I have the following query:
var result = _session.QueryOver<Entity>()
.Where(e => e.Property == value)
.SelectList(list => list
.Select(f => Projections.Concat("prefix-", e.BigIntProperty)).WithAlias(() => alias.Whatever)
...
)
.TransformUsing(Transformers.AliasToBean<Model>())
.Future<Model>();
The problem is that Projections.Concat() accepts only strings and since e.BigIntProperty is not, the above doesn't compile. Is there a way to cast e.BigIntProperty to string?
I tried something like the following, which doesn't work either:
.Select(f => Projections.Concat("prefix-", Projection.Cast(NHibernateUtil.String, e.BigIntProperty))).WithAlias(() => alias.Whatever)
, since Projections.Cast returns an IProjection and not a string.
Projections.Cast seems terribly limited in that it can't take arbitrary Projections. Luckily you can easily create your own custom projection that enables you to do that:
public static class CustomProjections
{
public static IProjection Concat(params IProjection[] projections)
{
return Projections.SqlFunction(
"concat",
NHibernateUtil.String,
projections);
}
}
Then, you'll be able to use your CustomProjections class like this:
var result = _session.QueryOver<Entity>()
.Where(e => e.Property == value)
.SelectList(list => list
.Select(CustomProjections.Concat(
Projections.Constant("prefix-"),
Projections.Cast(
NHibernateUtil.String,
Projections.Property<Entity>(e => e.BigIntProperty))))
.WithAlias(() => alias.Whatever)
...
)
.TransformUsing(Transformers.AliasToBean<Model>())
.Future<Model>();
I've already accepted Andrew's answer, but just for reference, you could use Projections.SqlFunction("concat", ...) directly which solves the whole issue since it can take IProjection's as arguments and not only string.
var result = _session.QueryOver<Entity>()
.Where(e => e.Property == value)
.SelectList(list => list
.Select(Projections.SqlFunction("concat",
NHibernateUtil.String,
Projections.Constant("prefix-"),
Projections.Cast(NHibernateUtil.String, Projections.Property<Entity>(e => e.BigIntProperty))))
.WithAlias(() => alias.Whatever)
...
)
.TransformUsing(Transformers.AliasToBean<Model>())
.Future<Model>();
NOTE: It seems that when calling either Projections.Concat(...) or Projections.SqlFunction("concat", ...), the query that is produced actually uses the + operator, e.g.:
SELECT (a + b) as foo FROM table
instead of:
SELECT concat(a, b) as foo FROM table
Of course, CONCAT is only available from MS SQL Server versions 2012 and above, so this is correct. Possibly the MsSQl2012Dialect could make use of the CONCAT, since CONCAT doesn't require that the arguments are varchar, they might as well be integers.
Unfortunately MsSQl2012Dialect doesn't do that, but it is very easy to build a custom Dialect:
public class CustomMsSql2012Dialect : MsSql2012Dialect
{
protected override void RegisterFunctions()
{
base.RegisterFunctions();
base.RegisterFunction("concat", new VarArgsSQLFunction(NHibernateUtil.String, "concat(", ",", ")"));
}
}
So, if you use version 2012 or above and you declare the above as your Dialect, you can ditch the Projections.Cast(...) part
Related
Finally tracked down my error which is a result of the query. I have an nhibernate query using a Restrictions.In. Problem is once query executes if no results returned query throws error immediately. Is there another restriction that I can use. I know if I was writing a linq query I could use the .Any to return bool value and go from there is there something similar I can do in this instance?
carMake is passed in
myQuery.JoinQueryOver(x => x.Car)
.Where(Restrictions.In("VIN",
Trades.Where(x => x.Car.Make.ToLower() == carMake.ToLower())
.Select(x => x.Car.PrimaryVIN)
.ToList()));
Assuming that Trades is a list of objects you can use .WhereRestrictionOn() instead. Try this (I split the code for better readability):
var vinsWithCarMake = Trades
.Where(x => x.Car.Make.ToLower() == carMake.ToLower())
.Select(x => x.Car.PrimaryVIN)
.ToList<string>();
var carAlias = null;
var result = myQuery.JoinAlias(x => x.Car, () => carAlias)
.WhereRestrictionOn(() => carAlias.VIN).IsInG<string>(vinsWithCarMake)
.List();
So I have the following where conditions
sessions = sessions.Where(y => y.session.SESSION_DIVISION.Any(x => x.DIVISION.ToUpper().Contains(SearchContent)) ||
y.session.ROOM.ToUpper().Contains(SearchContent) ||
y.session.COURSE.ToUpper().Contains(SearchContent));
I want to split this into multiple lines based on whether a string is empty for example:
if (!String.IsNullOrEmpty(Division)) {
sessions = sessions.Where(y => y.session.SESSION_DIVISION.Any(x => x.DIVISION.ToUpper().Contains(SearchContent)));
}
if (!String.IsNullOrEmpty(Room)) {
// this shoudl be OR
sessions = sessions.Where(y => y.session.ROOM.ToUpper().Contains(SearchContent));
}
if (!String.IsNullOrEmpty(course)) {
// this shoudl be OR
sessions = sessions.Where(y => y.session.COURSE.ToUpper().Contains(SearchContent));
}
If you notice I want to add multiple OR conditions split based on whether the Room, course, and Division strings are empty or not.
