HDBC-1.1.4: Haskell Database Connectivity

MaintainerJohn Goerzen <jgoerzen@complete.org>




Welcome to HDBC, the Haskell Database Connectivity library.

Written by John Goerzen, jgoerzen@complete.org



Welcome to HDBC, Haskell Database Connectivity.

HDBC provides an abstraction layer between Haskell programs and SQL relational databases. This lets you write database code once, in Haskell, and have it work with any number of backend SQL databases (MySQL, Oracle, PostgreSQL, ODBC-compliant databases, etc.)

HDBC is modeled loosely on Perl's DBI interface http://search.cpan.org/~timb/DBI/DBI.pm, though it has also been influenced by Python's DB-API v2, JDBC in Java, and HSQL in Haskell.

HDBC is a from-scratch effort. It is not a reimplementation of HSQL, though its purpose is the same.


Features of HDBC include:

  • Ability to use replacable parameters to let one query be executed multiple times (eliminates the need for an escape function)
  • Ability to access returned rows by column number
  • Ability to read data from the SQL server on-demand rather than reading the entire result set up front
  • HUnit testsuite for each backend driver
  • Well-defined standard API and easy backend driver implementation
  • Lazy reading of the entire result set (think hGetContents, but for the results of SELECT) (see sFetchAllRows)
  • Support for translation between Haskell and SQL types
  • Support for querying database server properties
  • Add-on package (hdbc-missingh) to integrate with MissingH, providing a database backend for AnyDBM.
  • Support for querying metadata such as column names.
  • Support for querying additional metadata (column types, etc.)

Available Drivers

Here is a list of known drivers as of March 28, 2006:

Sqlite v3
Available from http://software.complete.org/hdbc-sqlite3. Or, to participate in development, use darcs get --partial http://darcs.complete.org/hdbc-sqlite3
Available from http://software.complete.org/hdbc-postgresql. Or, to participate in development, use darcs get --partial http://darcs.complete.org/hdbc-postgresql
Available from http://software.complete.org/hdbc-odbc. Or, to partitipace in development, use darcs get --partial http://darcs.complete.org/hdbc-odbc
MySQL users are encouraged to use the ODBC driver, which works and has been tested against MySQL on both Linux/Unix and Windows platforms.

In addition, there is one integration package: hdbc-anydbm. This integrates with the AnyDBM library http://software.complete.org/anydbm. It lets any HDBC database act as a backend for the AnyDBM interface. Available from http://software.complete.org/hdbc-anydbm. Or, to participate in development, use darcs get --partial http://darcs.complete.org/hdbc-anydbm

The latest version of HDBC itself is available from http://software.complete.org/hdbc. Or, to participate in development, use darcs get --partial http://darcs.complete.org/hdbc.

Typing of transfer data

class Show a => SqlType a whereSource

Conversions to and from SqlValues and standard Haskell types.

Conversions are powerful; for instance, you can call fromSql on a SqlInt32 and get a String or a Double out of it. This class attempts to Do The Right Thing whenever possible, and will raise an error when asked to do something incorrect. In particular, when converting to any type except a Maybe, SqlNull as the input will cause an error to be raised.

Here are some notes about conversion:

  • Fractions of a second are not preserved on time values

See also nToSql, iToSql.

nToSql :: Integral a => a -> SqlValueSource

Converts any Integral type to a SqlValue by using toInteger.

iToSql :: Int -> SqlValueSource

Convenience function for using numeric literals in your program.

data SqlValue Source

The main type for expressing Haskell values to SQL databases.

This type is used to marshall Haskell data to and from database APIs. HDBC driver interfaces will do their best to use the most accurate and efficient way to send a particular value to the database server.

Values read back from the server are put in the most appropriate SqlValue type. fromSql can then be used to convert them into whatever type is needed locally in Haskell.

Most people will use toSql and fromSql instead of manipulating SqlValues directly.

The default representation of time values is an integer number of seconds. Databases such as PostgreSQL with builtin timestamp types can will see automatic conversion between these Haskell types to local types. Other databases can just use an int or a string.

This behavior also exists for other types. For instance, many databases don't have a Rational type, so they'll just use Haskell's show function and store a Rational as a string.

Two SqlValues are considered to be equal if one of these hold (first one that is true holds; if none are true, they are not equal): * Both are NULL * Both represent the same type and the encapsulated values are equal * The values of each, when converted to a string, are equal.


