Skip to content

zmactep/hasbolt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

96 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

HasBOLT

Travis GitHub Build hackage hackage-deps

Haskell driver for Neo4j 3+ (BOLT protocol)

Documentation

To build Haddock documentation run:

$ stack haddock

Usage example

{-# LANGUAGE OverloadedStrings #-}

import Database.Bolt

import Data.Default
import Data.Text
import Control.Monad
import Control.Monad.Except

-- Simple request can be done by using 'query' function. It returns a list of 'Record's which
-- are special dictionaries from 'Text' to any serializable 'Value'. You can extract this values by key using 'at' function.
nineties :: BoltActionT IO [Text]
nineties = do records <- query "MATCH (nineties:Movie) WHERE nineties.released >= 1990 AND nineties.released < 2000 RETURN nineties.title"
              forM records $ \record -> record `at` "nineties.title"

-- All types that can be processed by 'at' are instances of 'RecordValue' classtype.
-- You can implement new unpackers by your own.
--
-- If you want to perform some request multiple times with neo4j caching speedup,
-- you can use 'queryP' function that takes not only the Cypher request but also
-- a parameters dictionary.
genericABN :: RecordValue a => Text -> BoltActionT IO [a]
genericABN name = do toms' <- queryP "MATCH (tom:Person) WHERE tom.name CONTAINS {name} RETURN tom"
                                     (props ["name" =: name])
                     nodes <- forM toms' $ \record -> record `at` "tom"
                     forM nodes $ \node -> nodeProps node `at` "name"

-- Hasbolt has a special 'Node' type to unpack graph nodes. You also can find 'Relationship',
-- 'URelationship' and 'Path' as built-in types.
actorsByNameYear :: Text -> Int -> BoltActionT IO [Node]
actorsByNameYear name year = do toms' <- queryP "MATCH (n:Person {name: {props}.name, born: {props}.born}) RETURN n" 
                                                (props ["props" =: props ["name" =: name, "born" =: year]])
                                forM toms' $ \record -> record `at` "n"

actorsByName :: Text -> BoltActionT IO [Text]
actorsByName = genericABN

-- This request raises 'WrongMessageFormat' error, as it cannot unpack 'Text' values as 'Int's.
wrongType :: Text -> BoltActionT IO [Int]
wrongType = genericABN 

-- Database server answers with a 'ResponseError' exception on any syntax error or internal database problem.
typoInRequest :: Text -> BoltActionT IO [Text]
typoInRequest name = do toms' <- queryP "MATCH (tom:Person) WHERE tom.name CONTAINS {name} RETURN not_tom"
                                        (props ["name" =: name])
                        nodes <- forM toms' $ \record -> record `at` "tom"
                        forM nodes $ \node -> nodeProps node `at` "name"

-- 'RecordHasNoKey' is thrown in case of a wrong key usage in 'at'.
typoInField :: Text -> BoltActionT IO [Text]
typoInField name = do toms' <- queryP "MATCH (tom:Person) WHERE tom.name CONTAINS {name} RETURN tom" 
                                      (props ["name" =: name])
                      nodes <- forM toms' $ \record -> record `at` "not_tom"
                      forM nodes $ \node -> nodeProps node `at` "name"

main :: IO ()
main = do pipe <- connect $ def { user = "neo4j", password = "12345" }
          -- Prints nineties example from Movies tutorial
          putStrLn "Movies (nineties):" 
          titles <- run pipe nineties 
          forM_ titles print
          -- Prints all actors called Tom 
          putStrLn "\nActors (Tom):"
          toms' <- run pipe $ actorsByName "Tom" 
          forM_ toms' print
          -- Prints Tom Hanks 
          putStrLn "\nNodes (Tom Cruise):"
          nodes' <- run pipe $ actorsByNameYear "Tom Cruise" 1962 
          forM_ nodes' print
           -- Prints an error as type is wrong 
          putStrLn "\nWrong return type:"
          wtype' <- run pipe $ wrongType "Tom" `catchError`
                                 \e@(WrongMessageFormat _) -> liftIO (print e) >> pure [] 
          forM_ wtype' print
          -- Prints an error as the request contains a typo
          putStrLn "\nTypo in request:"
          typor' <- run pipe $ typoInRequest "Tom" `catchError`
                                 \(ResponseError e) -> liftIO (print e) >> pure [] 
          forM_ typor' print
          -- Prints an error as the field name contains a typo
          putStrLn "\nTypo in field:"
          typof' <- run pipe $ typoInField "Tom" `catchError` 
                                 \e@(RecordHasNoKey _) -> liftIO (print e) >> pure [] 
          forM_ typof' print
          close pipe

Notes

  • Do not forget to import Data.Default to use default connection configuration.
  • OverloadedStrings are very welcome, as the library doesn't use Strings at all.
  • You can use Database.Bolt.Lazy to work with lazy IO. In this case do not forget to read all the records before you send a next query.
  • See test/TransactionSpec.hs for an example of transactions usage.
  • Feel free to implement your own serialization procedures with Database.Bolt.Serialization module import.
  • Pipes work great with resource-pool.
  • For neo4j 3.4+ use version = 2 in connection configuration. This allows you to use new datatypes.
  • You can use both syntax variants to create properties dictionaries: fromList [("born", I 1962)] or props ["born" =: 1962].
  • Note that you have to make a type hint for Text values in the second construction, as Haskell cannot deduce it on its own.

New types

Neo4j 3.4+ implements BOLT v2 protocol (that still doesn't have any specification). Code inspection of neo4j sources led me to these new data types in v2. All of them are just structures with different signatures and fields.

  • Point2D
signature = 'X'
fields = { crs :: CoordinateReferenceSystem
         , x   :: Double
         , y   :: Double
         }
  • Point3D
signature = 'Y'
fields = { crs :: CoordinateReferenceSystem
         , x   :: Double
         , y   :: Double
         , z   :: Double
         }
  • Duration
signature = 'E'
fields = { months  :: Integer
         , days    :: Integer
         , seconds :: Integer
         , nanos   :: Integer
         }
  • Date
signature = 'D'
fields = { epochDays :: Integer
         }
  • Time
signature = 'T'
fields = { nanosOfDayLocal :: Integer
         , offsetSeconds   :: Integer
         }
  • LocalTime
signature = 't'
fields = { nanosOfDay :: Integer
         }
  • LocalDateTime
signature = 'd'
fields = { epochSeconds :: Integer
         , nano         :: Integer
         }
  • DateTimeWithZoneOffset
signature = 'F'
fields = { epochSecondsLocal :: Integer
         , nano              :: Integer
         , offsetSeconds     :: Integer
         }
  • DateTimeWithZoneName
signature = 'f'
fields = { epochSecondsLocal :: Integer
         , nano              :: Integer
         , zoneId            :: Integer
         }

Codes of Coordinate Reference Systems:

  • Cartesian 2D — 7203
  • Cartesian 3D — 9157
  • WGS-84 2D — 4326
  • WGS-84 3D — 4979

Example

λ> :set -XScopedTypeVariables 
λ> pipe <- connect $ def { user = "neo4j", password = "neo4j", version = 2 }
λ> point :: Value <- run pipe $ do records <- query "RETURN point({x: 1, y: 2, z: 3}) as point"
                                   (head records) `at` "point"
λ> point 
S (Structure {signature = 89, fields = [I 9157,F 1.0,F 2.0,F 3.0]})
λ> close pipe

Here 89 is an ASCII code for 'Y', 9157 shows default cartesian 3d coordinate reference system.