There are a few ways to go about this:
Apply the "where" to the original query each time, and then Union() the resulting queries.
var queries = new List<IQueryable<Session>>();
if (!String.IsNullOrEmpty(Division)) {
queries.Add(sessions.Where(y => y.session.SESSION_DIVISION.Any(x => x.DIVISION.ToUpper().Contains(SearchContent))));
}
if (!String.IsNullOrEmpty(Room)) {
// this shoudl be OR
queries.Add(sessions.Where(y => y.session.ROOM.ToUpper().Contains(SearchContent)));
}
if (!String.IsNullOrEmpty(course)) {
// this shoudl be OR
queries.Add(sessions.Where(y => y.session.COURSE.ToUpper().Contains(SearchContent)));
}
sessions = queries.Aggregate(sessions.Where(y => false), (q1, q2) => q1.Union(q2));
Do Expression manipulation to merge the bodies of your lambda expressions together, joined by OrElse expressions. (Complicated unless you've already got libraries to help you: after joining the bodies, you also have to traverse the expression tree to replace the parameter expressions. It can get sticky. See this post for details.
Use a tool like PredicateBuilder to do #2 for you.
.Where() assumes logical AND and as far as I know, there's no out of box solution to do it. If you want to separate OR statements, you may want to look into using Predicate Builder or Dynamic Linq.
You can create an extension method to conditionally apply the filter:
public static IQueryable<T> WhereIf<T>(
this IQueryable<T> source, bool condition,
Expression<Func<T, bool>> predicate)
{
return condition ? source.Where(predicate) : source;
}
And use it like this:
using static System.String;
...
var res = sessions
.WhereIf(!IsNullOrEmpty(Division), y => y.session.SESSION_DIVISION.ToUpper().Contains(SearchContent))
.WhereIf(!IsNullOrEmpty(Room), y => y.session.ROOM.ToUpper().Contains(SearchContent))
.WhereIf(!IsNullOrEmpty(course), y => y.session.COURSE.ToUpper().Contains(SearchContent)));
I want to do a sort of filtering chain to filter Receipt objects using queryOver functionality.
The chain can differ in length, according to the parameters user chooses on the screen.
Eventually, I want the chain to run somehow like this:
public IList<Receipt> RunFilters()
{
IQueryOver<Receipt, Receipt> currQuery = NHibernateHelper.Session.QueryOver<Receipt>();
foreach (var item in filters)
{
currQuery = item.RunFilter(currQuery);
}
return currQuery.List();
}
So, the question is - how RunFilter should be defined? I thought it should be
public IQueryOver<Receipt, Receipt> RunFilter(IQueryOver<Receipt, Receipt> prevFilter)
and they I can do filters like
return prevFilter.Where(receipt => receipt.TotalSum > 0);
But I can't do
return prevFilter.JoinQueryOver(v => v.Store).Where(vv => vv.Name.Equals(m_storeName));
Any ideas?
Thanks in advance
Victor
return prevFilter.JoinQueryOver(v => v.Store).Where(vv => vv.Name.Equals(m_storeName));
the above can be written as
Store storeAlias = null;
return prevFilter.JoinAlias(v => v.Store, () => storeAlias).Where(() => storeAlias.Name == m_storeName);
EDIT: fixed equation
I am trying to select a distinct list of values from a table whilst ordering on another column.
The only thing working for me so far uses magic strings and an object array. Any better (type-safe) way?
var projectionList = Projections.ProjectionList();
projectionList.Add(Projections.Property("FolderName"));
projectionList.Add(Projections.Property("FolderOrder"));
var list = Session.QueryOver<T>()
.Where(d => d.Company.Id == SharePointContextHelper.Current.CurrentCompanyId)
.OrderBy(t => t.FolderOrder).Asc
.Select(Projections.Distinct(projectionList))
.List<object[]>()
.ToList();
return list.Select(l => new Folder((string)l[0])).ToList();
btw, doing it with linq won't work, you must select FolderOrder otherwise you'll get a sql error (ORDER BY items must appear in the select list if SELECT DISTINCT is specified.
)
and then doing that gives a known error : Expression type 'NhDistinctExpression' is not supported by this SelectClauseVisitor. regarding using anonymous types with distinct
var q = Session.Query<T>()
.Where(d => d.Company.Id == SharePointContextHelper.Current.CurrentCompanyId)
.OrderBy(d => d.FolderOrder)
.Select(d => new {d.FolderName, d.FolderOrder})
.Distinct();
return q.ToList().Select(f => new Folder(f));
All seems a lot of hoops and complexity to do some sql basics....
