Package org.apache.accumulo.core.iterators

Class Combiner

java.lang.Object
org.apache.accumulo.core.iterators.WrappingIterator
org.apache.accumulo.core.iterators.Combiner
All Implemented Interfaces:
OptionDescriber, SortedKeyValueIterator<Key,Value>, YieldingKeyValueIterator<Key,Value>
Direct Known Subclasses:
TypedValueCombiner
public abstract class Combiner extends WrappingIterator implements OptionDescriber
A SortedKeyValueIterator that combines the Values for different versions (timestamp) of a Key within a row into a single Value. Combiner will replace one or more versions of a Key and their Values with the most recent Key and a Value which is the result of the reduce method. An IteratorSetting.Column which only specifies a column family will combine all Keys in that column family individually. Similarly, a IteratorSetting.Column which specifies a column family and column qualifier will combine all Keys in column family and qualifier individually. Combination is only ever performed on multiple versions and not across column qualifiers or column visibilities.

Implementations must provide a reduce method: public Value reduce(Key key, Iterator<Value> iter).

This reduce method will be passed the most recent Key and an iterator over the Values for all non-deleted versions of that Key. A combiner will not combine keys that differ by more than the timestamp.

This class and its implementations do not automatically filter out unwanted columns from those being combined, thus it is generally recommended to use a Combiner implementation with the ScannerBase.fetchColumnFamily(Text) or ScannerBase.fetchColumn(Text, Text) methods.

WARNING : Using deletes with Combiners may not work as intended. See setReduceOnFullCompactionOnly(IteratorSetting, boolean)

  • Field Details

  • Constructor Details

    • Combiner

      public Combiner()

  • Method Details

    • getTopKey

      public Key getTopKey()
      Description copied from interface: SortedKeyValueIterator
      Returns top key. Can be called 0 or more times without affecting behavior of next() or hasTop(). Note that in minor compaction scope and in non-full major compaction scopes the iterator may see deletion entries. These entries should be preserved by all iterators except ones that are strictly scan-time iterators that will never be configured for the minc or majc scopes. Deletion entries are only removed during full major compactions.

      For performance reasons, iterators reserve the right to reuse objects returned by getTopKey when SortedKeyValueIterator.next() is called, changing the data that the object references. Iterators that need to save an object returned by getTopKey ought to copy the object's data into a new object in order to avoid aliasing bugs.

      Specified by:
      getTopKey in interface SortedKeyValueIterator<Key,Value>
      Overrides:
      getTopKey in class WrappingIterator
      Returns:
      K

    • getTopValue

      public Value getTopValue()
      Description copied from interface: SortedKeyValueIterator
      Returns top value. Can be called 0 or more times without affecting behavior of next() or hasTop().

      For performance reasons, iterators reserve the right to reuse objects returned by getTopValue when SortedKeyValueIterator.next() is called, changing the underlying data that the object references. Iterators that need to save an object returned by getTopValue ought to copy the object's data into a new object in order to avoid aliasing bugs.

      Specified by:
      getTopValue in interface SortedKeyValueIterator<Key,Value>
      Overrides:
      getTopValue in class WrappingIterator
      Returns:
      V

    • hasTop

      public boolean hasTop()
      Description copied from interface: SortedKeyValueIterator
      Returns true if the iterator has more elements. Note that if this iterator has yielded (@see YieldingKeyValueIterator.enableYielding(YieldCallback)), this this method must return false.
      Specified by:
      hasTop in interface SortedKeyValueIterator<Key,Value>
      Overrides:
      hasTop in class WrappingIterator
      Returns:
      true if the iterator has more elements.

    • next

      public void next() throws IOException
      Description copied from interface: SortedKeyValueIterator
      Advances to the next K,V pair. Note that in minor compaction scope and in non-full major compaction scopes the iterator may see deletion entries. These entries should be preserved by all iterators except ones that are strictly scan-time iterators that will never be configured for the minc or majc scopes. Deletion entries are only removed during full major compactions.
      Specified by:
      next in interface SortedKeyValueIterator<Key,Value>
      Overrides:
      next in class WrappingIterator
      Throws:
      IOException - if an I/O error occurs.

    • seek

      public void seek(Range range, Collection<ByteSequence> columnFamilies, boolean inclusive) throws IOException
      Description copied from interface: SortedKeyValueIterator
      Seeks to the first key in the Range, restricting the resulting K,V pairs to those with the specified columns. An iterator does not have to stop at the end of the range. The whole range is provided so that iterators can make optimizations. Seek may be called multiple times with different parameters after SortedKeyValueIterator.init(org.apache.accumulo.core.iterators.SortedKeyValueIterator<K, V>, java.util.Map<java.lang.String, java.lang.String>, org.apache.accumulo.core.iterators.IteratorEnvironment) is called. Iterators that examine groups of adjacent key/value pairs (e.g. rows) to determine their top key and value should be sure that they properly handle a seek to a key in the middle of such a group (e.g. the middle of a row). Even if the client always seeks to a range containing an entire group (a,c), the tablet server could send back a batch of entries corresponding to (a,b], then reseek the iterator to range (b,c) when the scan is continued. columnFamilies is used, at the lowest level, to determine which data blocks inside of an RFile need to be opened for this iterator. This set of data blocks is also the set of locality groups defined for the given table. If no columnFamilies are provided, the data blocks for all locality groups inside of the correct RFile will be opened and seeked in an attempt to find the correct start key, regardless of the startKey in the range. In an Accumulo instance in which multiple locality groups exist for a table, it is important to ensure that columnFamilies is properly set to the minimum required column families to ensure that data from separate locality groups is not inadvertently read.
      Specified by:
      seek in interface SortedKeyValueIterator<Key,Value>
      Overrides:
      seek in class WrappingIterator
      Parameters:
      range - Range of keys to iterate over.
      columnFamilies - Collection of column families to include or exclude.
      inclusive - boolean that indicates whether to include (true) or exclude (false) column families.
      Throws:
      IOException - if an I/O error occurs.