SqlString String 
SqlByteString ByteString 
SqlWord32 Word32 
SqlWord64 Word64 
SqlInt32 Int32 
SqlInt64 Int64 
SqlInteger Integer 
SqlChar Char 
SqlBool Bool 
SqlDouble Double 
SqlRational Rational 
SqlEpochTime Integer

Representation of ClockTime or CalendarTime

SqlTimeDiff Integer

Representation of TimeDiff


NULL in SQL or Nothing in Haskell


Database Connections

class IConnection conn whereSource

Main database handle object.

An IConnection object is created by specific functions in the module for an individual database. That is, the connect function -- which creates this object -- is not standardized through the HDBC interface.

A connection is closed by a call to disconnect.

A call to commit is required to make sure that your changes get committed to the database. In other words, HDBC has no support for autocommit, which we consider an outdated notion.


disconnect :: conn -> IO ()Source

Disconnect from the remote database.

You do not need to explicitly close an IConnection object, but you may do so if you so desire. If you don't, the object will disconnect from the database in a sane way when it is garbage-collected. However, a disconnection may raise an error, so you are encouraged to explicitly call disconnect. Also, garbage collection may not run when the program terminates, and some databases really like an explicit disconnect.

So, bottom line is, you're best off calling disconnect directly, but the world won't end if you forget.

This function discards any data not committed already. Database driver implementators should explicitly call rollback if their databases don't do this automatically on disconnect.

Bad Things (TM) could happen if you call this while you have Statements active. In more precise language, the results in such situations are undefined and vary by database. So don't do it.

commit :: conn -> IO ()Source

Commit any pending data to the database.

Required to make any changes take effect.

rollback :: conn -> IO ()Source

Roll back to the state the database was in prior to the last commit or rollback.

run :: conn -> String -> [SqlValue] -> IO IntegerSource

Execute a single SQL query. Returns the number of rows modified (see execute for details). The second parameter is a list of replacement values, if any.

prepare :: conn -> String -> IO StatementSource

Prepares a statement for execution.

Question marks in the statement will be replaced by positional parameters in a later call to execute.

Please note that, depending on the database and the driver, errors in your SQL may be raised either here or by execute. Make sure you handle exceptions both places if necessary.

clone :: conn -> IO connSource

Create a new Connection object, pointed at the same server as this object is. This will generally establish a separate physical connection.

When you wish to establish multiple connections to a single server, the correct way to do so is to establish the first connection with the driver-specific connection function, and then clone it for each additional connection.

This can be important when a database doesn't provide much thread support itself, and the HDBC driver module must serialize access to a particular database.

This can also be a handy utility function whenever you need a separate connection to whatever database you are connected to already.

hdbcDriverName :: conn -> StringSource

The name of the HDBC driver module for this connection. Ideally would be the same as the database name portion of the Cabal package name. For instance, "sqlite3" or "odbc". This is the layer that is bound most tightly to HDBC.

hdbcClientVer :: conn -> StringSource

The version of the C (or whatever) client library that the HDBC driver module is bound to. The meaning of this is driver-specific. For an ODBC or similar proxying driver, this should be the version of the ODBC library, not the eventual DB client driver.

proxiedClientName :: conn -> StringSource

In the case of a system such as ODBC, the name of the database client/server in use, if available. For others, identical to hdbcDriverName.

proxiedClientVer :: conn -> StringSource

In the case of a system such as ODBC, the version of the database client in use, if available. For others, identical to hdbcClientVer. This is the next layer out past the HDBC driver.

dbServerVer :: conn -> StringSource

The version of the database server, if available.

dbTransactionSupport :: conn -> BoolSource

Whether or not the current database supports transactions. If False, then commit and rollback should be expected to raise errors.

MySQL is the only commonly-used database that is known to not support transactions entirely. Please see the MySQL notes in the ODBC driver for more information.

getTables :: conn -> IO [String]Source

The names of all tables accessible by the current connection, excluding special meta-tables (system tables).

You should expect this to be returned in the same manner as a result from Database.HDBC.fetchAllRows'.

All results should be converted to lowercase for you before you see them.

describeTable :: conn -> String -> IO [(String, SqlColDesc)]Source

Obtain information about the columns in a specific table. The String in the result set is the column name.

You should expect this to be returned in the same manner as a result from Database.HDBC.fetchAllRows'.

All results should be converted to lowercase for you before you see them.

Wrapped Connections

data ConnWrapper Source

Sometimes, it is annoying to use typeclasses with Haskell's type system. In those situations, you can use a ConnWrapper. You can create one with:

let wrapped = ConnWrapper iconn

You can then use this directly, since a ConnWrapper is also an IConnection. However, you will not be able to use private database functions on it.

