Develop Reference

How do you get mixed results when searching multiple types with ElasticSearch 2.x using NEST?

I'm pretty new to Elastic Search and stumbled upon this issue.
When searching multiple document types from the same index, the types are getting appended in the result documents set instead of being default sorted by boosted field.
My document types shares the same fields.
This is my search query:
var response = await client.SearchAsync<ProductListResponse>(s => s
.Type("product,productbundle")
.Index(index)
.From(from)
.Size(size)
.Query(fsq => fsq
.FunctionScore(c => c.Query(q => q
.MultiMatch(m => m.Query(request.Query)
.Fields(f => f
.Field(n => n.Name, 100.0)
.Field(n => n.NameWithoutSpecialChars, 100.0)
.Field(n => n.ProductName)
.Field(n => n.TeaserText)
.Field(n => n.Description)
.Field(n => n.Features)
.Field(n => n.Modules)
)
.Type(TextQueryType.PhrasePrefix)
)
).Functions(f => f
.FieldValueFactor(b => b
.Field(p => p.IsBoosted)
.Modifier(FieldValueFactorModifier.Log1P))
)
)
)
.Sort(ss => ss
.Descending(SortSpecialField.Score))
.PostFilter(filter => filter.Bool(b => b.Must(must => allFilters)))
.Source(sr => sr
.Include(fi => fi
.Field(f => f.Name)
.Field(n => n.ProductName)
.Field(n => n.TeaserText)
.Field(f => f.Image)
.Field(f => f.Thumbnail)
.Field(f => f.Url)
.Field(f => f.Features)
)
)
);
Any help is appreciated.
I would preferre not to adapt the product type with the additons to productbundle type..

I can confirm that .Type() does not mess with the order.
My issue was a boosted property not getting a value while indexing the bundles.

OrderBy deletes object property

I have the following code:
var test = _fitDbContext.MvLatestTestResult
.Include(r => r.LastTest)
.Include(r => r.LastTest.Testrun)
.Include(r => r.LastTest.Testrunconfig)
.Where(r => r.LastTest.Testrunconfig.Started.Value.Hour >= 0 && r.LastTest.Testrunconfig.Started.Value.Hour < 6)
.Where(r => r.LastTest.Testrun.Userc == "build")
.Where(r => r.ProductId == productId)
.GroupBy(r => r.LastTest.Testrunconfig.Started.Value.Date)
.OrderByDescending(r => r.Key.Date)
.Take(3)
.ToDictionary(group => group.Key, group => group.GroupBy(r => r.Configfilename));
Variable test looks like this:
You can see that LastTest is null. If I remove .OrderByDescending(r => r.Key.Date), it contains value.
Why does OrderByDescending remove LastTest value?
EDIT
Interesting thing I just found out. If I change order of GroupBy and OrderBy, everything works as expected. The problem is that OrderBy is executed with all records and takes a long time.
var test = _fitDbContext.MvLatestTestResult
.Include(r => r.LastTest)
.Include(r => r.LastTest.Testrun)
.Include(r => r.LastTest.Testrunconfig)
.Where(r => r.LastTest.Testrunconfig.Started.Value.Hour >= 0 && r.LastTest.Testrunconfig.Started.Value.Hour < 6)
.Where(r => r.LastTest.Testrun.Userc == "build")
.Where(r => r.ProductId == productId)
.OrderByDescending(r => r.Key.Date)
.GroupBy(r => r.LastTest.Testrunconfig.Started.Value.Date)
.Take(3)
.ToDictionary(group => group.Key, group => group.GroupBy(r => r.Configfilename));

