4 // Copyright (C) 2004-2005 Novell, Inc.
8 // Permission is hereby granted, free of charge, to any person obtaining a
9 // copy of this software and associated documentation files (the "Software"),
10 // to deal in the Software without restriction, including without limitation
11 // the rights to use, copy, modify, merge, publish, distribute, sublicense,
12 // and/or sell copies of the Software, and to permit persons to whom the
13 // Software is furnished to do so, subject to the following conditions:
15 // The above copyright notice and this permission notice shall be included in
16 // all copies or substantial portions of the Software.
18 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19 // IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20 // FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
21 // AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
22 // LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
23 // FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
24 // DEALINGS IN THE SOFTWARE.
28 using System
.Collections
;
29 using System
.Diagnostics
;
30 using System
.Globalization
;
33 using System
.Threading
;
35 using System
.Xml
.Serialization
;
37 using Lucene
.Net
.Analysis
;
38 using Lucene
.Net
.Analysis
.Standard
;
39 using Lucene
.Net
.Documents
;
40 using Lucene
.Net
.Index
;
41 using Lucene
.Net
.QueryParsers
;
42 using LNS
= Lucene
.Net
.Search
;
46 namespace Beagle
.Daemon
{
48 public class LuceneCommon
{
50 public delegate bool HitFilter (Hit hit
);
56 // 2: Changed format of timestamp strings
57 // 3: Schema changed to be more Dashboard-Match-like
58 // 4: Schema changed for files to include _Directory property
59 // 5: Changed analyzer to support stemming. Bumped version # to
60 // force everyone to re-index.
61 // 6: lots of schema changes as part of the general refactoring
62 // 7: incremented to force a re-index after our upgrade to lucene 1.4
63 // (in theory the file formats are compatible, we are seeing 'term
64 // out of order' exceptions in some cases)
65 // 8: another forced re-index, this time because of massive changes
66 // in the file system backend (it would be nice to have per-backend
67 // versioning so that we didn't have to purge all indexes just
68 // because one changed)
69 // 9: changed the way properties are stored, changed in conjunction
70 // with sane handling of multiple properties on hits.
71 // 10: changed to support typed and mutable properties
72 // 11: moved mime type and hit type into properties
73 // 12: added year-month and year-month-day resolutions for all
75 // 13: moved source into a property
76 // 14: allow wildcard queries to also match keywords
77 // 15: analyze PropertyKeyword field, and store all properties as
78 // lower case so that we're truly case insensitive.
79 // 16: add inverted timestamp to make querying substantially faster
80 private const int MAJOR_VERSION
= 16;
81 private int minor_version
= 0;
83 private string index_name
;
84 private string top_dir
;
86 private string fingerprint
;
87 private int last_item_count
= -1;
89 // This is the big index, containing document full-texts and
90 // data that is expensive to index.
91 private Lucene
.Net
.Store
.Directory primary_store
= null;
93 // This is the small index, containing document info that we
94 // expect to have change. Canonical example: file names.
95 private Lucene
.Net
.Store
.Directory secondary_store
= null;
97 //////////////////////////////////////////////////////////////////////////////
99 protected LuceneCommon (string index_name
, int minor_version
)
101 this.index_name
= index_name
;
102 this.minor_version
= minor_version
;
104 this.top_dir
= (Path
.IsPathRooted (index_name
)) ? index_name
: Path
.Combine (PathFinder
.IndexDir
, index_name
);
107 //////////////////////////////////////////////////////////////////////////////
109 protected string IndexName { get { return index_name; }
}
111 public Lucene
.Net
.Store
.Directory PrimaryStore { get { return primary_store; }
}
113 public Lucene
.Net
.Store
.Directory SecondaryStore { get { return secondary_store; }
}
115 public string Fingerprint { get { return fingerprint; }
}
117 public string TopDirectory { get { return top_dir; }
}
119 //////////////////////////////////////////////////////////////////////////////
121 protected TextCache text_cache
= null;
123 public TextCache TextCache
{
124 get { return text_cache; }
125 set { text_cache = value; }
128 //////////////////////////////////////////////////////////////////////////////
130 private string VersionFile
{
131 get { return Path.Combine (top_dir, "version"); }
134 private string FingerprintFile
{
135 get { return Path.Combine (top_dir, "fingerprint"); }
138 // Shouldn't really be public
139 public string PrimaryIndexDirectory
{
140 get { return Path.Combine (top_dir, "PrimaryIndex"); }
143 // Shouldn't really be public
144 public string SecondaryIndexDirectory
{
145 get { return Path.Combine (top_dir, "SecondaryIndex"); }
148 public string LockDirectory
{
149 get { return Path.Combine (top_dir, "Locks"); }
152 //////////////////////////////////////////////////////////////////////////////
154 // Deal with dangling locks
156 private bool IsDanglingLock (FileInfo info
)
158 Log
.Debug ("Checking for dangling locks...");
160 // It isn't even a lock file
161 if (! info
.Name
.EndsWith (".lock"))
168 reader
= new StreamReader (info
.FullName
);
169 pid
= reader
.ReadLine ();
173 // We couldn't read the lockfile, so it probably went away.
179 // Looks like the lock file was empty, which really
180 // shouldn't happen. It should contain the PID of
181 // the process which locked it. Lets be on the safe
182 // side and assume it's a dangling lock.
183 Log
.Warn ("Found an empty lock file, that shouldn't happen: {0}", info
.FullName
);
188 cmdline_file
= String
.Format ("/proc/{0}/cmdline", pid
);
192 reader
= new StreamReader (cmdline_file
);
193 cmdline
= reader
.ReadLine ();
196 // If we can't open that file, either:
197 // (1) The process doesn't exist
198 // (2) It does exist, but it doesn't belong to us.
199 // Thus it isn't an IndexHelper
200 // In either case, the lock is dangling --- if it
205 // The process exists, but isn't an IndexHelper.
206 // If the lock file is still there, it is dangling.
207 // FIXME: During one run of bludgeon I got a null reference
208 // exception here, so I added the cmdline == null check.
209 // Why exactly would that happen? Is this logic correct
210 // in that (odd and presumably rare) case?
211 if (cmdline
== null || cmdline
.IndexOf ("IndexHelper.exe") == -1)
214 // If we reach this point, we know:
215 // (1) The process still exists
217 // (3) It is an IndexHelper process
218 // Thus it almost certainly isn't a dangling lock.