To resolve the type-safety issue, the syntax is:
var projectionList = Projections.ProjectionList();
projectionList.Add(Projections.Property<T>(d => d.FolderName));
projectionList.Add(Projections.Property<T>(d => d.FolderOrder));
the object [] thing is unavoidable, unless you define a special class / struct to hold just FolderName and FolderOrder.
see this great introduction to QueryOver for type-saftey, which is most certainly supported.
best of luck.
I have a class with two computed columns. The formulas are select statements that grab counts from other tables, like so:
private const string VOTES_FORMULA = "(select count(v.id) from Votes v where v.BusinessID = Id)";
private const string SURVEY_FORMULA = "(select cast((case when exists (select * from surveys s where s.businessid = Id) then 1 else 0 end) as bit))";
// in my bootstrap code...
mappings.Override<Business>(map =>
{
map.IgnoreProperty(x => x.IsNewRecord);
map.IgnoreProperty(x => x.IdString);
map.Map(x => x.UserPassword).CustomType<EncryptedStringType>();
map.Map(x => x.HasTakenSurvey).Formula(SURVEY_FORMULA).Not.Insert().Not.Update();
map.Map(x => x.Votes).Formula(VOTES_FORMULA).Not.Insert().Not.Update();
});
This was all working fine with Fluent NHibernate 1.1 (using NHibernate 2.1), but I just upgraded to 1.2 (using NH 3.1) and it appears that Fluent NHibernate is ignoring the formulas. I'm getting an "invalid column name" exception for the two fields HasTakenSurvey and Votes because its' trying to query the columns directly rather than executing the formulas as directed. An example query:
exec sp_executesql N'select TOP (#p0) business0_.Id as Id0_, business0_.UserPassword as UserPass2_0_, business0_.HasTakenSurvey as HasTaken3_0_, business0_.Votes as Votes0_, business0_.Origin as Origin0_, business0_.SecurityToken as Security6_0_, business0_.BusinessName as Business7_0_, business0_.BusinessType as Business8_0_, business0_.BusinessImageUrl as Business9_0_, business0_.BusinessDescription as Busines10_0_, business0_.EmployeeCount as Employe11_0_, business0_.OwnerFirstName as OwnerFi12_0_, business0_.OwnerLastName as OwnerLa13_0_, business0_.UserPosition as UserPos14_0_, business0_.BusinessAddress1 as Busines15_0_, business0_.BusinessAddress2 as Busines16_0_, business0_.BusinessCity as Busines17_0_, business0_.BusinessState as Busines18_0_, business0_.BusinessPostal as Busines19_0_, business0_.BusinessCountry as Busines20_0_, business0_.UserBusinessPhone as UserBus21_0_, business0_.UserMobilePhone as UserMob22_0_, business0_.UserEmailAddress as UserEma23_0_, business0_.UserIpAddress as UserIpA24_0_, business0_.OptInReminders as OptInRe25_0_, business0_.OptInOffers as OptInOf26_0_, business0_.OptInSms as OptInSms0_, business0_.Created as Created0_, business0_.Modified as Modified0_ from dbo.Businesses business0_ order by business0_.BusinessName asc',N'#p0 int',#p0=25
Did the implementation change? What am I doing wrong?
As noted in the comments ConventionBuilder.Property.Always(x => x.Column(x.Property.Name)) was adding the column to all properties (and overriding the formula).
Adding .Columns.Clear() to the mapping should remove the column, so:
mappings.Override<Business>(map =>
{
map.IgnoreProperty(x => x.IsNewRecord);
map.IgnoreProperty(x => x.IdString);
map.Map(x => x.UserPassword).CustomType<EncryptedStringType>();
map.Map(x => x.HasTakenSurvey).Formula(SURVEY_FORMULA).Not.Insert().Not.Update().Columns.Clear();
map.Map(x => x.Votes).Formula(VOTES_FORMULA).Not.Insert().Not.Update().Columns.Clear();
});
Columns.Clear() solution provided by #david duffet also didn't work for me. The getter of Formula is private and so you can't filter on the Formula property (you get method-group cannot be converted to value or something like that). NH3.3, FNH 1.3.
My solution - create a custom Attribute in my Model project - IsNHibernateFormulaPropertyAttribute, apply it to my formula properties, and then check for that attribute in my naming convention logic using reflection:
private bool IsFormula(IPropertyInstance instance)
{
var propInfo = instance.Property.DeclaringType.GetProperty(instance.Property.Name);
if (propInfo != null)
{
return Attribute.IsDefined(propInfo, typeof(IsNHibernateFormulaPropertyAttribute));
}
return false;
}
public void Apply(IPropertyInstance instance)
{
if (!IsFormula(instance))
{
instance.Column(Convert(instance.Property.Name));
}
}