Exact search with ElasticSearch using EdgeNGram

I am using elastic for searching with greek and latin characters.My main problem is that I can't do exact searches.I am using edgeNgram filter on indexing, but I would like to control its min and max at search time according to my word length.For example if I type "titanox" I will firstly get "ΤΙΤΑΝΙΟΥ" and secondly "TITANOX".Here is my index creation :
var response = client.CreateIndex(index, s => s
.Settings(s1 => s1
.NumberOfShards(5)
.NumberOfReplicas(5)
.Analysis(a => a.TokenFilters(t => t.IcuTransform("greeklatin", it => it.Id("Greek-Latin; NFD; [:Nonspacing Mark:] Remove; NFC")//
.Direction(IcuTransformDirection.Forward)) //
.IcuTransform("latingreek", lg => lg.Id("Greek-Latin; NFD; [:Nonspacing Mark:] Remove; NFC")
.Direction(IcuTransformDirection.Reverse))
.EdgeNGram("greekedge", ed => ed.MaxGram(7)
.MinGram(1)
.Side(EdgeNGramSide.Front))
.Stop("greekstop", sw => sw.StopWords())
.Lowercase("greeklowercase", gl => gl.Language(Language.Greek.ToString()))
.KeywordMarker("greekkeywords", gk => gk.Keywords(""))
.Stemmer("greekstemmer", gs => gs.Language(Language.Greek.ToString())))
.Analyzers(a1 => a1
.Custom("greek", t => t.Tokenizer("standard")
.Filters("greekedge", "greekstop", "greeklowercase", "greekkeywords", "greekstemmer", "greeklatin")))))
.Mappings(m => m.Map(type, mt => mt.Properties(c => c.Text(c1 => c1.Name("id").Analyzer("greek"))
.Text(c2 => c2.Name("brand").Analyzer("greek"))
.Text(c3 => c3.Name("service").Analyzer("greek"))
.Text(c4 => c4.Name("servicegroupdesc").Analyzer("greek"))
.Text(c5 => c5.Name("servicecategorydesc).Analyzer("greek"))
.Text(c6 => c6.Name("partscategory").Analyzer("greek"))
.Text(c7 => c7.Name("partsid").Analyzer("greek"))
.Text(c8 => c8.Name("partsdesc").Analyzer("greek"))))));
and here my search
var response = client.Search<Cars>(n => n
.Index(index)
.Type(type)
.Query(m => m.MultiMatch(q => q
.Analyzer(analyzername)
//.MinimumShouldMatch("100%")
.Query("*" + searchWord + "*")
.Fields(f=>f.Field(fieldsForSearchList[0]))
.Fuzziness(Fuzziness.EditDistance(0))))
.Size(searchSize)
.From(0)
.TrackScores(true)
);

A solution that may can help is adding to this query a new query for boosting the word that users type.This can achieve more exact searches.

automapper optimization mapping (model to view model -> get latest version)

I am trying to optimize this part of code:
Mapper.CreateMap<Document, DocumentViewModel>()
.ForMember(g => g.Id, map => map.MapFrom(d => d.documentVersion.Where(v => v.version == d.documentVersion.Select(s => s.version).Max()).OrderByDescending(s => s.subVersion).First().Id))
.ForMember(g => g.IdRootDocument, map => map.MapFrom(d => d.Id))
.ForMember(g => g.certyficateType, map => map.MapFrom(d => d.documentVersion.Where(v => v.version == d.documentVersion.Select(s => s.version).Max()).OrderByDescending(s => s.subVersion).First().certyficateType))
I'm Using automapper, and I'm trying to optimize this part of code
In this part I'm trying mapping object from document to documentViewModel, in this complex model, source data always will be latest document version:
d => d.documentVersion.Where(v => v.version == d.documentVersion.Select(s => s.version).Max()).OrderByDescending(s => s.subVersion).First().myProportyX
Could anyone offer an example or suggestion as to how to approach optimization in this situation?

d => d.documentVersion.Where(v => v.version == d.documentVersion.Select(s => s.version).Max()).OrderByDescending(s => s.subVersion).First().myProporty
You are iterating quite a few times here, you may consider doing something like:
d.documentVersion.OrderByDescending(v => v.version).ThenByDescending(v => v.subVersion).First().myProperty
to reduce the number of iterations and only get the top version/subversion.

optimization mapping:
Mapper
.CreateMap<Document, DocumentViewModel>()
.ConvertUsing(doc =>
{
DocumentViewModel result = new DocumentViewModel();
DocumentVersion lastVersion = doc.documentVersion.Where(v => v.version == doc.documentVersion.Select(s => s.version).Max()).OrderByDescending(s => s.subVersion).First();
Mapper.Map(lastVersion, result);
return result;
});