219 // The process might be wedged, but that is
224 protected bool Exists ()
226 if (! (Directory
.Exists (top_dir
)
227 && File
.Exists (VersionFile
)
228 && File
.Exists (FingerprintFile
)
229 && Directory
.Exists (PrimaryIndexDirectory
)
230 && IndexReader
.IndexExists (PrimaryIndexDirectory
)
231 && Directory
.Exists (SecondaryIndexDirectory
)
232 && IndexReader
.IndexExists (SecondaryIndexDirectory
)
233 && Directory
.Exists (LockDirectory
)))
236 // Check the index's version number. If it is wrong,
237 // declare the index non-existent.
239 StreamReader version_reader
;
241 version_reader
= new StreamReader (VersionFile
);
242 version_str
= version_reader
.ReadLine ();
243 version_reader
.Close ();
245 int current_major_version
, current_minor_version
;
246 int i
= version_str
.IndexOf ('.');
249 current_major_version
= Convert
.ToInt32 (version_str
.Substring (0, i
));
250 current_minor_version
= Convert
.ToInt32 (version_str
.Substring (i
+1));
252 current_minor_version
= Convert
.ToInt32 (version_str
);
253 current_major_version
= 0;
256 if (current_major_version
!= MAJOR_VERSION
257 || (minor_version
>= 0 && current_minor_version
!= minor_version
)) {
258 Logger
.Log
.Debug ("Version mismatch in {0}", index_name
);
259 Logger
.Log
.Debug ("Index has version {0}.{1}, expected {2}.{3}",
260 current_major_version
, current_minor_version
,
261 MAJOR_VERSION
, minor_version
);
265 // Check the lock directory: If there is a dangling write lock,
266 // assume that the index is corrupted and declare it non-existent.
267 DirectoryInfo lock_dir_info
;
268 lock_dir_info
= new DirectoryInfo (LockDirectory
);
269 foreach (FileInfo info
in lock_dir_info
.GetFiles ()) {
270 if (IsDanglingLock (info
)) {
271 Logger
.Log
.Warn ("Found a dangling index lock on {0}", info
.FullName
);
279 private Lucene
.Net
.Store
.Directory
CreateIndex (string path
)
281 // Create a directory to put the index in.
282 Directory
.CreateDirectory (path
);
284 // Create a new store.
285 Lucene
.Net
.Store
.Directory store
;
286 store
= Lucene
.Net
.Store
.FSDirectory
.GetDirectory (path
, LockDirectory
, true);
288 // Create an empty index in that store.
290 writer
= new IndexWriter (store
, null, true);
296 // Create will kill your index dead. Use it with care.
297 // You don't need to call Open after calling Create.
298 protected void Create ()
300 if (minor_version
< 0)
303 // Purge any existing directories.
304 if (Directory
.Exists (top_dir
)) {
305 Logger
.Log
.Debug ("Purging {0}", top_dir
);
306 Directory
.Delete (top_dir
, true);
309 // Create any necessary directories.
310 Directory
.CreateDirectory (top_dir
);
311 Directory
.CreateDirectory (LockDirectory
);
313 // Create the indexes.
314 primary_store
= CreateIndex (PrimaryIndexDirectory
);
315 secondary_store
= CreateIndex (SecondaryIndexDirectory
);
317 // Generate and store the index fingerprint.
318 fingerprint
= GuidFu
.ToShortString (Guid
.NewGuid ());
320 writer
= new StreamWriter (FingerprintFile
, false);
321 writer
.WriteLine (fingerprint
);
324 // Store our index version information.
325 writer
= new StreamWriter (VersionFile
, false);
326 writer
.WriteLine ("{0}.{1}", MAJOR_VERSION
, minor_version
);
330 protected void Open ()
335 protected void Open (bool read_only_mode
)
337 // Read our index fingerprint.
339 reader
= new StreamReader (FingerprintFile
);
340 fingerprint
= reader
.ReadLine ();
343 // Create stores for our indexes.
344 primary_store
= Lucene
.Net
.Store
.FSDirectory
.GetDirectory (PrimaryIndexDirectory
, LockDirectory
, false, read_only_mode
);
345 secondary_store
= Lucene
.Net
.Store
.FSDirectory
.GetDirectory (SecondaryIndexDirectory
, LockDirectory
, false, read_only_mode
);
348 ////////////////////////////////////////////////////////////////
354 private class SingletonTokenStream
: TokenStream
{
356 private string singleton_str
;
358 public SingletonTokenStream (string singleton_str
)
360 this.singleton_str
= singleton_str
;
363 override public Lucene
.Net
.Analysis
.Token
Next ()
365 if (singleton_str
== null)
368 Lucene
.Net
.Analysis
.Token token
;
369 token
= new Lucene
.Net
.Analysis
.Token (singleton_str
, 0, singleton_str
.Length
);
371 singleton_str
= null;
377 // FIXME: This assumes everything being indexed is in English!
378 private class BeagleAnalyzer
: StandardAnalyzer
{
380 private char [] buffer
= new char [2];
381 private bool strip_extra_property_info
= false;
382 private bool tokenize_email_hostname
= false;
384 public BeagleAnalyzer (bool is_indexing_analyzer
)
386 if (is_indexing_analyzer
) {
387 this.strip_extra_property_info
= true;
388 this.tokenize_email_hostname
= true;
390 this.strip_extra_property_info
= false;
391 this.tokenize_email_hostname
= false;
395 public override TokenStream
TokenStream (string fieldName
, TextReader reader
)
397 bool is_text_prop
= false;
399 // Strip off the first two characters in a property.
400 // We store type information in those two characters, so we don't
401 // want to index them.
402 if (fieldName
.StartsWith ("prop:")) {
404 if (strip_extra_property_info
) {
405 // Skip everything up to and including the first :
409 } while (c
!= -1 && c
!= ':');
412 is_text_prop
= fieldName
.StartsWith ("prop:t");
414 // If this is non-text property, just return one token
415 // containing the entire string. We do this to avoid
416 // tokenizing keywords.
417 if (! is_text_prop
) {
418 // We don't want to lower case the token if it's
419 // not in the private namespace.
