Frictionless Data Package

Description

R package for working with Frictionless Data Package.

Features

Getting started

Installation

In order to install the latest distribution of R software to your computer you have to select one of the mirror sites of the Comprehensive R Archive Network, select the appropriate link for your operating system and follow the wizard instructions.

For windows users you can:

  1. Go to CRAN
  2. Click download R for Windows
  3. Click Base (This is what you want to install R for the first time)
  4. Download the latest R version
  5. Run installation file and follow the instrustions of the installer.

(Mac) OS X and Linux users may need to follow different steps depending on their system version to install R successfully and it is recommended to read the instructions on CRAN site carefully.

Even more detailed installation instructions can be found in R Installation and Administration manual.

To install RStudio, you can download RStudio Desktop with Open Source License and follow the wizard instructions:

  1. Go to RStudio
  2. Click download on RStudio Desktop
  3. Download on RStudio Desktop free download
  4. Select the appropriate file for your system
  5. Run installation file

To install the datapackage package it is necessary to install first devtools package to make installation of github packages available.

# Install devtools package if not already
install.packages("devtools")

Install datapackage.r

# And then install the development version from github
devtools::install_github("frictionlessdata/datapackage-r")

Load package

# load the package using
library(datapackage.r)

Examples

Code examples in this readme requires R 3.3 or higher, You could see even more examples in vignettes directory.

descriptor <- '{
  "resources": [
    {
      "name": "example",
      "profile": "tabular-data-resource",
      "data": [
        ["height", "age", "name"],
        [180, 18, "Tony"],
        [192, 32, "Jacob"]
      ],
      "schema":  {
        "fields": [
          {"name": "height", "type": "integer" },
          {"name": "age", "type": "integer" },
          {"name": "name", "type": "string" }
        ]
      }
    }
  ]
}'

dataPackage <- Package.load(descriptor)
dataPackage
## <Package>
##   Public:
##     addResource: function (descriptor) 
##     clone: function (deep = FALSE) 
##     commit: function (strict = NULL) 
##     descriptor: active binding
##     errors: active binding
##     getResource: function (name) 
##     infer: function (pattern) 
##     initialize: function (descriptor = list(), basePath = NULL, strict = FALSE, 
##     profile: active binding
##     removeResource: function (name) 
##     resourceNames: active binding
##     resources: active binding
##     save: function (target, type = "json") 
##     valid: active binding
##   Private:
##     basePath_: C:/Users/akis_/Documents/datapackage-r
##     build_: function () 
##     currentDescriptor_: list
##     currentDescriptor_json: NULL
##     descriptor_: NULL
##     errors_: list
##     nextDescriptor_: list
##     pattern_: NULL
##     profile_: Profile, R6
##     resources_: list
##     resources_length: NULL
##     strict_: FALSE
resource <- dataPackage$getResource('example')
# convert to json and add indentation with jsonlite prettify function
jsonlite::prettify(helpers.from.list.to.json(resource$read()))
## [
##     [
##         180,
##         18,
##         "Tony"
##     ],
##     [
##         192,
##         32,
##         "Jacob"
##     ]
## ]
## 

Documentation

Json objects are not included in R base data types. Jsonlite package is internally used to convert json data to list objects. The input parameters of functions could be json strings, files or lists and the outputs are in list format to easily further process your data in R environment and exported as desired. The examples below show how to use jsonlite package to convert the output back to json adding indentation whitespace. More details about handling json you can see jsonlite documentation or vignettes here.

Package

A class for working with data packages. It provides various capabilities like loading local or remote data package, inferring a data package descriptor, saving a data package descriptor and many more.

Consider we have some local csv files in a data directory. Let’s create a data package based on this data using a Package class:

inst/extdata/readme_example/cities.csv

city,location
london,"51.50,-0.11"
paris,"48.85,2.30"
rome,"41.89,12.51"

inst/extdata/readme_example/population.csv

city,year,population
london,2017,8780000
paris,2017,2240000
rome,2017,2860000

First we create a blank data package:

dataPackage <- Package.load()