How to partially project a child object with many fields in nHibernate

I have the following nHibernate query that select a course based on its course id and then return selected fields for the course object on the initial fetch, and the query executes with no issues.
MatchMode option = ...
CourseItem courseAlias = null;
TeacherItem teacherAlias = null;
var query = session.QueryOver<CourseItem>()
.JoinAlias(c => c.Teacher, () => teacherAlias)
.Where(c => c.CourseID.IsInsensitiveLike(strNumber, option))
.SelectList(list => list
.Select(c => c.CourseID).WithAlias(() => courseAlias.CourseID)
.Select(c => c.IsActive).WithAlias(() => courseAlias.IsActive)
.Select(c => c.CourseDesc).WithAlias(() => courseAlias.CourseDesc)
.Select(c => c.Teacher).WithAlias(() => courseAlias.Teacher))
.TransformUsing(Transformers.AliasToBean<CourseItem>())
.List<CourseItem>();
I wanted to go a step further with the query to only return a partial teacher object, let's say i just wanted to return the ID and Name. So, I updated the projected list to as follows:
var query = session.QueryOver<CourseItem>()
.JoinAlias(c => c.Teacher, () => teacherAlias)
.Where(c => c.CourseID.IsInsensitiveLike(strNumber, option))
.SelectList(list => list
.Select(c => c.CourseID).WithAlias(() => courseAlias.CourseID)
.Select(c => c.IsActive).WithAlias(() => courseAlias.IsActive)
.Select(c => c.CourseDesc).WithAlias(() => courseAlias.CourseDesc)
.Select(c => c.Teacher.ID).WithAlias(() => courseAlias.Teacher.ID)
.Select(c => c.Teacher.Name).WithAlias(() => courseAlias.Teacher.Name))
.TransformUsing(Transformers.AliasToBean<CourseItem>())
.List<CourseItem>();
The query doesn't work because nHibernate has no idea how to resovled based on Teacher.ID and Teacher.Name. Any thoughts on whether it's possible to NOT fetch the entire child object back to a parent object?
I've tried the following query and it works this is not my fully desired outcome
var query = session.QueryOver<CourseItem>(() => courseAlias)
.JoinAlias(() => courseAlias.Teacher, () => teacherAlias)
.Where(() => courseAlias.CourseID.IsInsensitiveLike(strNumber, option))
.SelectList(list => list
.Select(() => courseAlias.CourseID)
.Select(() => courseAlias.IsActive)
.Select(() => courseAlias.CourseDesc)
.Select(() => teacher.ID)
.Select(() => teacher.Name))
.List<object[]>();
I can query the right values but unable to transform it back correctly to the Course / teacher data type.
Any thoughts?
thanks!

We can indeed use custom transformer. There is one, which I am using for a really very very deep projections (inlcuding dynamic objects - 5.1.13. component, dynamic-component)
DeepTransformer<TEntity>
Take it (if needed adjust it) and your final query could be like this
// just the last lines are different
var query = session.QueryOver<CourseItem>()
.JoinAlias(c => c.Teacher, () => teacherAlias)
.Where(c => c.CourseID.IsInsensitiveLike(strNumber, option))
.SelectList(list => list
.Select(c => c.CourseID).WithAlias(() => courseAlias.CourseID)
.Select(c => c.IsActive).WithAlias(() => courseAlias.IsActive)
.Select(c => c.CourseDesc).WithAlias(() => courseAlias.CourseDesc)
// the native WitAlias would not work, it uses expression
// to extract just the last property
//.Select(c => c.Teacher.ID).WithAlias(() => courseAlias.Teacher.ID)
//.Select(c => c.Teacher.Name).WithAlias(() => courseAlias.Teacher.Name))
// so we can use this way to pass the deep alias
.Select(Projections.Property(() => teacherAlias.ID).As("Teacher.ID"))
.Select(Projections.Property(() => teacherAlias.Name).As("Teacher.Name"))
// instead of this
// .TransformUsing(Transformers.AliasToBean<CourseItem>())
// use this
.TransformUsing(new DeepTransformer<CourseItem>())
And in case, that your aliases do match to property names, that transformer will built the object tree...