421 TokenStream singleton_stream
= new SingletonTokenStream (reader
.ReadToEnd ());
423 if (fieldName
.StartsWith ("prop:k:" + LuceneQueryingDriver
.PrivateNamespace
))
424 return singleton_stream
;
426 return new LowerCaseFilter (singleton_stream
);
428 } else if (fieldName
== "PropertyKeyword")
429 return new LowerCaseFilter (new SingletonTokenStream (reader
.ReadToEnd ()));
431 TokenStream outstream
;
432 outstream
= base.TokenStream (fieldName
, reader
);
434 if (fieldName
== "Text"
435 || fieldName
== "HotText"
436 || fieldName
== "PropertyText"
438 outstream
= new NoiseEmailHostFilter (outstream
, tokenize_email_hostname
);
439 outstream
= new PorterStemFilter (outstream
);
446 static private Analyzer indexing_analyzer
= new BeagleAnalyzer (true);
447 static private Analyzer query_analyzer
= new BeagleAnalyzer (false);
449 static protected Analyzer IndexingAnalyzer { get { return indexing_analyzer; }
}
450 static protected Analyzer QueryAnalyzer { get { return query_analyzer; }
}
452 ////////////////////////////////////////////////////////////////
455 // Dealing with properties
458 static private char TypeToCode (PropertyType type
)
461 case PropertyType
.Text
: return 't';
462 case PropertyType
.Keyword
: return 'k';
463 case PropertyType
.Date
: return 'd';
465 throw new Exception ("Bad property type: " + type
);
468 static private PropertyType
CodeToType (char c
)
471 case 't': return PropertyType
.Text
;
472 case 'k': return PropertyType
.Keyword
;
473 case 'd': return PropertyType
.Date
;
476 throw new Exception ("Bad property code: " + c
);
479 static private string TypeToWildcardField (PropertyType type
)
482 case PropertyType
.Text
: return "PropertyText";
483 case PropertyType
.Keyword
: return "PropertyKeyword";
484 case PropertyType
.Date
: return "PropertyDate";
490 // Exposing this is a little bit suspicious.
491 static protected string PropertyToFieldName (PropertyType type
, string key
)
493 return String
.Format ("prop:{0}:{1}", TypeToCode (type
), key
);
497 static private void AddDateFields (string field_name
, Property prop
, Document doc
)
499 DateTime dt
= StringFu
.StringToDateTime (prop
.Value
);
502 f
= new Field ("YM:" + field_name
,
503 StringFu
.DateTimeToYearMonthString (dt
),
504 false, // never store
505 true, // always index
506 false); // never tokenize
509 f
= new Field ("D:" + field_name
,
510 StringFu
.DateTimeToDayString (dt
),
511 false, // never store
512 true, // always index
513 false); // never tokenize
517 static protected void AddPropertyToDocument (Property prop
, Document doc
)
519 if (prop
== null || prop
.Value
== null)
522 // Don't actually put properties in the UnindexedNamespace
523 // in the document. A horrible (and yet lovely!) hack.
524 if (prop
.Key
.StartsWith (StringFu
.UnindexedNamespace
))
529 if (prop
.IsSearched
) {
530 string wildcard_field
= TypeToWildcardField (prop
.Type
);
531 if (wildcard_field
!= null) {
532 f
= new Field (wildcard_field
,
534 false, // never stored
535 true, // always indexed
536 true); // always tokenize (just lowercases for keywords; full analysis for text)
539 if (prop
.Type
== PropertyType
.Date
)
540 AddDateFields (wildcard_field
, prop
, doc
);
545 coded_value
= String
.Format ("{0}:{1}",
546 prop
.IsSearched
? 's' : '_',
549 string field_name
= PropertyToFieldName (prop
.Type
, prop
.Key
);
551 f
= new Field (field_name
,
554 true, // always index
555 true); // always tokenize (strips off type code for keywords and lowercases)
558 if (prop
.Type
== PropertyType
.Date
)
559 AddDateFields (field_name
, prop
, doc
);
562 static protected Property
GetPropertyFromDocument (Field f
, Document doc
, bool from_primary_index
)
564 // Note: we don't use the document that we pass in,
565 // but in theory we could. At some later point we
566 // might need to split a property's data across two or
567 // more fields in the document.
573 field_name
= f
.Name ();
574 if (field_name
.Length
< 7
575 || ! field_name
.StartsWith ("prop:"))
579 field_value
= f
.StringValue ();
582 prop
= new Property ();
583 prop
.Type
= CodeToType (field_name
[5]);
584 prop
.Key
= field_name
.Substring (7);
585 prop
.Value
= field_value
.Substring (2);
586 prop
.IsSearched
= (field_value
[0] == 's');
587 prop
.IsMutable
= ! from_primary_index
;
588 prop
.IsStored
= f
.IsStored ();
593 //////////////////////////////////////////////////////////////////////////////
596 // Dealing with documents
599 static protected void BuildDocuments (Indexable indexable
,
600 out Document primary_doc
,
601 out Document secondary_doc
)
603 primary_doc
= new Document ();
604 secondary_doc
= null;
608 f
= Field
.Keyword ("Uri", UriFu
.UriToEscapedString (indexable
.Uri
));
611 if (indexable
.ParentUri
!= null) {
612 f
= Field
.Keyword ("ParentUri", UriFu
.UriToEscapedString (indexable
.ParentUri
));
616 if (indexable
.ValidTimestamp
) {
617 // Note that we also want to search in the
618 // Timestamp field when we do a wildcard date
619 // query, so that's why we also add a wildcard
620 // field for each item here.
622 string wildcard_field
= TypeToWildcardField (PropertyType
.Date
);
624 string str
= StringFu
.DateTimeToString (indexable
.Timestamp
);
625 f
= Field
.Keyword ("Timestamp", str
);
627 f
= Field
.UnStored (wildcard_field
, str
);
630 // Create an inverted timestamp so that we can
631 // sort by timestamp at search-time.
632 long timeval
= Convert
.ToInt64 (str
);
633 f
= Field
.UnStored ("InvertedTimestamp", (Int64
.MaxValue
- timeval
).ToString ());
636 str
= StringFu
.DateTimeToYearMonthString (indexable
.Timestamp
);
637 f
= Field
.Keyword ("YM:Timestamp", str
);
639 f
= Field
.UnStored ("YM:" + wildcard_field
, str
);
642 str
= StringFu
.DateTimeToDayString (indexable
.Timestamp
);
643 f
= Field
.Keyword ("D:Timestamp", str
);
645 f
= Field
.UnStored ("D:" + wildcard_field
, str
);
649 if (indexable
.NoContent
) {
650 // If there is no content, make a note of that
651 // in a special property.
