fakedata: Library for producing fake data

[ bsd3, fake, fakedata, library, random ] [ Propose Tags ]

Please see the README on GitHub at https://github.com/psibi/fakedata#readme

[Skip to Readme]
Versions [RSS] [faq], 0.2.0, 0.2.1, 0.2.2, 0.3.0, 0.3.1, 0.4.0, 0.5.0, 0.6.0, 0.6.1, 0.7.0, 0.7.1, 0.8.0, 1.0, 1.0.1
Change log ChangeLog.md
Dependencies attoparsec, base (>=4.7 && <5), bytestring, containers, directory, exceptions, fakedata-parser, filepath, hashable, random, string-random, template-haskell, text, time, transformers, unordered-containers, vector, yaml [details]
License BSD-3-Clause
Copyright Sibi Prabakaran
Author Sibi Prabakaran
Maintainer sibi@psibi.in
Category Random, Fake, FakeData
Home page https://github.com/psibi/fakedata#readme
Bug tracker https://github.com/psibi/fakedata/issues
Source repo head: git clone https://github.com/psibi/fakedata
Uploaded by psibi at 2021-06-26T13:12:46Z
Distributions LTSHaskell:0.8.0, NixOS:0.8.0, Stackage:1.0.1
Downloads 4894 total (40 in the last 30 days)
Rating (no votes yet) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Hackage Matrix CI
Docs available [build log]
Last success reported on 2021-06-26 [all 1 reports]


[Index] [Quick Jump]


Maintainer's Corner

For package maintainers and hackage trustees


Readme for fakedata-1.0.1

[back to package description]

fakedata Hackage StackageNightly StackageLTS Build Status

Table of Contents


This library is a port of Ruby's faker. It's a library for producing fake data such as names, addressess and phone numbers. Note that it directly uses the source data from that library, so the quality of fake data is quite high!

This packages comes in handy when you have to generate large amount of real like data for various purposes. I have personally used it for websites where it needs some realistic data in the initial stage, loading database with real like values etc. There are companies which have used this for sophisphicated testing purpose.

Additionly, there are two other packages for creating generators which is useful property testing:


Generating address

~/g/fakedata (master) $ stack ghci
λ> import Faker
λ> import Faker.Address
λ> address <- generate fullAddress
λ> address
"Apt. 298 340 Ike Mission, Goldnertown, FL 19488-9259"

Generating name

λ> fullName <- generate name
λ> fullName
"Sherryl Steuber"

Generate quotes from the movie Back to the Future

λ> import Faker.Movie.BackToTheFuture
λ> import Faker.Combinators
λ> qs <- generateNonDeterministic $ listOf 5 quotes
λ> qs
[ "Yes. Yes. I'm George. George McFly. I'm your density. I mean, your destiny."
, "Hello? Hello? Anybody home? Huh? Think, McFly. Think! I gotta have time to get them retyped. Do you realize what would happen if I hand in my reports in your handwriting? I'll get fired. You wouldn't want that to happen, would ya? Would ya?"
, "Lorraine. My density has brought me to you."
, "See you in about 30 years."
, "You really think I ought to swear?"

Combining Fake datas

{-#LANGUAGE RecordWildCards#-}

import Faker
import Faker.Name
import Faker.Address
import Data.Text

data Person = Person {
    personName :: Text,
    personAddress :: Text
} deriving (Show, Eq)

fakePerson :: Fake Person
fakePerson = do
    personName <- name
    personAddress <- fullAddress
    pure $ Person{..}

main :: IO ()
main = do
    person <- generate fakePerson
    print person

And on executing them:

$ stack name.hs
  { personName = "Sherryl Steuber"
  , personAddress = "Apt. 298 340 Ike Mission, Goldnertown, FL 19488-9259"

You would have noticed in the above output that the name and address are the same as generated before in the GHCi REPL. That's because, by default all the generated data are deterministic. If you want a different set of output each time, you would have to modify the random generator output:

main :: IO ()
main = do
    gen <- newStdGen
    let settings = setRandomGen gen defaultFakerSettings
    person <- generateWithSettings settings fakePerson
    print person

And on executing the program, you will get a different output:

  { personName = "Ned Effertz Sr."
  , personAddress = "Suite 158 1580 Schulist Mall, Schulistburgh, NY 15804-3392"

The above program can be even minimized like this:

main :: IO ()
main = do
    let settings = setNonDeterministic defaultFakerSettings
    person <- generateWithSettings settings fakePerson
    print person

Or even better:

main :: IO ()
main = do
    person <- generateNonDeterministic fakePerson
    print person



λ> import Faker.Address
λ> item <- generateNonDeterministic $ listOf 5 country
λ> item
["Ecuador","French Guiana","Faroe Islands","Canada","Armenia"]


λ> item <- generate $ oneof [country, fullAddress]
λ> item
"Suite 599 599 Brakus Flat, South Mason, MT 59962-6876"


λ> import qualified Faker.Address as AD
λ> item :: Text <- generate $ suchThat AD.country (\x -> (T.length x > 5))
λ> item
λ> item :: Text <- generate $ suchThat AD.country (\x -> (T.length x > 8))
λ> item
"French Guiana"

For seeing the full list of combinators, see the module documentation of Faker.Combinators.

Using the FakeT transformer

When generating values, you may want to perform some side-effects.

import Control.Monad.IO.Class
import Control.Monad.Logger
import Data.Text
import Data.Text.IO
import Faker.ChuckNorris

logQuote :: (MonadIO m, MonadLogger m) => m ()
logQuote = do
  userName <- liftIO getLine
  quote <- generateNonDeterministic fact
  $(logInfo) $ "Chuck Norris" userName quote

This works fine for one-off generation - but if you try to repeatedly generate values, you will run into performance trouble.

import Control.Monad (replicateM)

slowFunction :: (MonadIO m, MonadLogger m) => m ()
slowFunction = replicateM 1000 logQuote

This is because generating a Fake parses the data files and builds a cache for future use. Using the Monad instance on Fake shares that cache between Fakes, making faking fast. But in the above code, a new Fake is generated each time - so the cache is discarded, and performance is much worse.

It's better to use the FakeT monad transformer when writing such code, to get the benefits of sharing the cache, as well as being able to perform side effects. FakeT comes with the mtl-style MonadFake class, for easy use with your monad stack, which lets you lift Fakes with liftFake.

import Faker.Class

betterLogQuote :: (MonadIO m, MonadLogger m, MonadFake m) => m ()
betterLogQuote = do
  userName <- liftIO getLine
  quote <- liftFake fact
  $(logInfo) $ "Chuck Norris" userName quote

slowFunction can be rewritten to be much faster, because the FakeT is shared between all the calls to fact.

fastFunction :: (MonadIO m, MonadLogger m) => m ()
fastFunction = generateNonDeterministic go
    go :: FakeT m ()
    go = replicateM 1000 logQuote

Comparision with other libraries

There are two other libraries in the Hackage providing fake data:

The problem with both the above libraries is that the library covers only a very small amount of fake data source. I wanted to have an equivalent functionality with something like faker. Also, most of the combinators in this packages has been inspired (read as taken) from the fake library. Also, fakedata offers fairly good amount of support of different locales. Also since we rely on an external data source, we get free updates and high quality data source with little effort. Also, it's easier to extend the library with it's own data source if we want to do it that way.


Benjamin Curtis for his Ruby faker library from which the data source is taken from.

Icons made by Freepik from Flaticon.