/// <summary> Internal helper method used by check that iterates over
/// the keys of readerFieldToValIds and generates a Collection
/// of Insanity instances whenever two (or more) ReaderField instances are
/// found that have an ancestery relationships.
///
/// </summary>
/// <seealso cref="InsanityType.SUBREADER">
/// </seealso>
private List<Insanity> CheckSubreaders(MapOfSets<int,CacheEntry> valIdToItems, MapOfSets<ReaderField,int> readerFieldToValIds)
{
List<Insanity> insanity = new List<Insanity>(23);
Dictionary<ReaderField, Dictionary<ReaderField, ReaderField>> badChildren = new Dictionary<ReaderField, Dictionary<ReaderField, ReaderField>>(17);
MapOfSets<ReaderField, ReaderField> badKids = new MapOfSets<ReaderField, ReaderField>(badChildren); // wrapper
IDictionary<int, Dictionary<CacheEntry, CacheEntry>> viToItemSets = valIdToItems.GetMap();
IDictionary<ReaderField, Dictionary<int, int>> rfToValIdSets = readerFieldToValIds.GetMap();
Dictionary<ReaderField, ReaderField> seen = new Dictionary<ReaderField, ReaderField>(17);
foreach (ReaderField rf in rfToValIdSets.Keys)
{
if (seen.ContainsKey(rf))
continue;
System.Collections.IList kids = GetAllDecendentReaderKeys(rf.readerKey);
for (int i = 0; i < kids.Count; i++)
{
ReaderField kid = new ReaderField(kids[i], rf.fieldName);
if (badChildren.ContainsKey(kid))
{
// we've already process this kid as RF and found other problems
// track those problems as our own
badKids.Put(rf, kid);
badKids.PutAll(rf, badChildren[kid]);
badChildren.Remove(kid);
}
else if (rfToValIdSets.ContainsKey(kid))
{
// we have cache entries for the kid
badKids.Put(rf, kid);
}
if (!seen.ContainsKey(kid))
{
seen.Add(kid, kid);
}
}
if (!seen.ContainsKey(rf))
{
seen.Add(rf, rf);
}
}
// every mapping in badKids represents an Insanity
foreach (ReaderField parent in badChildren.Keys)
{
Dictionary<ReaderField,ReaderField> kids = badChildren[parent];
List<CacheEntry> badEntries = new List<CacheEntry>(kids.Count * 2);
// put parent entr(ies) in first
{
foreach (int val in rfToValIdSets[parent].Keys)
{
badEntries.AddRange(viToItemSets[val].Keys);
}
}
// now the entries for the descendants
foreach (ReaderField kid in kids.Keys)
{
foreach (int val in rfToValIdSets[kid].Keys)
{
badEntries.AddRange(viToItemSets[val].Keys);
}
}
insanity.Add(new Insanity(InsanityType.SUBREADER, "Found caches for decendents of " + parent.ToString(), badEntries.ToArray()));
}
return insanity;
}
/// <summary> Internal helper method used by check that iterates over
/// valMismatchKeys and generates a Collection of Insanity
/// instances accordingly. The MapOfSets are used to populate
/// the Insantiy objects.
/// </summary>
/// <seealso cref="InsanityType.VALUEMISMATCH">
/// </seealso>
private List<Insanity> CheckValueMismatch(MapOfSets<int,CacheEntry> valIdToItems, MapOfSets<ReaderField,int> readerFieldToValIds, Dictionary<ReaderField,ReaderField> valMismatchKeys)
{
List<Insanity> insanity = new List<Insanity>(valMismatchKeys.Count * 3);
if (!(valMismatchKeys.Count == 0))
{
// we have multiple values for some ReaderFields
IDictionary<ReaderField,Dictionary<int,int>> rfMap = readerFieldToValIds.GetMap();
IDictionary<int,Dictionary<CacheEntry,CacheEntry>> valMap = valIdToItems.GetMap();
foreach (ReaderField rf in valMismatchKeys.Keys)
{
List<CacheEntry> badEntries = new List<CacheEntry>(valMismatchKeys.Count * 2);
foreach (int val in rfMap[rf].Keys)
{
foreach (CacheEntry entry in valMap[val].Keys)
{
badEntries.Add(entry);
}
}
insanity.Add(new Insanity(InsanityType.VALUEMISMATCH, "Multiple distinct value objects for " + rf.ToString(), badEntries.ToArray()));
}
}
return insanity;
}
/// <summary> Tests a CacheEntry[] for indication of "insane" cache usage.
/// <p/>
/// NOTE:FieldCache CreationPlaceholder objects are ignored.
/// (:TODO: is this a bad idea? are we masking a real problem?)