653 prop
= Property
.NewBool ("beagle:NoContent", true);
654 AddPropertyToDocument (prop
, primary_doc
);
658 // Since we might have content, add our text
663 reader
= indexable
.GetTextReader ();
664 if (reader
!= null) {
665 f
= Field
.Text ("Text", reader
);
669 reader
= indexable
.GetHotTextReader ();
670 if (reader
!= null) {
671 f
= Field
.Text ("HotText", reader
);
676 // Store the Type and MimeType in special properties
678 if (indexable
.HitType
!= null) {
680 prop
= Property
.NewUnsearched ("beagle:HitType", indexable
.HitType
);
681 AddPropertyToDocument (prop
, primary_doc
);
684 if (indexable
.MimeType
!= null) {
686 prop
= Property
.NewUnsearched ("beagle:MimeType", indexable
.MimeType
);
687 AddPropertyToDocument (prop
, primary_doc
);
690 if (indexable
.Source
!= null) {
692 prop
= Property
.NewUnsearched ("beagle:Source", indexable
.Source
);
693 AddPropertyToDocument (prop
, primary_doc
);
696 // Store the other properties
698 foreach (Property prop
in indexable
.Properties
) {
699 Document target_doc
= primary_doc
;
700 if (prop
.IsMutable
) {
701 if (secondary_doc
== null) {
702 secondary_doc
= new Document ();
703 f
= Field
.Keyword ("Uri", UriFu
.UriToEscapedString (indexable
.Uri
));
704 secondary_doc
.Add (f
);
706 target_doc
= secondary_doc
;
709 AddPropertyToDocument (prop
, target_doc
);
713 static protected Document
RewriteDocument (Document old_secondary_doc
,
714 Indexable prop_only_indexable
)
716 Hashtable seen_props
;
717 seen_props
= new Hashtable ();
720 new_doc
= new Document ();
723 uri_f
= Field
.Keyword ("Uri", UriFu
.UriToEscapedString (prop_only_indexable
.Uri
));
726 Logger
.Log
.Debug ("Rewriting {0}", prop_only_indexable
.DisplayUri
);
728 // Add the new properties to the new document. To
729 // delete a property, set the Value to null... then it
730 // will be added to seen_props (so the old value will
731 // be ignored below), but AddPropertyToDocument will
732 // return w/o doing anything.
733 foreach (Property prop
in prop_only_indexable
.Properties
) {
734 seen_props
[prop
.Key
] = prop
;
735 AddPropertyToDocument (prop
, new_doc
);
736 Logger
.Log
.Debug ("New prop '{0}' = '{1}'", prop
.Key
, prop
.Value
);
739 // Copy the other properties from the old document to the
740 // new one, skipping any properties that we got new values
741 // for out of the Indexable.
742 if (old_secondary_doc
!= null) {
743 foreach (Field f
in old_secondary_doc
.Fields ()) {
745 prop
= GetPropertyFromDocument (f
, old_secondary_doc
, false);
746 if (prop
!= null && ! seen_props
.Contains (prop
.Key
)) {
747 Logger
.Log
.Debug ("Old prop '{0}' = '{1}'", prop
.Key
, prop
.Value
);
748 AddPropertyToDocument (prop
, new_doc
);
756 static protected Uri
GetUriFromDocument (Document doc
)
759 uri
= doc
.Get ("Uri");
761 throw new Exception ("Got document from Lucene w/o a URI!");
762 return UriFu
.EscapedStringToUri (uri
);
765 static protected Hit
DocumentToHit (Document doc
)
770 hit
.Uri
= GetUriFromDocument (doc
);
773 str
= doc
.Get ("ParentUri");
775 hit
.ParentUri
= UriFu
.EscapedStringToUri (str
);
777 hit
.Timestamp
= StringFu
.StringToDateTime (doc
.Get ("Timestamp"));
779 AddPropertiesToHit (hit
, doc
, true);
781 // Get the Type and MimeType from the properties.
782 hit
.Type
= hit
.GetFirstProperty ("beagle:HitType");
783 hit
.MimeType
= hit
.GetFirstProperty ("beagle:MimeType");
784 hit
.Source
= hit
.GetFirstProperty ("beagle:Source");
789 static protected void AddPropertiesToHit (Hit hit
, Document doc
, bool from_primary_index
)
791 foreach (Field f
in doc
.Fields ()) {
793 prop
= GetPropertyFromDocument (f
, doc
, from_primary_index
);
795 hit
.AddProperty (prop
);
800 //////////////////////////////////////////////////////////////////////////////
803 // Handle the index's item count
806 public int GetItemCount ()
808 if (last_item_count
< 0) {
810 reader
= GetReader (PrimaryStore
);
811 last_item_count
= reader
.NumDocs ();
812 ReleaseReader (reader
);
814 return last_item_count
;
817 // We should set the cached count of index items when IndexReaders
818 // are open and available, so calls to GetItemCount will return immediately.
820 protected bool HaveItemCount { get { return last_item_count >= 0; }
}
822 protected void SetItemCount (IndexReader reader
)
824 last_item_count
= reader
.NumDocs ();
827 public void SetItemCount (int count
)
829 last_item_count
= count
;
832 protected void AdjustItemCount (int delta
)
834 if (last_item_count
>= 0)
835 last_item_count
+= delta
;
838 //////////////////////////////////////////////////////////////////////////////
841 // Access to the stemmer and list of stop words
844 static PorterStemmer stemmer
= new PorterStemmer ();
846 static public string Stem (string str
)
848 return stemmer
.Stem (str
);
851 public static bool IsStopWord (string stemmed_word
)
853 return ArrayFu
.IndexOfString (StopAnalyzer
.ENGLISH_STOP_WORDS
, stemmed_word
) != -1;
856 //////////////////////////////////////////////////////////////////////////////
859 // Special Hit Filtering classes
862 static private bool TrueHitFilter (Hit hit
)
867 static private HitFilter true_hit_filter
= new HitFilter (TrueHitFilter
);
869 public class OrHitFilter
{
871 private ArrayList all
= new ArrayList ();
872 private bool contains_known_true
= false;
874 public void Add (HitFilter hit_filter
)
876 if (hit_filter
== true_hit_filter
)
877 contains_known_true
= true;
878 all
.Add (hit_filter
);
881 public bool HitFilter (Hit hit
)
883 if (contains_known_true
)
885 foreach (HitFilter hit_filter
in all
)
886 if (hit_filter (hit
))
892 public class AndHitFilter
{
894 private ArrayList all
= new ArrayList ();
896 public void Add (HitFilter hit_filter
)
898 all
.Add (hit_filter
);
901 public bool HitFilter (Hit hit
)
903 foreach (HitFilter hit_filter
in all
)
904 if (! hit_filter (hit
))
910 public class NotHitFilter
{
913 public NotHitFilter (HitFilter original
)
915 this.original
= original
;
918 public bool HitFilter (Hit hit
)
920 return ! original (hit
);
924 //////////////////////////////////////////////////////////////////////////////
930 static private LNS
.Query
StringToQuery (string field_name
,
934 ArrayList tokens
= new ArrayList ();
936 // Use the analyzer to extract the query's tokens.