    • reduce

      public abstract Value reduce(Key key, Iterator<Value> iter)
      Reduces a list of Values into a single Value.
      Parameters:
      key - The most recent version of the Key being reduced.
      iter - An iterator over the Values for different versions of the key.
      Returns:
      The combined Value.

    • init

      public void init(SortedKeyValueIterator<Key,Value> source, Map<String,String> options, IteratorEnvironment env) throws IOException
      Description copied from interface: SortedKeyValueIterator
      Initializes the iterator. Data should not be read from the source in this method.
      Specified by:
      init in interface SortedKeyValueIterator<Key,Value>
      Overrides:
      init in class WrappingIterator
      Parameters:
      source - SortedKeyValueIterator source to read data from.
      options - Map map of string option names to option values.
      env - IteratorEnvironment environment in which iterator is being run, provided by Accumulo itself and is expected to be non-null.
      Throws:
      IOException - unused.

    • deepCopy

      public SortedKeyValueIterator<Key,Value> deepCopy(IteratorEnvironment env)
      Description copied from interface: SortedKeyValueIterator
      Creates a deep copy of this iterator as though seek had not yet been called. init should be called on an iterator before deepCopy is called. init should not need to be called on the copy that is returned by deepCopy; that is, when necessary init should be called in the deepCopy method on the iterator it returns. The behavior is unspecified if init is called after deepCopy either on the original or the copy. A proper implementation would call deepCopy on the source.
      Specified by:
      deepCopy in interface SortedKeyValueIterator<Key,Value>
      Overrides:
      deepCopy in class WrappingIterator
      Parameters:
      env - IteratorEnvironment environment in which iterator is being run, provided by Accumulo itself and is expected to be non-null.
      Returns:
      SortedKeyValueIterator a copy of this iterator (with the same source and settings).

    • describeOptions

      public OptionDescriber.IteratorOptions describeOptions()
      Description copied from interface: OptionDescriber
      Gets an iterator options object that contains information needed to configure this iterator. This object will be used by the accumulo shell to prompt the user to input the appropriate information.
      Specified by:
      describeOptions in interface OptionDescriber
      Returns:
      an iterator options object

    • validateOptions

      public boolean validateOptions(Map<String,String> options)
      Description copied from interface: OptionDescriber
      Check to see if an options map contains all options required by an iterator and that the option values are in the expected formats.
      Specified by:
      validateOptions in interface OptionDescriber
      Parameters:
      options - a map of option names to option values
      Returns:
      true if options are valid, false otherwise (IllegalArgumentException preferred)

    • setColumns

      public static void setColumns(IteratorSetting is, List<IteratorSetting.Column> columns)
      A convenience method to set which columns a combiner should be applied to. For each column specified, all versions of a Key which match that @{link IteratorSetting.Column} will be combined individually in each row. This method is likely to be used in conjunction with ScannerBase.fetchColumnFamily(Text) or ScannerBase.fetchColumn(Text,Text).
      Parameters:
      is - iterator settings object to configure
      columns - a list of columns to encode as the value for the combiner column configuration

    • setCombineAllColumns

      public static void setCombineAllColumns(IteratorSetting is, boolean combineAllColumns)
      A convenience method to set the "all columns" option on a Combiner. This will combine all columns individually within each row.
      Parameters:
      is - iterator settings object to configure
      combineAllColumns - if true, the columns option is ignored and the Combiner will be applied to all columns

    • setReduceOnFullCompactionOnly

      public static void setReduceOnFullCompactionOnly(IteratorSetting is, boolean reduceOnFullCompactionOnly)
      Combiners may not work correctly with deletes. Sometimes when Accumulo compacts the files in a tablet, it only compacts a subset of the files. If a delete marker exists in one of the files that is not being compacted, then data that should be deleted may be combined. See ACCUMULO-2232 for more information. For correctness deletes should not be used with columns that are combined OR this option should be set to true.

      When this method is set to true all data is passed through during partial major compactions and no reducing is done. Reducing is only done during scan and full major compactions, when deletes can be correctly handled. Only reducing on full major compactions may have negative performance implications, leaving lots of work to be done at scan time.

      When this method is set to false, combiners will log an error if a delete is seen during any compaction. This can be suppressed by adjusting logging configuration. Errors will not be logged more than once an hour per Combiner, regardless of how many deletes are seen.

      This method was added in 1.6.5 and 1.7.1. If you want your code to work in earlier versions of 1.6 and 1.7 then do not call this method. If not set this property defaults to false in order to maintain compatibility.

      Since:
      1.6.5, 1.7.1, 1.8.0