucket.org/graumannlabtools/datapackager/downloads/remote_rds.Rds',
    root              = pkg_root)

  6. Investigate the resulting data_catalogue & file system structure

     str(data_catalogue)
 list.files(pkg_root, recursive = TRUE)

  7. Remove one of the tracked data sets

     data_catalogue <- remove_data(
   object_to_remove = 'data_dummy.tsv',
   root             = pkg_root)

    • Investigate the resulting data_catalogue

        str(data_catalogue)

    • (Crudely) investigate the result in the file system

        list.files(pkg_root, recursive = TRUE)

  8. (Remotely) install the result & use the internal functionality to test data integrity

    devtools::install(pkg_root)
devtools::test(pkg_root)

  9. Clean up the example package & -root

    pkg_root %>%
  basename() %>%
  remove.packages()
unlink(pkg_root, recursive = TRUE)

Exemplary use case "seperately distributed data"

  1. Create an empty packaging skeleton:

     pkg_root <- tempdir() %>%
   file.path('packagetest')
 data_catalogue <- init(
   root = pkg_root)

  2. Attempt to add (access restricted) remote data

     data_catalogue <- include_data(
   object_to_include = 'https://bitbucket.org/graumannlabtools/datapackager-restricted-access/downloads/remote_rda.rda',
   root              = pkg_root)
 # --> Can't access URL: Client error: (401) Unauthorized

  3. Authenticatedly add remote data

     data_catalogue <- include_data(
   object_to_include = 'https://bitbucket.org/graumannlabtools/datapackager-restricted-access/downloads/remote_rda.rda',
   root              = pkg_root,
   user              = 'datapackageR_user',
   password          = 'datapackageR_user',
   distributable     = FALSE)

 data_catalogue <- include_data(
   object_to_include = 'https://bitbucket.org/graumannlabtools/datapackager-restricted-access/downloads/remote_rds.Rds',
   root              = pkg_root,
   user              = 'datapackageR_user',
   password          = 'datapackageR_user',
   distributable     = FALSE)

    • Investigate the resulting data_catalogue

        str(data_catalogue)

    • (Crudely) investigate the result in the file system

        list.files(pkg_root, recursive = TRUE)

  4. Simulate a package infrastructure shared without the remote & authentication-protected data

     list.files(
     pkg_root,
     recursive  = TRUE,
     pattern    = '^data[[:punct:]]remote',
     full.names = TRUE) %>%
   unlink()

    • (Crudely) investigate the result in the file system

        list.files(pkg_root, recursive = TRUE)

  5. Use datapackageR for attempted retreival of the missing data

     retrieve_missing_remote_data(pkg_root)
 # --> Can't access URL: Client error: (401) Unauthorized

  6. Repeat and properly authenticate (& implicitly check downloads against the stored hashes)

     retrieve_missing_remote_data(
   pkg_root,
   user     = 'datapackageR_user',
   password = 'datapackageR_user')

    • (Crudely) investigate the result in the file system

        list.files(pkg_root, recursive = TRUE)

  7. Clean up the example package & -root

     pkg_root %>%
   basename() %>%
   remove.packages()
 unlink(pkg_root, recursive = TRUE)