937 // This code is taken from Lucene's query parser.
938 TokenStream source
= QueryAnalyzer
.TokenStream (field_name
, new StringReader (text
));
940 Lucene
.Net
.Analysis
.Token token
;
942 token
= source
.Next ();
945 } catch (IOException
) {
949 tokens
.Add (token
.TermText ());
953 } catch (IOException
) {
957 if (tokens
.Count
== 0)
960 LNS
.PhraseQuery query
= new LNS
.PhraseQuery ();
962 foreach (string token
in tokens
) {
964 term
= new Term (field_name
, token
);
966 if (term_list
!= null)
967 term_list
.Add (term
);
974 // Date Range Handling
977 // This function will break down dates to discrete chunks of
978 // time to avoid expanding RangeQuerys as much as possible.
979 // For example, searching for
981 // YMD(5 May 2005, 16 Oct 2006)
983 // would break down into three queries:
985 // (YM(May 2005) AND D(5,31)) OR
986 // YM(Jun 2005, Sep 2006) OR
987 // (YM(Oct 2006) AND D(1,16))
989 static private DateTime lower_bound
= new DateTime (1970, 1, 1);
991 // FIXME: we should probably boost this sometime around 2030.
992 // Mark your calendar.
993 static private DateTime upper_bound
= new DateTime (2038, 12, 31);
995 static private Term
NewYearMonthTerm (string field_name
, int y
, int m
)
997 return new Term ("YM:" + field_name
, String
.Format ("{0}{1:00}", y
, m
));
1000 static private LNS
.Query
NewYearMonthQuery (string field_name
, int y
, int m
)
1002 return new LNS
.TermQuery (NewYearMonthTerm (field_name
, y
, m
));
1005 static private LNS
.Query
NewYearMonthQuery (string field_name
, int y1
, int m1
, int y2
, int m2
)
1007 return new LNS
.RangeQuery (NewYearMonthTerm (field_name
, y1
, m1
),
1008 NewYearMonthTerm (field_name
, y2
, m2
),
1009 true); // query is inclusive
1012 static private Term
NewDayTerm (string field_name
, int d
)
1014 return new Term ("D:" + field_name
, String
.Format ("{0:00}", d
));
1017 static private LNS
.Query
NewDayQuery (string field_name
, int d1
, int d2
)
1019 return new LNS
.RangeQuery (NewDayTerm (field_name
, d1
),
1020 NewDayTerm (field_name
, d2
),
1021 true); // query is inclusive
1024 private class DateRangeHitFilter
{
1026 public DateTime StartDate
;
1027 public DateTime EndDate
;
1029 public bool HitFilter (Hit hit
)
1031 // First, check the Timestamp
1032 if (Key
== QueryPart_DateRange
.AllPropertiesKey
1033 || Key
== QueryPart_DateRange
.TimestampKey
) {
1036 if (StartDate
<= dt
&& dt
<= EndDate
)
1038 if (Key
== QueryPart_DateRange
.TimestampKey
)
1042 if (Key
== QueryPart_DateRange
.AllPropertiesKey
) {
1043 // Walk through all of the properties, and see if any
1044 // date properties fall inside the range.
1045 foreach (Property prop
in hit
.Properties
) {
1046 if (prop
.Type
== PropertyType
.Date
) {
1048 dt
= StringFu
.StringToDateTime (prop
.Value
);
1049 if (StartDate
<= dt
&& dt
<= EndDate
)
1055 // Walk through all of the properties with the given key,
1056 // and see if any of them fall inside of the range.
1058 values
= hit
.GetProperties (Key
);
1059 foreach (string v
in values
) {
1061 dt
= StringFu
.StringToDateTime (v
);
1062 if (StartDate
<= dt
&& dt
<= EndDate
)
1070 static private LNS
.Query
GetDateRangeQuery (QueryPart_DateRange part
, out HitFilter hit_filter
)
1073 if (part
.Key
== QueryPart_DateRange
.AllPropertiesKey
)
1074 field_name
= TypeToWildcardField (PropertyType
.Date
);
1075 else if (part
.Key
== QueryPart_DateRange
.TimestampKey
)
1076 field_name
= "Timestamp";
1078 field_name
= PropertyToFieldName (PropertyType
.Date
, part
.Key
);
1080 // FIXME: We could optimize this and reduce the size of our range
1081 // queries if we actually new the min and max date that appear in
1082 // any properties in the index. We would need to inspect the index to
1083 // determine that at start-up, and then track it as new documents
1084 // get added to the index.
1085 if (part
.StartDate
< lower_bound
)
1086 part
.StartDate
= lower_bound
;
1087 if (part
.EndDate
> upper_bound
|| part
.EndDate
== DateTime
.MinValue
)
1088 part
.EndDate
= upper_bound
;
1090 // Swap the start and end dates if they come in reversed.
1091 if (part
.StartDate
> part
.EndDate
) {
1093 swap
= part
.StartDate
;
1094 part
.StartDate
= part
.EndDate
;
1095 part
.EndDate
= swap
;
1098 // Set up our hit filter to cull out the bad dates.
1099 DateRangeHitFilter drhf
;
1100 drhf
= new DateRangeHitFilter ();
1101 drhf
.Key
= part
.Key
;
1102 drhf
.StartDate
= part
.StartDate
;
1103 drhf
.EndDate
= part
.EndDate
;
1104 hit_filter
= new HitFilter (drhf
.HitFilter
);
1106 Logger
.Log
.Debug ("Building new date range query");
1107 Logger
.Log
.Debug ("Start: {0}", part
.StartDate
);
1108 Logger
.Log
.Debug ("End: {0}", part
.EndDate
);
1110 int y1
, m1
, d1
, y2
, m2
, d2
;
1111 y1
= part
.StartDate
.Year
;
1112 m1
= part
.StartDate
.Month
;
1113 d1
= part
.StartDate
.Day
;
1114 y2
= part
.EndDate
.Year
;
1115 m2
= part
.EndDate
.Month
;
1116 d2
= part
.EndDate
.Day
;
1118 LNS
.BooleanQuery top_level_query
;
1119 top_level_query
= new LNS
.BooleanQuery ();
1121 // A special case: both the start and the end of our range fall
1122 // in the same month.