Or, you can use withWConn.

withWConn :: forall b. ConnWrapper -> (forall conn. IConnection conn => conn -> b) -> bSource

Unwrap a ConnWrapper and pass the embedded IConnection to a function. Example:

withWConn wrapped run $ "SELECT * from foo where bar = 1" []

Preparing Queries

sRun :: IConnection conn => conn -> String -> [Maybe String] -> IO IntegerSource

Like run, but take a list of Maybe Strings instead of SqlValues.

quickQuery' :: IConnection conn => conn -> String -> [SqlValue] -> IO [[SqlValue]]Source

Strict version of quickQuery.

quickQuery :: IConnection conn => conn -> String -> [SqlValue] -> IO [[SqlValue]]Source

A quick way to do a query. Similar to preparing, executing, and then calling fetchAllRows on a statement. See also quickQuery'

Transaction Handling

This section concerns itself with writing (updating) a database.

In HDBC, as with many RDBMS implementations, every write to the database occurs within a transaction. No changes are visible (outside the current transaction) until a commit operation occurs, in which case all changes since the transaction started are atomically committed. Also, there is a rollback operation that can undo all changes since the transaction started.

HDBC does everything within a transaction. A transaction is implicitly entered when a connection to a database is established, and a transaction is implicitly entered after each call to commit or rollback as well.

The practical effect of this is that you must call commit after making changes to a database in order for those changes to become visible. You don't have to call commit after every change, just after a batch of them.

(Exceptions exist for databases that don't offer a high level of transaction isolation; but you should always play it safe and commit anyway.)

Database developers will also be experienced with the atomicity benefits of transactions, an explanation of which is outside the scope of this manual.

Errors occuring at the database level can leave a transaction in an indeterminate state, depending on the database. Some databases will refuse all queries until the next commit or rollback. The safe thing to do is to issue a commit or rollback after trapping any SqlError. Alternatively, you could use withTransaction, which will automatically handle this detail for you.

withTransaction :: IConnection conn => conn -> (conn -> IO a) -> IO aSource

Execute some code. If any uncaught exception occurs, run rollback and re-raise it. Otherwise, run commit and return.

This function, therefore, encapsulates the logical property that a transaction is all about: all or nothing.

The IConnection object passed in is passed directly to the specified function as a convenience.

This function traps all uncaught exceptions, not just SqlErrors. Therefore, you will get a rollback for any exception that you don't handle. That's probably what you want anyway.

Since all operations in HDBC are done in a transaction, this function doesn't issue an explicit "begin" to the server. You should ideally have called Database.HDBC.commit or Database.HDBC.rollback before calling this function. If you haven't, this function will commit or rollback more than just the changes made in the included action.

If there was an error while running rollback, this error will not be reported since the original exception will be propogated back. (You'd probably like to know about the root cause for all of this anyway.) Feedback on this behavior is solicited.

Connection Inquiries



execute :: Statement -> [SqlValue] -> IO IntegerSource

Execute the prepared statement, passing in the given positional parameters (that should take the place of the question marks in the call to prepare).

For non-SELECT queries, the return value is the number of rows modified, if known. If no rows were modified, you get 0. If the value is unknown, you get -1. All current HDBC drivers support this function and should never return -1.

For SELECT queries, you will always get 0.

This function should automatically call finish() to finish the previous execution, if necessary.

sExecute :: Statement -> [Maybe String] -> IO IntegerSource

Like execute, but take a list of Maybe Strings instead of SqlValues.

executeMany :: Statement -> [[SqlValue]] -> IO ()Source

Execute the query with many rows. The return value is the return value from the final row as if you had called execute on it.

Due to optimizations that are possible due to different databases and driver designs, this can often be significantly faster than using execute multiple times since queries need to be compiled only once.

This is most useful for non-SELECT statements.

sExecuteMany :: Statement -> [[Maybe String]] -> IO ()Source

Like executeMany, but take a list of Maybe Strings instead of SqlValues.

Fetching Results

fetchRow :: Statement -> IO (Maybe [SqlValue])Source

Fetches one row from the DB. Returns Nothing if there are no more rows. Will automatically call finish when the last row is read.

fetchRowAL :: Statement -> IO (Maybe [(String, SqlValue)])Source

Like fetchRow, but instead of returning a list, return an association list from column name to value.

The keys of the column names are lowercase versions of the data returned by getColumnNames. Please heed the warnings there. Additionally, results are undefined if multiple columns are returned with identical names.

fetchRowMap :: Statement -> IO (Maybe (Map String SqlValue))Source

Similar to fetchRowAL, but return a Map instead of an association list.

sFetchRow :: Statement -> IO (Maybe [Maybe String])Source

Like fetchRow, but return a list of Maybe Strings instead of SqlValues.

fetchAllRows :: Statement -> IO [[SqlValue]]Source

Lazily fetch all rows from an executed Statement.