/// <p/>
/// </summary>
public Insanity[] Check(CacheEntry[] cacheEntries)
{
if (null == cacheEntries || 0 == cacheEntries.Length)
return new Insanity[0];
if (null != ramCalc)
{
for (int i = 0; i < cacheEntries.Length; i++)
{
cacheEntries[i].EstimateSize(ramCalc);
}
}
// the indirect mapping lets MapOfSet dedup identical valIds for us
//
// maps the (valId) identityhashCode of cache values to
// sets of CacheEntry instances
MapOfSets<int,CacheEntry> valIdToItems = new MapOfSets<int,CacheEntry>(new Dictionary<int,Dictionary<CacheEntry,CacheEntry>>(17));
// maps ReaderField keys to Sets of ValueIds
MapOfSets<ReaderField,int> readerFieldToValIds = new MapOfSets<ReaderField,int>(new Dictionary<ReaderField,Dictionary<int,int>>(17));
//
// any keys that we know result in more then one valId
// TODO: This will be a HashSet<T> when we start using .NET Framework 3.5
Dictionary<ReaderField, ReaderField> valMismatchKeys = new Dictionary<ReaderField, ReaderField>();
// iterate over all the cacheEntries to get the mappings we'll need
for (int i = 0; i < cacheEntries.Length; i++)
{
CacheEntry item = cacheEntries[i];
System.Object val = item.GetValue();
if (val is Mono.Lucene.Net.Search.CreationPlaceholder)
continue;
ReaderField rf = new ReaderField(item.GetReaderKey(), item.GetFieldName());
System.Int32 valId = val.GetHashCode();
// indirect mapping, so the MapOfSet will dedup identical valIds for us
valIdToItems.Put(valId, item);
if (1 < readerFieldToValIds.Put(rf, valId))
{
if (!valMismatchKeys.ContainsKey(rf))
{
valMismatchKeys.Add(rf, rf);
}
}
}
List<Insanity> insanity = new List<Insanity>(valMismatchKeys.Count * 3);
insanity.AddRange(CheckValueMismatch(valIdToItems, readerFieldToValIds, valMismatchKeys));
insanity.AddRange(CheckSubreaders(valIdToItems, readerFieldToValIds));
return insanity.ToArray();
}
/// <summary>
/// Tests a CacheEntry[] for indication of "insane" cache usage.
/// <p>
/// <B>NOTE:</b>FieldCache CreationPlaceholder objects are ignored.
/// (:TODO: is this a bad idea? are we masking a real problem?)
/// </p>
/// </summary>
public Insanity[] Check(params FieldCache.CacheEntry[] cacheEntries)
{
if (null == cacheEntries || 0 == cacheEntries.Length)
{
return new Insanity[0];
}
if (EstimateRam)
{
for (int i = 0; i < cacheEntries.Length; i++)
{
cacheEntries[i].EstimateSize();
}
}
// the indirect mapping lets MapOfSet dedup identical valIds for us
// maps the (valId) identityhashCode of cache values to
// sets of CacheEntry instances
MapOfSets<int, FieldCache.CacheEntry> valIdToItems = new MapOfSets<int, FieldCache.CacheEntry>(new Dictionary<int, HashSet<FieldCache.CacheEntry>>(17));
// maps ReaderField keys to Sets of ValueIds
MapOfSets<ReaderField, int> readerFieldToValIds = new MapOfSets<ReaderField, int>(new Dictionary<ReaderField, HashSet<int>>(17));
// any keys that we know result in more then one valId
ISet<ReaderField> valMismatchKeys = new HashSet<ReaderField>();
// iterate over all the cacheEntries to get the mappings we'll need
for (int i = 0; i < cacheEntries.Length; i++)
{
FieldCache.CacheEntry item = cacheEntries[i];
object val = item.Value;
// It's OK to have dup entries, where one is eg
// float[] and the other is the Bits (from
// getDocWithField())
if (val is Bits)
{
continue;
}
if (val is Lucene.Net.Search.FieldCache.CreationPlaceholder)
{
continue;
}
ReaderField rf = new ReaderField(item.ReaderKey, item.FieldName);
int valId = val.GetHashCode();
// indirect mapping, so the MapOfSet will dedup identical valIds for us
valIdToItems.Put(valId, item);
if (1 < readerFieldToValIds.Put(rf, valId))
{
valMismatchKeys.Add(rf);
}
}
List<Insanity> insanity = new List<Insanity>(valMismatchKeys.Count * 3);
insanity.AddRange(CheckValueMismatch(valIdToItems, readerFieldToValIds, valMismatchKeys));
insanity.AddRange(CheckSubreaders(valIdToItems, readerFieldToValIds));
return insanity.ToArray();
}
/// <summary>
/// Internal helper method used by check that iterates over
/// the keys of readerFieldToValIds and generates a Collection
/// of Insanity instances whenever two (or more) ReaderField instances are
/// found that have an ancestry relationships.