1123 if (y1
== y2
&& m1
== m2
) {
1125 ym_query
= NewYearMonthQuery (field_name
, y1
, m1
);
1127 // If our range only covers a part of the month, do a range query on the days.
1128 if (d1
!= 1 || d2
!= DateTime
.DaysInMonth (y2
, m2
)) {
1129 LNS
.BooleanQuery sub_query
;
1130 sub_query
= new LNS
.BooleanQuery ();
1131 sub_query
.Add (ym_query
, true, false);
1132 sub_query
.Add (NewDayQuery (field_name
, d1
, d2
), true, false);
1133 top_level_query
.Add (sub_query
, false, false);
1135 top_level_query
.Add (ym_query
, false, false);
1140 // Handle a partial month at the beginning of our range.
1142 LNS
.BooleanQuery sub_query
;
1143 sub_query
= new LNS
.BooleanQuery ();
1144 sub_query
.Add (NewYearMonthQuery (field_name
, y1
, m1
), true, false);
1145 sub_query
.Add (NewDayQuery (field_name
, d1
, DateTime
.DaysInMonth (y1
, m1
)), true, false);
1146 top_level_query
.Add (sub_query
, false, false);
1155 // And likewise, handle a partial month at the end of our range.
1156 if (d2
< DateTime
.DaysInMonth (y2
, m2
)) {
1157 LNS
.BooleanQuery sub_query
;
1158 sub_query
= new LNS
.BooleanQuery ();
1159 sub_query
.Add (NewYearMonthQuery (field_name
, y2
, m2
), true, false);
1160 sub_query
.Add (NewDayQuery (field_name
, 1, d2
), true, false);
1161 top_level_query
.Add (sub_query
, false, false);
1170 // Generate the query for the "middle" of our period, if it is non-empty
1171 if (y1
< y2
|| ((y1
== y2
) && m1
<= m2
))
1172 top_level_query
.Add (NewYearMonthQuery (field_name
, y1
, m1
, y2
, m2
),
1176 return top_level_query
;
1179 // search_subset_uris is a list of Uris that this search should be
1181 static protected void QueryPartToQuery (QueryPart abstract_part
,
1182 bool only_build_primary_query
,
1183 ArrayList term_list
,
1184 out LNS
.Query primary_query
,
1185 out LNS
.Query secondary_query
,
1186 out HitFilter hit_filter
)
1188 primary_query
= null;
1189 secondary_query
= null;
1191 // By default, we assume that our lucene queries will return exactly the
1192 // matching set of objects. We need to set the hit filter if further
1193 // refinement of the search results is required. (As in the case of
1194 // date range queries, for example.) We essentially have to do this
1195 // to make OR queries work correctly.
1196 hit_filter
= true_hit_filter
;
1198 // The exception is when dealing with a prohibited part. Just return
1199 // null for the hit filter in that case. This works since
1200 // prohibited parts are not allowed inside of OR queries.
1201 if (abstract_part
.Logic
== QueryPartLogic
.Prohibited
)
1204 if (abstract_part
== null)
1207 if (abstract_part
is QueryPart_Text
) {
1208 QueryPart_Text part
= (QueryPart_Text
) abstract_part
;
1210 if (! (part
.SearchFullText
|| part
.SearchTextProperties
))
1213 LNS
.BooleanQuery p_query
= new LNS
.BooleanQuery ();
1214 LNS
.BooleanQuery s_query
= new LNS
.BooleanQuery ();
1216 if (part
.SearchFullText
) {
1218 subquery
= StringToQuery ("Text", part
.Text
, term_list
);
1219 if (subquery
!= null)
1220 p_query
.Add (subquery
, false, false);
1222 // FIXME: HotText is ignored for now!
1223 // subquery = StringToQuery ("HotText", part.Text);
1224 // if (subquery != null)
1225 // p_query.Add (subquery, false, false);
1228 if (part
.SearchTextProperties
) {
1230 subquery
= StringToQuery ("PropertyText", part
.Text
, term_list
);
1231 if (subquery
!= null) {
1232 p_query
.Add (subquery
, false, false);
1233 // Properties can live in either index
1234 if (! only_build_primary_query
)
1235 s_query
.Add (subquery
.Clone () as LNS
.Query
, false, false);
1239 term
= new Term ("PropertyKeyword", part
.Text
.ToLower ()); // make sure text is lowercased
1240 // FIXME: terms are already added in term_list. But they may have been tokenized
1241 // The term here is non-tokenized version. Should this be added to term_list ?
1242 // term_list is used to calculate scores
1243 if (term_list
!= null)
1244 term_list
.Add (term
);
1245 subquery
= new LNS
.TermQuery (term
);
1246 p_query
.Add (subquery
, false, false);
1247 // Properties can live in either index
1248 if (! only_build_primary_query
)
1249 s_query
.Add (subquery
.Clone () as LNS
.Query
, false, false);
1252 primary_query
= p_query
;
1253 if (! only_build_primary_query
)
1254 secondary_query
= s_query
;
1259 if (abstract_part
is QueryPart_Wildcard
) {
1260 QueryPart_Wildcard part
= (QueryPart_Wildcard
) abstract_part
;
1262 LNS
.BooleanQuery p_query
= new LNS
.BooleanQuery ();
1263 LNS
.BooleanQuery s_query
= new LNS
.BooleanQuery ();
1268 // Lower case the terms for searching
1269 string query_string_lower
= part
.QueryString
.ToLower ();
1271 // Search text content
1272 term
= new Term ("Text", query_string_lower
);
1273 subquery
= new LNS
.WildcardQuery (term
);
1274 p_query
.Add (subquery
, false, false);
1275 term_list
.Add (term
);
1277 // Search text properties
1278 term
= new Term ("PropertyText", query_string_lower
);
1279 subquery
= new LNS
.WildcardQuery (term
);
1280 p_query
.Add (subquery
, false, false);
1281 // Properties can live in either index
1282 if (! only_build_primary_query
)
1283 s_query
.Add (subquery
.Clone () as LNS
.Query
, false, false);
1284 term_list
.Add (term
);
1286 // Search property keywords
1287 term
= new Term ("PropertyKeyword", query_string_lower
);
1288 term_list
.Add (term
);
1289 subquery
= new LNS
.WildcardQuery (term
);
1290 p_query
.Add (subquery
, false, false);
1291 // Properties can live in either index
1292 if (! only_build_primary_query
)
1293 s_query
.Add (subquery
.Clone () as LNS
.Query
, false, false);
1295 primary_query
= p_query
;
1296 if (! only_build_primary_query
)
1297 secondary_query
= s_query
;
1302 if (abstract_part
is QueryPart_Property
) {
1303 QueryPart_Property part
= (QueryPart_Property
) abstract_part
;
1306 if (part
.Key
== QueryPart_Property
.AllProperties
) {
1307 field_name
= TypeToWildcardField (part
.Type
);
1308 // FIXME: probably shouldn't just return silently
1309 if (field_name
== null)
1312 field_name
= PropertyToFieldName (part
.Type
, part
.Key
);
1314 if (part
.Type
== PropertyType
.Text
)
1315 primary_query
= StringToQuery (field_name
, part
.Value
, term_list
);
1318 term
= new Term (field_name
, part
.Value
.ToLower ());
1319 if (term_list
!= null)
1320 term_list
.Add (term
);
1321 primary_query
= new LNS
.TermQuery (term
);
1324 // Properties can live in either index
1325 if (! only_build_primary_query
&& primary_query
!= null)
1326 secondary_query
= primary_query
.Clone () as LNS
.Query
;
1331 if (abstract_part
is QueryPart_DateRange
) {
1333 QueryPart_DateRange part
= (QueryPart_DateRange
) abstract_part
;
1335 primary_query
= GetDateRangeQuery (part
, out hit_filter
);
1336 // Date properties can live in either index
1337 if (! only_build_primary_query
&& primary_query
!= null)
1338 secondary_query
= primary_query
.Clone () as LNS
.Query
;
1340 // If this is a prohibited part, invert our hit filter.
