Resource R

The resourcer package is meant to access resources identified by a URL in a uniform way whether it references a dataset (stored in a file, a SQL table, a MongoDB collection etc.) or a computation unit (system commands, web services etc.). Usually some credentials will be defined, and an additional data format information can be provided to help dataset coercing to a data.frame object.

The main concepts are:

• Resource, access to a resource (dataset or computation unit) is described by an object with URL, optional credentials and optional data format properties,
• ResourceResolver, a ResourceClient factory based on the URL scheme and available in a resolvers registry,
• ResourceClient, realizes the connection with the dataset or the computation unit described by a Resource,
• FileResourceGetter, connect to a file described by a resource,
• DBIResourceConnector, establish a DBI connection.

Install

Install from CRAN:

install.packages("resourcer")

The resourcer has quite some suggested dependencies. These are only suggestions, meaning that it will depend on the kind of resource that will be accessed at runtime.

Tidy files

• haven: Import and Export 'SPSS', 'Stata' and 'SAS' Files
• readr: Read Rectangular Text Data
• readxl: Read Excel Files
• dplyr: A Grammar of Data Manipulation

Databases

• dbplyr: A 'dplyr' Back End for Databases
• DBI: R Database Interface
• RMariaDB: Database Interface and 'MariaDB' Driver
• RPostgres: 'Rcpp' Interface to 'PostgreSQL'
• sparklyr: R Interface to Apache Spark
• RPresto: DBI Connector to Presto
• nodbi: 'NoSQL' Database Connector
• mongolite: Fast and Simple 'MongoDB' Client for R

Remote computation server

• ssh: Secure Shell (SSH) Client for R

System dependencies

R packages often depend on system libraries or other software external to R. These dependencies are not automatically installed.

See the provided example script for installing the system requirements, per R package, for a Ubuntu 18.04 system: install-system-requirements-ubuntu18.sh

File Resources

These are resources describing a file. If the file is in a remote location, it must be downloaded before being read. The data format specification of the resource helps to find the appropriate file reader.

File Getter

The file locations supported by default are:

• file, local file system,
• http(s), web address, basic authentication,
• gridfs, MongoDB file store,
• scp, file copy through SSH,
• opal, Opal file store.

This can be easily applied to other file locations by extending the FileResourceGetter class. An instance of the new file resource getter is to be registered so that the FileResourceResolver can operate as expected.

registerFileResourceGetter(MyFileLocationResourceGetter$new())

File Data Format

The data format specified within the Resource object, helps at finding the appropriate file reader. Currently supported data formats are:

• the data formats that have a reader in tidyverse: readr (csv, csv2, tsv, ssv, delim), haven (spss, sav, por, dta, stata, sas, xpt), readxl (excel, xls, xlsx). This can be easily applied to other data file formats by extending the FileResourceClient class.
• the R data format that can be loaded in a child R environment from which object of interest will be retrieved.

Usage example that reads a local SPSS file:

# make a SPSS file resource
res <- resourcer::newResource(
  name = "CNSIM1",
  url = "file:///data/CNSIM1.sav",
  format = "spss"
)

# coerce the csv file in the opal server to a data.frame
df <- as.data.frame(res)

To support other file data format, extend the FileResourceClient class with the new data format reader implementation. Associate factory class, an extension of the ResourceResolver class is also to be implemented and registered.

registerResourceResolver(MyFileFormatResourceResolver$new())

Database Resources

DBI Connectors

DBI is a set of virtual classes that are are used to abstract the SQL database connections and operations within R. Then any DBI implementation can be used to access to a SQL table. Which DBI connector to be used is an information that can be extracted from the scheme part of the resource's URL. For instance a resource URL starting with postgres:// will require the RPostgres driver. To separate the DBI connector instanciation from the DBI interface interactions in the SQLResourceClient, a DBIResourceConnector registry is to be populated. The currently supported SQL database connectors are:

• mariadb MariaDB connector,
• mysql MySQL connector,
• postgres or postgresql Postgres connector,
• presto, presto+http or presto+https Presto connector,
• spark, spark+http or spark+https Spark connector.