Now we’re ready to infer a data package descriptor based on data files we have. Because we have two csv files we use glob pattern csv:

jsonlite::toJSON(dataPackage$infer('csv'), pretty = TRUE)
## {
##   "profile": ["tabular-data-package"],
##   "resources": [
##     {
##       "path": ["cities.csv"],
##       "profile": ["tabular-data-resource"],
##       "encoding": ["utf-8"],
##       "name": ["cities"],
##       "format": ["csv"],
##       "mediatype": ["text/csv"],
##       "schema": {
##         "fields": [
##           {
##             "name": ["city"],
##             "type": ["string"],
##             "format": ["default"]
##           },
##           {
##             "name": ["location"],
##             "type": ["string"],
##             "format": ["default"]
##           }
##         ],
##         "missingValues": [
##           [""]
##         ]
##       }
##     },
##     {
##       "path": ["population.csv"],
##       "profile": ["tabular-data-resource"],
##       "encoding": ["utf-8"],
##       "name": ["population"],
##       "format": ["csv"],
##       "mediatype": ["text/csv"],
##       "schema": {
##         "fields": [
##           {
##             "name": ["city"],
##             "type": ["string"],
##             "format": ["default"]
##           },
##           {
##             "name": ["year"],
##             "type": ["integer"],
##             "format": ["default"]
##           },
##           {
##             "name": ["population"],
##             "type": ["integer"],
##             "format": ["default"]
##           }
##         ],
##         "missingValues": [
##           [""]
##         ]
##       }
##     }
##   ]
## }
jsonlite::toJSON(dataPackage$descriptor, pretty = TRUE)
## {
##   "profile": ["tabular-data-package"],
##   "resources": [
##     {
##       "path": ["cities.csv"],
##       "profile": ["tabular-data-resource"],
##       "encoding": ["utf-8"],
##       "name": ["cities"],
##       "format": ["csv"],
##       "mediatype": ["text/csv"],
##       "schema": {
##         "fields": [
##           {
##             "name": ["city"],
##             "type": ["string"],
##             "format": ["default"]
##           },
##           {
##             "name": ["location"],
##             "type": ["string"],
##             "format": ["default"]
##           }
##         ],
##         "missingValues": [
##           [""]
##         ]
##       }
##     },
##     {
##       "path": ["population.csv"],
##       "profile": ["tabular-data-resource"],
##       "encoding": ["utf-8"],
##       "name": ["population"],
##       "format": ["csv"],
##       "mediatype": ["text/csv"],
##       "schema": {
##         "fields": [
##           {
##             "name": ["city"],
##             "type": ["string"],
##             "format": ["default"]
##           },
##           {
##             "name": ["year"],
##             "type": ["integer"],
##             "format": ["default"]
##           },
##           {
##             "name": ["population"],
##             "type": ["integer"],
##             "format": ["default"]
##           }
##         ],
##         "missingValues": [
##           [""]
##         ]
##       }
##     }
##   ]
## }

An infer method has found all our files and inspected it to extract useful metadata like profile, encoding, format, Table Schema etc. Let’s tweak it a little bit:

dataPackage$descriptor$resources[[2]]$schema$fields[[2]]$type <- 'year'
dataPackage$commit()
## [1] TRUE
dataPackage$valid
## [1] TRUE

Because our resources are tabular we could read it as a tabular data:

jsonlite::toJSON(dataPackage$getResource("population")$read(keyed = TRUE),auto_unbox = FALSE,pretty = TRUE)
## [
##   {
##     "city": ["london"],
##     "year": [2017],
##     "population": [8780000]
##   },
##   {
##     "city": ["paris"],
##     "year": [2017],
##     "population": [2240000]
##   },
##   {
##     "city": ["rome"],
##     "year": [2017],
##     "population": [2860000]
##   }
## ]

Let’s save our descriptor on the disk. After it we could update our datapackage.json as we want, make some changes etc:

dataPackage.save('datapackage.json')

To continue the work with the data package we just load it again but this time using local datapackage.json:

dataPackage <- Package.load('datapackage.json')
# Continue the work

It was onle basic introduction to the Package class. To learn more let’s take a look on Package class API reference.

