bobo Issue Tracker Rss Feed

Commented Issue: NuGet package? [3]

Shcherbachev — Sat, 15 Dec 2012 13:43:29 GMT

Since all the other Lucene.Net libs have NuGet packages, I was hoping this one did too. It doesn't seem so. Am I missing something?
Comments: You are right. We do need a NuGet package for Bobo.

Created Issue: NuGet package? [3]

tommck — Wed, 12 Dec 2012 17:16:27 GMT

Since all the other Lucene.Net libs have NuGet packages, I was hoping this one did too. It doesn't seem so. Am I missing something?

Commented Issue: PathFacetHandler fails on partial match [2]

Shcherbachev — Mon, 24 Sep 2012 12:14:23 GMT

For an index only containing deep paths (e.g. A/B/C, A/B/D, but *NOT* A) a parent value (A) will not match.

PathFacetHandler.GetFilters(List<int>, string, int, bool) identifies the range of terms matching the selected path.

It does this by finding the first path prefix match in the ordered termList and selecting subsequent terms until the prefix no longer matches.

Current code calls IndexOf(T) on a list from dataCache.valArray.GetInnerList() - a List<T>, returning the index or -1.
But the code expects to be calling the IndexOf(object) from a ITermValueList, which does a BinarySearch, returning the index or a negative number that is the bitwise complement of the index of the next larger element.

Although TermValueList<T> implements ITermValueList and derives from List<T>, List<T>'s IndexOf() isn't virtual (and has a different signature - T vs object), so different IndexOf() methods are called when it's cast as List<T> or ITermValueList.

Calling List<T>.IndexOf() instead of ITermValueList.IndexOf() means that indexes containing an exact match for the parent, will succeed - and indexes not containing an exact match for the parent, will fail.

Possible fix (closer to original Java sources) follows.

diff --git a/BoboBrowse.Net/Facets/impl/PathFacetHandler.cs b/BoboBrowse.Net/Facets/impl/PathFacetHandler.cs
--- a/BoboBrowse.Net/Facets/impl/PathFacetHandler.cs
+++ b/BoboBrowse.Net/Facets/impl/PathFacetHandler.cs
@@ -123,7 +123,7 @@

private void GetFilters(List<int> intSet, string val, int depth, bool strict)
{
- List<string> termList = dataCache.valArray.GetInnerList();
+ ITermValueList termList = dataCache.valArray;
int index = termList.IndexOf(val);

int startDepth = GetPathDepth(val);
@@ -141,7 +141,7 @@

for (int i = index; i < termList.Count; ++i)
{
- string path = termList[i];
+ string path = termList.Get(i);
if (path.StartsWith(val))
{
if (!strict || GetPathDepth(path) - startDepth == depth)

Comments: Thank you. Sources updated.

Edited Issue: PathFacetHandler fails on partial match [2]

Shcherbachev — Mon, 24 Sep 2012 12:14:23 GMT

Created Issue: PathFacetHandler fails on partial match [2]

ivanjh — Sat, 15 Sep 2012 02:14:56 GMT

Edited Issue: BrowseSelection fails with hyphenated values [1]

Shcherbachev — Wed, 12 Sep 2012 15:07:13 GMT

I believe Lucene terms are ordered using Lucene.Net.Index.Term.CompareTo() which relies upon string.CompareOrdinal().
BoboBrowse.Facets.data.TermStringList.IndexOf() uses System.Collections.Generic.List<T>.BinarySearch() with the default StringComparison.CurrentCulture.

These sort orders are different (hyphens as an example):
StringComparer.CurrentCulture.Compare("AB", "A-B") = -1
StringComparer.Ordinal.Compare("AB", "A-B") = 21

Because of this, a binary search will fail to find existing items.

Possible patch follows.

diff --git a/BoboBrowse.Net/Facets/data/TermStringList.cs b/BoboBrowse.Net/Facets/data/TermStringList.cs
--- a/BoboBrowse.Net/Facets/data/TermStringList.cs
+++ b/BoboBrowse.Net/Facets/data/TermStringList.cs
@@ -30,7 +30,7 @@

public override int IndexOf(object o)
{
- return BinarySearch((string) o);
+ return BinarySearch((string)o, System.StringComparer.Ordinal);
}
}

Commented Issue: BrowseSelection fails with hyphenated values [1]

Shcherbachev — Wed, 12 Sep 2012 15:06:21 GMT

Created Issue: BrowseSelection fails with hyphenated values [1]

ivanjh — Wed, 12 Sep 2012 12:21:22 GMT