1341 if (part
.Logic
== QueryPartLogic
.Prohibited
) {
1343 nhf
= new NotHitFilter (hit_filter
);
1344 hit_filter
= new HitFilter (nhf
.HitFilter
);
1350 if (abstract_part
is QueryPart_Or
) {
1351 QueryPart_Or part
= (QueryPart_Or
) abstract_part
;
1353 // Assemble a new BooleanQuery combining all of the sub-parts.
1354 LNS
.BooleanQuery p_query
;
1355 p_query
= new LNS
.BooleanQuery ();
1357 LNS
.BooleanQuery s_query
= null;
1358 if (! only_build_primary_query
)
1359 s_query
= new LNS
.BooleanQuery ();
1361 primary_query
= p_query
;
1362 secondary_query
= s_query
;
1364 OrHitFilter or_hit_filter
= null;
1366 foreach (QueryPart sub_part
in part
.SubParts
) {
1367 LNS
.Query p_subq
, s_subq
;
1368 HitFilter sub_hit_filter
; // FIXME: This is (and must be) ignored
1369 // FIXME: Any subpart in an OR which has a hit filter won't work
1370 // correctly, because we can't tell which part of an OR we matched
1371 // against to filter correctly. This affects date range queries.
1372 QueryPartToQuery (sub_part
, only_build_primary_query
,
1374 out p_subq
, out s_subq
, out sub_hit_filter
);
1376 p_query
.Add (p_subq
, false, false);
1378 s_query
.Add (s_subq
, false, false);
1379 if (sub_hit_filter
!= null) {
1380 if (or_hit_filter
== null)
1381 or_hit_filter
= new OrHitFilter ();
1382 or_hit_filter
.Add (sub_hit_filter
);
1386 if (or_hit_filter
!= null)
1387 hit_filter
= new HitFilter (or_hit_filter
.HitFilter
);
1392 throw new Exception ("Unhandled QueryPart type! " + abstract_part
.ToString ());
1395 static protected LNS
.Query
UriQuery (string field_name
, Uri uri
)
1397 return new LNS
.TermQuery (new Term (field_name
, UriFu
.UriToEscapedString (uri
)));
1400 static protected LNS
.Query
UriQuery (string field_name
, ICollection uri_list
)
1402 return UriQuery (field_name
, uri_list
, null);
1405 static protected LNS
.Query
UriQuery (string field_name
, ICollection uri_list
, LNS
.Query extra_requirement
)
1407 if (uri_list
.Count
== 0)
1411 max_clauses
= LNS
.BooleanQuery
.GetMaxClauseCount ();
1414 N
= 1 + (uri_list
.Count
- 1) / max_clauses
;
1416 LNS
.BooleanQuery top_query
;
1417 top_query
= new LNS
.BooleanQuery ();
1420 if (extra_requirement
!= null) {
1421 top_query
.Add (extra_requirement
, true, false);
1425 ArrayList bottom_queries
= null;
1428 bottom_queries
= new ArrayList ();
1429 for (int i
= 0; i
< N
; ++i
) {
1430 LNS
.BooleanQuery bq
;
1431 bq
= new LNS
.BooleanQuery ();
1432 bottom_queries
.Add (bq
);
1433 top_query
.Add (bq
, false, false);
1437 foreach (Uri uri
in uri_list
) {
1439 subquery
= UriQuery (field_name
, uri
);
1441 LNS
.BooleanQuery target
;
1445 target
= (LNS
.BooleanQuery
) bottom_queries
[cursor
];
1451 target
.Add (subquery
, false, false);
1457 ///////////////////////////////////////////////////////////////////////////////////
1459 public int SegmentCount
{
1461 DirectoryInfo dir_info
;
1462 int p_count
= 0, s_count
= 0;
1464 dir_info
= new DirectoryInfo (PrimaryIndexDirectory
);
1465 foreach (FileInfo file_info
in dir_info
.GetFiles ())
1466 if (file_info
.Extension
== ".cfs")
1469 dir_info
= new DirectoryInfo (SecondaryIndexDirectory
);
1470 foreach (FileInfo file_info
in dir_info
.GetFiles ())
1471 if (file_info
.Extension
== ".cfs")
1474 return p_count
> s_count
? p_count
: s_count
;
1478 ///////////////////////////////////////////////////////////////////////////////////
1480 // Cache IndexReaders on a per-Lucene index basis, since they
1481 // are extremely expensive to create. Note that using this
1482 // only makes sense in situations where the index only
1483 // possibly might change from underneath us, but most of the
1484 // time probably won't. This means it makes sense to do
1485 // this in LuceneQueryingDriver.cs, but it doesn't in
1486 // LuceneIndexingDriver.cs.