Package.load(descriptor, basePath, strict=FALSE)

Constructor to instantiate Package class.

package$valid

package$errors

package$profile

package$descriptor

package$resources

package$resourceNames

package$getResource(name)

Get data package resource by name.

package$addResource(descriptor)

Add new resource to data package. The data package descriptor will be validated with newly added resource descriptor.

package$removeResource(name)

Remove data package resource by name. The data package descriptor will be validated after resource descriptor removal.

package$infer(pattern=FALSE)

Infer a data package metadata. If pattern is not provided only existent resources will be inferred (added metadata like encoding, profile etc). If pattern is provided new resoures with file names mathing the pattern will be added and inferred. It commits changes to data package instance.

package$commit(strict)

Update data package instance if there are in-place changes in the descriptor.

dataPackage <- Package.load('{ 
 "name": "package",
 "resources": [{
  "name": "resource",
  "data": ["data"]
 }]
 }')

dataPackage$descriptor$name # package
## [1] "package"
dataPackage$descriptor$name <- 'renamed-package'
dataPackage$commit() # TRUE
## [1] TRUE
dataPackage$descriptor$name # renamed-package
## [1] "renamed-package"

package.save(target)

For now only descriptor will be saved.

Save data package to target destination.

Resource

A class for working with data resources. You can read or iterate tabular resources using the iter/read methods and all resource as bytes using rowIter/rowRead methods.

Consider we have some local csv file. It could be inline data or remote link - all supported by Resource class (except local files for in-brower usage of course). But say it’s cities.csv for now:

city,location
london,"51.50,-0.11"
paris,"48.85,2.30"
rome,N/A

Let’s create and read a resource. We use static Resource$load method instantiate a resource. Because resource is tabular we could use resourceread method with a keyed option to get an array of keyed rows:

resource <- Resource.load('{"path": "cities.csv"}')
resource$tabular
## [1] TRUE
jsonlite::toJSON(resource$read(keyed = TRUE), pretty = TRUE)
## [
##   {
##     "city": ["london"],
##     "location": ["\"51.50 -0.11\""]
##   },
##   {
##     "city": ["paris"],
##     "location": ["\"48.85 2.30\""]
##   },
##   {
##     "city": ["rome"],
##     "location": ["\"41.89 12.51\""]
##   }
## ]

As we could see our locations are just a strings. But it should be geopoints. Also Rome’s location is not available but it’s also just a N/A string instead of null. First we have to infer resource metadata:

jsonlite::toJSON(resource$infer(), pretty = TRUE)
## {
##   "path": ["cities.csv"],
##   "profile": ["tabular-data-resource"],
##   "encoding": ["utf-8"],
##   "name": ["cities"],
##   "format": ["csv"],
##   "mediatype": ["text/csv"],
##   "schema": {
##     "fields": [
##       {
##         "name": ["city"],
##         "type": ["string"],
##         "format": ["default"]
##       },
##       {
##         "name": ["location"],
##         "type": ["string"],
##         "format": ["default"]
##       }
##     ],
##     "missingValues": [
##       [""]
##     ]
##   }
## }
jsonlite::toJSON(resource$descriptor, pretty = TRUE)
## {
##   "path": ["cities.csv"],
##   "profile": ["tabular-data-resource"],
##   "encoding": ["utf-8"],
##   "name": ["cities"],
##   "format": ["csv"],
##   "mediatype": ["text/csv"],
##   "schema": {
##     "fields": [
##       {
##         "name": ["city"],
##         "type": ["string"],
##         "format": ["default"]
##       },
##       {
##         "name": ["location"],
##         "type": ["string"],
##         "format": ["default"]
##       }
##     ],
##     "missingValues": [
##       [""]
##     ]
##   }
## }
# resource$read( keyed = TRUE )
# # Fails with a data validation error

Let’s fix not available location. There is a missingValues property in Table Schema specification. As a first try we set missingValues to N/A in resource$descriptor.schema. Resource descriptor could be changed in-place but all changes should be commited by resource$commit():

resource$descriptor$schema$missingValues <- 'N/A'
resource$commit()
## [1] TRUE
resource$valid # FALSE
## [1] FALSE
resource$errors
## [[1]]
## [1] "Descriptor validation error:\n            data.schema.missingValues - is the wrong type"

As a good citiziens we’ve decided to check out recource descriptor validity. And it’s not valid! We should use an array for missingValues property. Also don’t forget to have an empty string as a missing value:

resource$descriptor$schema[['missingValues']] <- list('', 'N/A')
resource$commit()
## [1] TRUE
resource$valid # TRUE
## [1] TRUE

