The intent is to develop a high quality, high level driver similar to pycassa.
Please see the Haddocks hosted on HackageDB:
This module offers low-level functionality that is still much more pleasant than using Thrift directly.
import Database.Cassandra.Basic test :: IO () test = do pool <- createCassandraPool defServers 3 300 "TestKeySpace" insert pool "testCF" "key1" QUORUM [col "col1" "value1"] getCol pool "testCF" "key1" "col1" QUORUM
This is the primary high level functionality module. Its use is recommended above the other options.
- Columns can be any Haskell type with a CasType instance. See
Database.Cassandra.Packfor what you can use there out of the box.
- You can choose how to encode/decode your column content. Out of the box, we support Show/Read, ToJSON/FromJSON, Serialize and SafeCopy.
Example usage: JSON-encoded columns
import Database.Cassandra.Marshall import Data.Aeson type Name = String type Age = Int data Person = Person Name Age -- Define JSON serialization for our data structure instance ToJSON Person where toJSON (Person nm age) = toJSON (nm,age) instance FromJSON Person where parseJSON v = do (nm, age) <- parseJSON v return $ Person nm age test :: Person -> IO () test p@(Person nm age) = do pool <- createCassandraPool defServers 3 300 "TestKeySpace" -- I can use any string-like key and don't have to explicitly -- convert person to ByteString. runCas pool $ insertCol casJSON "testCF" "people" nm QUORUM p -- Automatically de-serialized back to a datatype res <- runCas pool $ getCol casJSON "testCF" "people" nm QUORUM case res of Just (Person nm age) -> return age Nothing -> error "Oh NO!!!"
- We now use retry >= 0.4 and hence exceptions machinery instead of monad-control.
- Updated to work with monad-control instead of MonadCatchIO.
- Now using retry >= 0.3 for the same reason.
- More CasType instances in Pack module for completeness. Preference is shifting to using native Haskell types directly.
Database.Cassandra.JSONmodule is now formally deprecated and will be removed in 0.7. Please switch to
Marshall, which can replace it in entirety.
Database.Cassandra.Marshallthat is now intended to be the primary module to be used in all Cassandra operations. This module supercedes and replaces the
Database.Cassandra.JSONhigh level module. Building on top of the Basic module, Marshall allows user to pick the serialization methodology for each of the operations. We provide out of box support for JSON, cereal, safecopy and plain old show/read.
- A new
TTimeStamptype makes it easier to have timestamps as Long-encoded columns.
- There is now simple support for pagination of columns in wide rows,
CPS-style. See the
- Numerous other fixes and tweaks
This version packs a fairly large changeset. It will almost definitely break your code, although the fix/adjustment is likely to be minor.
- Vastly enhanced the Database.Cassandra.Pack module to represent types that Cassandra can sort and validate.
- Added CasType typeclass that offers
decodeCasmethods that handle conversions to/from the binary ByteString representation.
- Composite columns are now supported through tuples. Just pick two or more CasType instances and put them in a tuple to automatically trigger composite column conversion. Keep in mind that your ColumnFamily schema must be configured right or else you'll get runtime exceptions.
- Added a bunch of newtype wrappers to directly map to types Cassandra
knows. These include
TUtf8and some others.
- Changed several methods in Basic and JSON modules to expect CasType column key values instead of concrete ByteString.
- Added the useful
unpackColfunctions to smoothly handle column key type conversions when working with the Basic module.
- Made the Cas monad a simple type synonym for ReaderT CPool.
- Added the
JSONto make it easier to discard key names and just get the column contents.
- Numerous fixes and minor tweaks.
- Added MonadCassandra typeclass, which is now used by all operations by default.
- Added a default Cas moand that instantiates MonadCassandra for convenience.
- All Basic module ops now return results directly instead of an Either wrapper. Each operation may raise a CassandraException.
- Connection pooling now builds on top of the resource-pool library to initiate connections to multiple servers in round-robin fashion.
- Basic.insert now knows how to insert a SuperColumn.
- Add support for counters and batch mutators
- Add support for database/admin operations
Would love to get contributions, bug reports, suggestions, feedback from the community.