1488 private class ReaderAndVersion
{
1490 public IndexReader Reader
;
1491 public long Version
;
1492 public int Refcount
;
1494 public ReaderAndVersion (IndexReader reader
, long version
)
1496 this.Reader
= reader
;
1497 this.Version
= version
;
1502 static private Hashtable directory_rav_map
= new Hashtable ();
1503 static private Hashtable reader_rav_map
= new Hashtable ();
1505 static public LNS
.IndexSearcher
GetSearcher (Lucene
.Net
.Store
.Directory directory
)
1507 IndexReader reader
= GetReader (directory
);
1509 return new LNS
.IndexSearcher (reader
);
1512 static public IndexReader
GetReader (Lucene
.Net
.Store
.Directory directory
)
1517 lock (reader_rav_map
) {
1518 ReaderAndVersion rav
= (ReaderAndVersion
) directory_rav_map
[directory
];
1521 version
= IndexReader
.GetCurrentVersion (directory
);
1522 reader
= IndexReader
.Open (directory
);
1524 rav
= new ReaderAndVersion (reader
, version
);
1527 directory_rav_map
[directory
] = rav
;
1528 reader_rav_map
[reader
] = rav
;
1533 version
= IndexReader
.GetCurrentVersion (directory
);
1535 if (version
!= rav
.Version
) {
1536 UnrefReaderAndVersion_Unlocked (rav
);
1538 reader
= IndexReader
.Open (directory
);
1540 rav
= new ReaderAndVersion (reader
, version
);
1543 directory_rav_map
[directory
] = rav
;
1544 reader_rav_map
[reader
] = rav
;
1552 static private void UnrefReaderAndVersion_Unlocked (ReaderAndVersion rav
)
1556 if (rav
.Refcount
== 0) {
1557 rav
.Reader
.Close ();
1558 reader_rav_map
.Remove (rav
.Reader
);
1562 static public void ReleaseReader (IndexReader reader
)
1564 lock (reader_rav_map
) {
1565 ReaderAndVersion rav
= (ReaderAndVersion
) reader_rav_map
[reader
];
1567 UnrefReaderAndVersion_Unlocked (rav
);
1571 static public void ReleaseSearcher (LNS
.IndexSearcher searcher
)
1573 IndexReader reader
= searcher
.GetIndexReader ();
1576 ReleaseReader (reader
);
1579 ///////////////////////////////////////////////////////////////////////////////////
1582 // Various ways to grab lots of hits at once.
1583 // These should never be used for querying, only for utility
1587 public int GetBlockOfHits (int cookie
,
1588 Hit
[] block_of_hits
)
1590 IndexReader primary_reader
;
1591 IndexReader secondary_reader
;
1592 primary_reader
= GetReader (PrimaryStore
);
1593 secondary_reader
= GetReader (SecondaryStore
);
1596 request_size
= block_of_hits
.Length
;
1597 if (request_size
> primary_reader
.NumDocs ())
1598 request_size
= primary_reader
.NumDocs ();
1601 max_doc
= primary_reader
.MaxDoc ();
1605 random
= new Random ();
1606 cookie
= random
.Next (max_doc
);
1609 int original_cookie
;
1610 original_cookie
= cookie
;
1612 Hashtable primary_docs
, secondary_docs
;
1613 primary_docs
= UriFu
.NewHashtable ();
1614 secondary_docs
= UriFu
.NewHashtable ();
1616 // Load the primary documents
1617 for (int i
= 0; i
< request_size
; ++i
) {
1619 if (! primary_reader
.IsDeleted (cookie
)) {
1621 doc
= primary_reader
.Document (cookie
);
1622 primary_docs
[GetUriFromDocument (doc
)] = doc
;
1626 if (cookie
>= max_doc
) // wrap around
1629 // If we somehow end up back where we started,
1631 if (cookie
== original_cookie
)
1635 // If necessary, load the secondary documents
1636 if (secondary_reader
!= null) {
1637 LNS
.IndexSearcher searcher
;
1638 searcher
= new LNS
.IndexSearcher (secondary_reader
);
1640 LNS
.Query uri_query
;
1641 uri_query
= UriQuery ("Uri", primary_docs
.Keys
);
1644 hits
= searcher
.Search (uri_query
);
1645 for (int i
= 0; i
< hits
.Length (); ++i
) {
1648 secondary_docs
[GetUriFromDocument (doc
)] = doc
;
1654 ReleaseReader (primary_reader
);
1655 ReleaseReader (secondary_reader
);
1657 // Now assemble the hits
1659 foreach (Uri uri
in primary_docs
.Keys
) {
1660 Document primary_doc
, secondary_doc
;
1661 primary_doc
= primary_docs
[uri
] as Document
;
1662 secondary_doc
= secondary_docs
[uri
] as Document
;
1665 hit
= DocumentToHit (primary_doc
);
1666 if (secondary_doc
!= null)
1667 AddPropertiesToHit (hit
, secondary_doc
, false);
1669 block_of_hits
[j
] = hit
;
1673 // null-pad the array, if necessary
1674 for (; j
< block_of_hits
.Length
; ++j
)
1675 block_of_hits
[j
] = null;
1678 // Return the new cookie
1682 // For a large index, this will be very slow and will consume
1683 // a lot of memory. Don't call it without a good reason!
1684 // We return a hashtable indexed by Uri.
1685 public Hashtable
GetAllHitsByUri ()
1688 all_hits
= UriFu
.NewHashtable ();
1690 IndexReader primary_reader
;
1691 IndexReader secondary_reader
;
1692 primary_reader
= GetReader (PrimaryStore
);
1693 secondary_reader
= GetReader (SecondaryStore
);
1695 // Load everything from the primary index
1697 max_doc
= primary_reader
.MaxDoc ();
1698 for (int i
= 0; i
< max_doc
; ++i
) {
1700 if (primary_reader
.IsDeleted (i
))
1704 doc
= primary_reader
.Document (i
);
1707 hit
= DocumentToHit (doc
);
1708 all_hits
[hit
.Uri
] = hit
;
1711 // Now add in everything from the secondary index, if it exists
1712 if (secondary_reader
!= null) {
1713 max_doc
= secondary_reader
.MaxDoc ();
1714 for (int i
= 0; i
< max_doc
; ++i
) {
1716 if (secondary_reader
.IsDeleted (i
))
1720 doc
= secondary_reader
.Document (i
);
1723 uri
= GetUriFromDocument (doc
);
1726 hit
= (Hit
) all_hits
[uri
];
1728 AddPropertiesToHit (hit
, doc
, false);
1732 ReleaseReader (primary_reader
);
1733 ReleaseReader (secondary_reader
);