/// </summary>
/// <seealso cref= InsanityType#SUBREADER </seealso>
private ICollection<Insanity> CheckSubreaders(MapOfSets<int, FieldCache.CacheEntry> valIdToItems, MapOfSets<ReaderField, int> readerFieldToValIds)
{
List<Insanity> insanity = new List<Insanity>(23);
Dictionary<ReaderField, HashSet<ReaderField>> badChildren = new Dictionary<ReaderField, HashSet<ReaderField>>(17);
MapOfSets<ReaderField, ReaderField> badKids = new MapOfSets<ReaderField, ReaderField>(badChildren); // wrapper
IDictionary<int, HashSet<FieldCache.CacheEntry>> viToItemSets = valIdToItems.Map;
IDictionary<ReaderField, HashSet<int>> rfToValIdSets = readerFieldToValIds.Map;
HashSet<ReaderField> seen = new HashSet<ReaderField>();
//IDictionary<ReaderField, ISet<int>>.KeyCollection readerFields = rfToValIdSets.Keys;
foreach (ReaderField rf in rfToValIdSets.Keys)
{
if (seen.Contains(rf))
{
continue;
}
IList<object> kids = GetAllDescendantReaderKeys(rf.ReaderKey);
foreach (object kidKey in kids)
{
ReaderField kid = new ReaderField(kidKey, rf.FieldName);
if (badChildren.ContainsKey(kid))
{
// we've already process this kid as RF and found other problems
// track those problems as our own
badKids.Put(rf, kid);
badKids.PutAll(rf, badChildren[kid]);
badChildren.Remove(kid);
}
else if (rfToValIdSets.ContainsKey(kid))
{
// we have cache entries for the kid
badKids.Put(rf, kid);
}
seen.Add(kid);
}
seen.Add(rf);
}
// every mapping in badKids represents an Insanity
foreach (ReaderField parent in badChildren.Keys)
{
HashSet<ReaderField> kids = badChildren[parent];
List<FieldCache.CacheEntry> badEntries = new List<FieldCache.CacheEntry>(kids.Count * 2);
// put parent entr(ies) in first
{
foreach (int value in rfToValIdSets[parent])
{
badEntries.AddRange(viToItemSets[value]);
}
}
// now the entries for the descendants
foreach (ReaderField kid in kids)
{
foreach (int value in rfToValIdSets[kid])
{
badEntries.AddRange(viToItemSets[value]);
}
}
FieldCache.CacheEntry[] badness = new FieldCache.CacheEntry[badEntries.Count];
badness = badEntries.ToArray();//LUCENE TO-DO had param of badness first
insanity.Add(new Insanity(InsanityType.SUBREADER, "Found caches for descendants of " + parent.ToString(), badness));
}
return insanity;
}
/// <summary>
/// Internal helper method used by check that iterates over
/// valMismatchKeys and generates a Collection of Insanity
/// instances accordingly. The MapOfSets are used to populate
/// the Insanity objects. </summary>
/// <seealso cref= InsanityType#VALUEMISMATCH </seealso>
private ICollection<Insanity> CheckValueMismatch(MapOfSets<int, FieldCache.CacheEntry> valIdToItems, MapOfSets<ReaderField, int> readerFieldToValIds, ISet<ReaderField> valMismatchKeys)
{
List<Insanity> insanity = new List<Insanity>(valMismatchKeys.Count * 3);
if (valMismatchKeys.Count != 0)
{
// we have multiple values for some ReaderFields
IDictionary<ReaderField, HashSet<int>> rfMap = readerFieldToValIds.Map;
IDictionary<int, HashSet<FieldCache.CacheEntry>> valMap = valIdToItems.Map;
foreach (ReaderField rf in valMismatchKeys)
{
IList<FieldCache.CacheEntry> badEntries = new List<FieldCache.CacheEntry>(valMismatchKeys.Count * 2);
foreach (int value in rfMap[rf])
{
foreach (FieldCache.CacheEntry cacheEntry in valMap[value])
{
badEntries.Add(cacheEntry);
}
}
FieldCache.CacheEntry[] badness = new FieldCache.CacheEntry[badEntries.Count];
badness = badEntries.ToArray(); //LUCENE TO-DO had param of badness before
insanity.Add(new Insanity(InsanityType.VALUEMISMATCH, "Multiple distinct value objects for " + rf.ToString(), badness));
}
}
return insanity;
}