All good. It looks like we’re ready to read our data again:

jsonlite::toJSON(resource$read( keyed = TRUE ), pretty = TRUE)
## [
##   {
##     "city": ["london"],
##     "location": ["\"51.50 -0.11\""]
##   },
##   {
##     "city": ["paris"],
##     "location": ["\"48.85 2.30\""]
##   },
##   {
##     "city": ["rome"],
##     "location": ["\"41.89 12.51\""]
##   }
## ]

Now we see that: - locations are arrays with numeric lattide and longitude - Rome’s location is a native JavaScript null

And because there are no errors on data reading we could be sure that our data is valid againt our schema. Let’s save our resource descriptor:

resource$save('dataresource.json')

Let’s check newly-crated dataresource.json. It contains path to our data file, inferred metadata and our missingValues tweak:

{
    "path": "data.csv",
    "profile": "tabular-data-resource",
    "encoding": "utf-8",
    "name": "data",
    "format": "csv",
    "mediatype": "text/csv",
    "schema": {
        "fields": [
            {
                "name": "city",
                "type": "string",
                "format": "default"
            },
            {
                "name": "location",
                "type": "geopoint",
                "format": "default"
            }
        ],
        "missingValues": [
            "",
            "N/A"
        ]
    }
}

If we decide to improve it even more we could update the dataresource.json file and then open it again using local file name:

resource <- Resource.load('dataresource.json')
# Continue the work

It was onle basic introduction to the Resource class. To learn more let’s take a look on Resource class API reference.

Resource$load(descriptor, basePath, strict=FALSE)

Constructor to instantiate Resource class.

resource$valid

resource$errors

resource$profile

resource$descriptor

resource$name

resource$inline

resource$local

resource$remote

resource$multipart

resource$tabular

resource$source

Combination of resource$source and resource$inline/local/remote/multipart provides predictable interface to work with resource data.

resource$headers

Only for tabular resources

resource$schema

Only for tabular resources

It returns Schema instance to interact with data schema. Read API documentation - tableschema.Schema.

resource$iter(keyed, extended, cast=TRUE, relations=FALSE, stream=FALSE)

Only for tabular resources

Iter through the table data and emits rows cast based on table schema (async for loop). Data casting could be disabled.

resource$read(keyed, extended, cast=TRUE, relations=FALSE, limit)

Only for tabular resources

Read the whole table and returns as array of rows. Count of rows could be limited.

resource$checkRelations()

Only for tabular resources

It checks foreign keys and raises an exception if there are integrity issues.

resource$rawIter(stream = FALSE)

Iterate over data chunks as bytes. If stream is true Node Stream will be returned.

resource$rawRead()

Returns resource data as bytes.

resource$infer()

Infer resource metadata like name, format, mediatype, encoding, schema and profile. It commits this changes into resource instance.

resource$commit(strict)

Update resource instance if there are in-place changes in the descriptor.

resource$save(target)

For now only descriptor will be saved.

Save resource to target destination.

Profile

A component to represent JSON Schema profile from Profiles Registry:

profile <- Profile.load('data-package')
profile$name # data-package
## [1] "data-package"
profile$jsonschema # List of JSON Schema contents
valid_errors <- profile$validate(descriptor)
valid <- valid_errors$valid # TRUE if valid descriptor
valid
## [1] TRUE

Profile.load(profile)

Constuctor to instantiate Profile class.

Profile$name()

Profile$jsonschema()

Profile$validate(descriptor)

Validate a data package descriptor against the Profile$

Validate

A standalone function to validate a data package descriptor:

valid_errors <- validate('{"name": "Invalid Datapackage"}')

validate(descriptor)

A standalone function to validate a data package descriptor:

Infer

A standalone function to infer a data package descriptor.