To support another SQL database having a DBI driver, extend the DBIResourceConnector class and register it:

registerDBIResourceConnector(MyDBResourceConnector$new())

Use dplyr

Having the data stored in the database allows to handle large (common SQL databases) to big (PrestoDB, Spark) datasets using dplyr which will delegate as much as possible operations to the database.

Document Databases

NoSQL databases can be described by a resource. The nodbi can be used here. Currently only connection to MongoDB database is supported using URL scheme mongodb or mongodb+srv.

Computation Resources

Computation resources are resources on which tasks/commands can be triggerred and from which resulting data can be retrieved.

Example of computation resource that connects to a server through SSH:

# make an application resource on a ssh server
res <- resourcer::newResource(
  name = "supercomp1",
  url = "ssh://server1.example.org/work/dir?exec=plink,ls",
  identity = "sshaccountid",
  secret = "sshaccountpwd"
)

# get ssh client from resource object
client <- resourcer::newResourceClient(res) # does a ssh::ssh_connect()

# execute commands
files <- client$exec("ls") # exec 'cd /work/dir && ls'

# release connection
client$close() # does ssh::ssh_disconnect(session)

Extending Resources

There are several ways to extend the Resources handling. These are based on different R6 classes having a isFor(resource) function:

• If the resource is a file located at a place not already handled, write a new FileResourceGetter subclass and register an instance of it with the function registerFileResourceGetter().
• If the resource is a SQL engine having a DBI connector defined, write a new DBIResourceConnector subclass and register an instance of it with the function registerDBIResourceConnector().
• If the resource is in a domain specific web application or database, write a new ResourceResolver subclass and register an instance of it with the function registerResourceResolver(). This ResourceResolver object will create the appropriate ResourceClient object that matches your needs.

The design of the URL that will describe your new resource should not overlap an existing one, otherwise the different registries will return the first instance for which the isFor(resource) is TRUE. In order to distinguish resource locations, the URL's scheme can be extended, for instance the scheme for accessing a file in a Opal server is opal+https so that the credentials be applied as needed by Opal.

Resource Forms

As it can be error prone to define a new resource, when a URL is complex, or when there is a limited choice of formats or when credentials can be on different types, it is recommended to declare the resources forms and factory functions within the R package. This resource declaration is to be done in javascript, as this is a very commonly used language for building graphical user interfaces.

These files are expected to be installed at the root of the package folder (then in the source code of the R package, they will be declared in the inst/resources folder), so that an external application can lookup statically the packages having declared some resources.

The configuration file inst/resources/resource.js is a javascript file which contains an object with the properties:

• settings, a JSON object that contains the description and the documentation of the web forms (based on the json-schema specification).
• asResource, a javascript function that will convert the data captured from one of the declared web forms into a data structure representing the resource object.

As an example (see also resourcer's resource.js):

var myPackage = {
  settings: {
    "title": "MyPackage resources",
    "description": "MyPackage resources are for etc.",
    "web": "https://github.com/org/myPackage",
    "categories": [
      {
        "name": "my-format",
        "title": "My data format",
        "description": "Data are files in my format, that will be read by myPackage etc."
      }
    ],
    "types": [
      {
        "name": "my-format-http",
        "title": "My data format - HTTP",
        "description": "Data are files in my format, that will be downloaded from a HTTP server etc.",
        "tags": ["my-format", "http"],
        "parameters": {},
        "credentials": {}
      }
    ]
  },
  asResource: function(type, name, params, credentials) {
    // make a resource object from arguments, using type to drive 
    // what params/credentials properties are to be used
    // a basic example of resource object:
    return {
      "name": name,
      "url": params.url,
      "format": params.format,
      "identity": credentials.username,
      "secret": credentials.password
    };
  }
}

The specifications for the resource.js file are the following:

• settings object:

• category object:

• type object:

• asResource function: a javascript function which signature is function(type, name, params, credentials) where:
  • type, the form name used to capture the resource parameters and credentials,
  • name, the name to apply to the resource,
  • params, the captured parameters,
  • credentials, the captured credentials.

The name of the root object must follow the pattern: <R package> (note that any dots (.) in the R package name are to be replaced by underscores (_)).