[Haskell-cafe] Announce: HDBI-1.3 and friends
Vagif Verdi
vagif.verdi at gmail.com
Wed Oct 16 16:12:25 UTC 2013
When discussing databases it is hard to understand each other without
knowing which exact database we are talking about.
I assume you are using MS Sql server (just like me) since you keep
mentioning dotnet.
There are few things you should know about specifics of haskell community
regarding windows and especially MS Sql server.
Both windows and MS Sql server are not first class citizens here.
Most haskellers use open source databases like Postgres and Mysql and they
use those from linux/mac, not from windows.
I had to fix a few bugs in hdbc-odbc driver that were specific to ms sql
server and freetds linux driver myself. Since maintainers of hdbc-odbc do
not even have a ms sql server to reproduce the problem and test it.
If you are looking for a specific Ms Sql Server feature, most likely you
will have to implement it yourself.
I'm trying to keep db access patterns really simple and do not use any
specific to Ms Sql features. And even in this case every time something
updates, like say freetds driver, everything breaks.
Just yesterday i fixed the problem with new freetds driver not fetching
scope_identity() anymore. I had to add set nocount on everywhere i use
scope_identity. Even though on previous version of either freetds or
unixodbc it was working fine.
So using Ms Sql server from haskell is still a shaky experience.
On Wednesday, October 16, 2013 8:49:59 AM UTC-7, Gauthier Segay wrote:
>
> Hello Vagif,
>
> It's definitely possible to work around this feature not being here;
> though I won't call it premature optimization as it's really
> convenient and logical to batch statements in a single roundtrip when
> you can do so, when the statements have coherence; I'm sure there are
> a lot of stored procedures in the wild that gives back multiple
> resultsets too.
>
> I'm sorry if it sounded like a plea to get this feature implemented,
> but I thought it might be worth considering in API design, it's there
> in jdbc (
> http://docs.oracle.com/javase/7/docs/api/java/sql/Statement.html#getMoreResults%28%29)
>
> and ado.net (
> http://msdn.microsoft.com/en-us/library/system.data.idatareader.nextresult.aspx).
>
>
> I appreciate the work being done on HDBI, I'll give it a go whenever
> I'll work with postgresql, and check-out new drivers coming, I'll keep
> an eye on it's evolution.
>
> Well at least I got the thing I was discussing understood as it wasn't
> clear in the initial exchanges due to my poor explanation :)
>
> On Wed, Oct 16, 2013 at 5:22 PM, Vagif Verdi <vagif... at gmail.com<javascript:>>
> wrote:
> > Over my more than 20 years career working with many different databases
> and
> > languages you are the first person i meet who requests such a feature.
> >
> > Premature optimization is harmful.
> >
> > You are looking for a solution to a problem that practically never
> manifests
> > itself.
> >
> > Especially in a server side application that are usually hosted close to
> sql
> > server and on a very fat bandwidth.
> >
> > In terms of the performance much more important question is connection
> > pooling.
> >
> > You are basically asking a single programmer who just started a quite
> big
> > project to spend his time on a feature practically no one is interested
> with
> > while he has hundreds of much more pressing issues to implement.
> >
> > I'd say at this point your only course of action is to implement it
> > yourself, if it is really that important to you.
> >
> >
> >
> > On Wednesday, October 16, 2013 6:42:56 AM UTC-7, Gauthier Segay wrote:
> >>
> >> 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 <s9gf... at gmail.com>
> wrote:
> >> > 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 <gauthie... at gmail.com>
> >> >>
> >> >> > 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.html.
>
> >> >> > 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 <s9gf... at gmail.com>
> >> >> 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.html.
>
> >> >> > 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 <gauthie... at gmail.com>
> >> >> >>
> >> >> >> 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-parameters)
>
> >> >> >> * fetching multiple results returned from single statement
> >> >> >>
> >> >> >>
> >> >> >>
> >> >> >> (
> http://stackoverflow.com/questions/19159287/hdbc-and-multiple-resultsets-in-a-single-statement-only-first-resultset-returne)
>
> >> >> >> (in .net this is done via
> >> >> >>
> >> >> >>
> >> >> >>
> >> >> >>
> http://msdn.microsoft.com/en-us/library/system.data.idatareader.nextresult.aspx)
>
> >> >> >>
> >> >> >> Is there any plan to get this supported in HDBI?
> >> >> >>
> >> >> >> On Tue, Oct 15, 2013 at 7:55 PM, Alexey Uimanov <
> s9gf... at gmail.com>
> >> >> >> 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
> >> >> >> > Haskel... at haskell.org
> >> >> >> > http://www.haskell.org/mailman/listinfo/haskell-cafe
> >> >> >> >
> >> >> >
> >> >> >
> >> >> >
> >> >> > _______________________________________________
> >> >> > Haskell-Cafe mailing list
> >> >> > Haskel... at haskell.org
> >> >> > http://www.haskell.org/mailman/listinfo/haskell-cafe
> >> >> >
> >> >
> >> >
> >> >
> >> > _______________________________________________
> >> > Haskell-Cafe mailing list
> >> > Haskel... at haskell.org
> >> > http://www.haskell.org/mailman/listinfo/haskell-cafe
> >> >
> >> _______________________________________________
> >> Haskell-Cafe mailing list
> >> Haskel... at haskell.org
> >> http://www.haskell.org/mailman/listinfo/haskell-cafe
> >
> >
> > _______________________________________________
> > Haskell-Cafe mailing list
> > Haskel... at haskell.org <javascript:>
> > http://www.haskell.org/mailman/listinfo/haskell-cafe
> >
> _______________________________________________
> Haskell-Cafe mailing list
> Haskel... at haskell.org <javascript:>
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/haskell-cafe/attachments/20131016/c5779058/attachment-0001.html>
More information about the Haskell-Cafe
mailing list