descriptor <- infer("csv",basePath = '.')
jsonlite::toJSON(descriptor, pretty = TRUE)
## {
##   "profile": ["tabular-data-package"],
##   "resources": [
##     {
##       "path": ["cities.csv"],
##       "profile": ["tabular-data-resource"],
##       "encoding": ["utf-8"],
##       "name": ["cities"],
##       "format": ["csv"],
##       "mediatype": ["text/csv"],
##       "schema": {
##         "fields": [
##           {
##             "name": ["city"],
##             "type": ["string"],
##             "format": ["default"]
##           },
##           {
##             "name": ["location"],
##             "type": ["string"],
##             "format": ["default"]
##           }
##         ],
##         "missingValues": [
##           [""]
##         ]
##       }
##     },
##     {
##       "path": ["population.csv"],
##       "profile": ["tabular-data-resource"],
##       "encoding": ["utf-8"],
##       "name": ["population"],
##       "format": ["csv"],
##       "mediatype": ["text/csv"],
##       "schema": {
##         "fields": [
##           {
##             "name": ["city"],
##             "type": ["string"],
##             "format": ["default"]
##           },
##           {
##             "name": ["year"],
##             "type": ["integer"],
##             "format": ["default"]
##           },
##           {
##             "name": ["population"],
##             "type": ["integer"],
##             "format": ["default"]
##           }
##         ],
##         "missingValues": [
##           [""]
##         ]
##       }
##     }
##   ]
## }

infer(pattern, basePath)

Infer a data package descriptor.

Foreign Keys

The package supports foreign keys described in the Table Schema specification. It means if your data package descriptor use resources[]$schema$foreignKeys property for some resources a data integrity will be checked on reading operations.

Consider we have a data package:

DESCRIPTOR <- '{
  "resources": [
    {
      "name": "teams",
      "data": [
        ["id", "name", "city"],
        ["1", "Arsenal", "London"],
        ["2", "Real", "Madrid"],
        ["3", "Bayern", "Munich"]
      ],
      "schema": {
        "fields": [
          {"name": "id", "type": "integer"},
          {"name": "name", "type": "string"},
          {"name": "city", "type": "string"}
        ],
        "foreignKeys": [
          {
            "fields": "city",
            "reference": {"resource": "cities", "fields": "name"}
          }
        ]
      }
    }, {
      "name": "cities",
      "data": [
        ["name", "country"],
        ["London", "England"],
        ["Madrid", "Spain"]
      ]
    }
  ]
}'

Let’s check relations for a teams resource:

package <- Package.load(DESCRIPTOR)
teams <- package$getResource('teams')
teams$checkRelations()
## Error: Foreign key 'city' violation in row '4'
# tableschema.exceptions.RelationError: Foreign key "['city']" violation in row "4"

As we could see there is a foreign key violation. That’s because our lookup table cities doesn’t have a city of Munich but we have a team from there. We need to fix it in cities resource:

package$descriptor$resources[[2]]$data <- rlist::list.append(package$descriptor$resources[[2]]$data, list('Munich', 'Germany'))
package$commit()
## [1] TRUE
teams <- package$getResource('teams')
teams$checkRelations()
## [1] TRUE
# TRUE

Fixed! But not only a check operation is available. We could use relations argument for resource$iter/read methods to dereference a resource relations:

jsonlite::toJSON(teams$read(keyed = TRUE, relations = FALSE), pretty =  TRUE)
## [
##   {
##     "id": [1],
##     "name": ["Arsenal"],
##     "city": ["London"]
##   },
##   {
##     "id": [2],
##     "name": ["Real"],
##     "city": ["Madrid"]
##   },
##   {
##     "id": [3],
##     "name": ["Bayern"],
##     "city": ["Munich"]
##   }
## ]

Instead of plain city name we’ve got a dictionary containing a city data. These resource$iter/read methods will fail with the same as resource$check_relations error if there is an integrity issue. But only if relations = TRUE flag is passed.

Errors

errors$DataPackageError

Base class for the all package errors. If there are more than one error you could get an additional information from the error object:

tryCatch({
  # some lib action
}, error = function() {
  error # you have N cast errors (see error.errors)
  if (error$multiple) {
    for ( error in error$errors) {
        error # cast error M is ...
    }
  }
})

Changelog - News

In NEWS.md described only breaking and the most important changes. The full changelog could be found in nicely formatted commit history.

Contributing

The project follows the Open Knowledge International coding standards. There are common commands to work with the project.Recommended way to get started is to create, activate and load the package environment. To install package and development dependencies into active environment:

devtools::install_github("frictionlessdata/datapackage-r", dependencies=TRUE)

To make test:

  test_that(description, {
    expect_equal(test, expected result)
  })

To run tests:

devtools::test()

more detailed information about how to create and run tests you can find in testthat package

Github