You can think of this as hGetContents applied to a database result set.

The result of this is a lazy list, and each new row will be read, lazily, from the database as the list is processed.

When you have exhausted the list, the Statement will be finished.

Please note that the careless use of this function can lead to some unpleasant behavior. In particular, if you have not consumed the entire list, then attempt to finish or re-execute the statement, and then attempt to consume more elements from the list, the result will almost certainly not be what you want.

But then, similar caveats apply with hGetContents.

Bottom line: this is a very convenient abstraction; use it wisely.

Use fetchAllRows' if you need something that is strict, without all these caveats.

fetchAllRows' :: Statement -> IO [[SqlValue]]Source

Strict version of fetchAllRows. Does not have the side-effects of fetchAllRows, but forces the entire result set to be buffered in memory.

fetchAllRowsAL :: Statement -> IO [[(String, SqlValue)]]Source

Like fetchAllRows, but instead of returning a list for each row, return an association list for each row, from column name to value.

See fetchRowAL for more details.

fetchAllRowsMap :: Statement -> IO [Map String SqlValue]Source

Like fetchAllRowsAL, but return a list of Maps instead of a list of association lists.

sFetchAllRows :: Statement -> IO [[Maybe String]]Source

Like fetchAllRows, but return Maybe Strings instead of SqlValues.

getColumnNames :: Statement -> IO [String]Source

Returns a list of the column names in the result. For maximum portability, you should not assume that information is available until after an execute function has been run.

Information is returned here directly as returned by the underlying database layer. Note that different databases have different rules about capitalization of return values and about representation of names of columns that are not simple columns. For this reason, it is suggested that you treat this information for display purposes only. Failing that, you should convert to lower (or upper) case, and use AS clauses for anything other than simple columns.

A simple getColumnNames implementation could simply apply map fst to the return value of describeResult.

Statement Inquires

describeResult :: Statement -> IO [(String, SqlColDesc)]Source

Obtain information about the columns in the result set. Must be run only after execute. The String in the result set is the column name.

You should expect this to be returned in the same manner as a result from Database.HDBC.fetchAllRows'.

All results should be converted to lowercase for you before you see them.

Please see caveats under getColumnNames for information on the column name field here.


finish :: Statement -> IO ()Source

Abort a query in progress -- usually not needed.

originalQuery :: Statement -> StringSource

The original query that this Statement was prepared with.


data SqlError Source

The main HDBC exception object. As much information as possible is passed from the database through to the application through this object.

Errors generated in the Haskell layer will have seNativeError set to -1.



catchSql :: IO a -> (SqlError -> IO a) -> IO aSource

Execute the given IO action.

If it raises a SqlError, then execute the supplied handler and return its return value. Otherwise, proceed as normal.

handleSql :: (SqlError -> IO a) -> IO a -> IO aSource

Like catchSql, with the order of arguments reversed.

sqlExceptions :: Exception -> Maybe SqlErrorSource

Given an Exception, return Just SqlError if it was an SqlError, or Nothing otherwise. Useful with functions like catchJust.

handleSqlError :: IO a -> IO aSource

Catches SqlErrors, and re-raises them as IO errors with fail. Useful if you don't care to catch SQL errors, but want to see a sane error message if one happens. One would often use this as a high-level wrapper around SQL calls.

Column Types

These are defined in Database.HDBC.ColTypes but are available to programs importing Database.HDBC by default as well. See Database.HDBC.ColTypes for documentation.


FIXME: this is draft information

Thread support in a generalized interface such as HDBC can be complicated because support for threading varies across database interfaces.

However, applications using HDBC should be able to rely upon at least a few basic guarantees:

  • The HDBC modules may freely be imported and used across all threads.
  • HDBC modules may also freely share database connections and statements; the database or HDBC driver will be responsible for locking if necessary.

I use "share" in the same sense as Python's DB-API: multiple threads may use the resource without wrapping it in any lock.

However, there are some caveats to the above:

  • Not all databases support more than one active statement for a single connection. Therefore, for maximum portability, you should use a different connection to the database for each simultaneous query you wish to use. FIXME: describe when a statement is active.
  • Not all databases may support the level of multithreading described above. For those that don't, safe access will be restriced in the HDBC driver by using locks. Therefore, you can write portable code, but you only get real multithreading when databases really support it. Details of thread support should be documented in the HDBC driver for each specific database.

Copyright and License

Copyright (C) 2005-2007 John Goerzen jgoerzen@complete.org

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version.

This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License along with this library; if not, write to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA

Please see COPYING in the source distribution for the full license.