
no the use case is to issue a bunch of disparate select statements,
i.e. retrieving data from several tables in a single roundtrip.
select 1,2,3
select 'a', 'b'
In this case, I'd like to call the API in this manner (or something better):
-- statement is prepared
rows1 <- fetchAllRows statement
_ <- nextResults statement
rows2 <- fetchAllRows statement
-- continue
right now, with hdbc-odbc (nextResults doesn't exist though so I don't
call it) rows2 will be an empty list, I'd expect it to be[(SqlText
"a"), (SqlText "b")]
I assume in the case of odbc, the related C function is SQLMoreResults
http://msdn.microsoft.com/en-us/library/ms714673%28v=vs.85%29.aspx
On Wed, Oct 16, 2013 at 3:16 PM, Alexey Uimanov
If I understand correctly you just want to re-execute the same query multiple times. There is method `reset` in HDBI already. Just call `reset` and statement return to it's initial state just after `prepare`, then execute `fetchAll` to get the result. Or just create new statement with `prepare` using the same query string or getting it from old statement with `originalQuery`.
2013/10/16 Gauthier Segay
use `runFetchAll` it will return constructed Seq with results. But I would use conduits and conduit `selectAll` http://hackage.haskell.org/package/hdbi-conduit-1.3.0/docs/Data-Conduit-HDBI.... Conduits are realy simple and effective.
I'm unsure I expressed the question properly, multiple results (rows) is definitely supported but multiple resultset does not seem to be.
There are occurrences where you issue multiple select statements in a single roundtrip to the database, each select with potentially different row layout.
After reading your answer, I actually tried (hdbc) calling the fetchRows function several times on Statement but it won't return anything past the first resultset.
It seems the .net NextResults() approach is good because it let's you check whether or not there is a next resultset to fetch rows from, I think it would be necessary to have a similar approach if this is going to be supported.
On Wed, Oct 16, 2013 at 6:03 AM, Alexey Uimanov
wrote: ability to use named parameters
Yes. I dont know when It will be done, but it is on github issues already https://github.com/s9gf4ult/hdbi/issues/3 .
fetching multiple results returned from single statement
use `runFetchAll` it will return constructed Seq with results. But I would use conduits and conduit `selectAll`
http://hackage.haskell.org/package/hdbi-conduit-1.3.0/docs/Data-Conduit-HDBI.... Conduits are realy simple and effective.
If you need something like nextresult just use method `fetch` of `Statement` until it return Nothing. It is just the same. Or use ResumableSink from conduits (I would).
2013/10/16 Gauthier Segay
Thanks for the announcement / library, I have started using hdbc (and it's odbc driver) recently and had two concerns with it so far:
* ability to use named parameters
(http://stackoverflow.com/questions/19137803/does-database-hdbc-support-named...) * fetching multiple results returned from single statement
(http://stackoverflow.com/questions/19159287/hdbc-and-multiple-resultsets-in-...) (in .net this is done via
http://msdn.microsoft.com/en-us/library/system.data.idatareader.nextresult.a...)
Is there any plan to get this supported in HDBI?
On Tue, Oct 15, 2013 at 7:55 PM, Alexey Uimanov
wrote: Hello haskellers!
HDBI is the fork of HDBC but reworked. It supports SQlite3 and Postgresql for now. It also supports streaming with coduits. There is TH deriving mechanism to map database rows to Haskell structures and back. HDBI is trying to become simple but still powerfull and flexible database interface. It must be suitable to become the common RDBMS interface for higher level interfaces like persistent or haskelldb.
The documentation is not very good, while I have no enough time to make some.
In this version changed typeclass signatures of Connection and Statement. Now methods `run` and `execute` get any instance of `ToRow` and method `fetch` return an instance of `FromRow`. Note that [SqlValue] is also an instance of `FromRow` and `ToRow` typeclasses so you do not loose the control. Methods `fromRow` and `toRow` for [SqlValue] are just `id`. SQlite and Postgresql drivers are fixed as well as hdbi-conduit.
There is also new helper functions, like `onei :: Integer -> [SqlValue]` which helps you to execute queries with one parameter or execute many queries consistinf of one parameter.
Prelude Database.HDBI Database.HDBI.SQlite> :set -XScopedTypeVariables Prelude Database.HDBI Database.HDBI.SQlite> :set -XOverloadedStrings Prelude Database.HDBI Database.HDBI.SQlite> c <- connectSqlite3 ":memory:" Prelude Database.HDBI Database.HDBI.SQlite> runRaw c "create table test(val integer)" Prelude Database.HDBI Database.HDBI.SQlite> withTransaction c $ runMany c "insert into test(val) values (?)" $ map one [1..1000]
<interactive>:7:76: Warning: Defaulting the following constraint(s) to type `Integer' .........
Prelude Database.HDBI Database.HDBI.SQlite> r :: (Maybe Integer) <- runFetchOne c "select sum(val) from test" () Prelude Database.HDBI Database.HDBI.SQlite> r Just 500500
Note here that the empty set is used as a parameter of query in `runFetchOne`. Empty set is an instance of `FromRow` and `ToRow` and return an empty list of [SqlValue]. Use empty list as a parameters is bad idea, because we could instantiate some another list of things, suppose the [Integer] as `FromRow` and `ToRow` instance. So it would lead to ambigous type because [SqlValue] is also a list instantiating `FromRow` and `ToRow`.
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe