Package org.apache.solr.common.cloud

Class CompositeIdRouter

  • public class CompositeIdRouter
extends HashBasedRouter
    CompositeIdRouter partitions ids based on a SEPARATOR, hashes each partition and merges the hashes together to map the id to a slice. This allows bucketing of like groups of ids.

    Allows basic separation split between 32 bits as example given below, or using bitsSeparator can specify exact bits 

     Example inputs:
 user!uniqueid
 app!user!uniqueid
 user/4!uniqueid
 app/2!user/4!uniqueid
    Lets say you had a set of records you want to index together such as a contact in a database, using a prefix of contact!contactid would allow all contact ids to be bucketed together. 
     An Example:
 If the id "contact!0000000KISS is passed 😘
 Take "contact"" and hash it with murmurhash3_x86_32
 result: -541354036
 bits: 11011111101110111001011111001100

 Take 0000000KISS and hash it with murmurhash3_x86_32
 result: 2099700320
 bits: 01111101001001101110001001100000

 Now we take the bits and apply a mask, since this is 32 bits the mask is the first 16 bits or the last 16 bits
 So uppermask = 0xFFFF0000  11111111111111110000000000000000
 So we bitwise AND to get half the original hash and only the upper 16 bits for 00T
 11011111101110111001011111001100
 11111111111111110000000000000000
 ________________________________
 11011111101110110000000000000000

 lowermask = 0x0000FFFF 00000000000000001111111111111111
 So we bitwise AND and get the lower 16 bits of the original hash for 0000000KISS
 01111101001001101110001001100000
 00000000000000001111111111111111
 ________________________________
 00000000000000001110001001100000

 Now we combine the hashes with a bitwise OR
 11011111101110110000000000000000
 00000000000000001110001001100000
 ________________________________
 11011111101110111110001001100000

 11011111101110111110001001100000 is the hash we return, bucketing the suffixed by prefix type prefix!suffix

    • Field Detail

      • SEPARATOR

        public static final char SEPARATOR
        This character separates a composite ID into a leading route key and the rest.

        Importantly, it's also used at the end of a provided route key parameter (which appears in many places) to designate a hash range which translates to a list of slices. If a route key does not end with this character, then semantically the key points to a single slice that holds a doc with that ID.

        See Also:
        Constant Field Values

    • Constructor Detail

      • CompositeIdRouter

        public CompositeIdRouter()

    • Method Detail

      • getRouteKeyWithSeparator

        public int getRouteKeyWithSeparator​(byte[] id,
                                    int idOffset,
                                    int idLength)
        Parse out the route key from id up to and including the SEPARATOR, returning it's length. If no route key is detected then 0 is returned.

      • getRouteKeyNoSuffix

        public String getRouteKeyNoSuffix​(String id)
        Parse out the route key from id. It will not have the "bits" suffix or separator, if present. If there is no route key found then null is returned.

      • keyHashRange

        public DocRouter.Range keyHashRange​(String routeKey)
        Get Range for a given CompositeId based route key
        Parameters:
        routeKey - to return Range for
        Returns:
        Range for given routeKey

      • getSearchRangeSingle

        public DocRouter.Range getSearchRangeSingle​(String shardKey,
                                            SolrParams params,
                                            DocCollection collection)
        Description copied from class: DocRouter
        This method is consulted to determine what search range (the part of the hash ring) should be queried for a request when an explicit shards parameter was not used. This method only accepts a single shard key (or null).
        Overrides:
        getSearchRangeSingle in class DocRouter

      • getSearchSlicesSingle

        public Collection<Slice> getSearchSlicesSingle​(String shardKey,
                                               SolrParams params,
                                               DocCollection collection)
        Description copied from class: DocRouter
        This method is consulted to determine what slices should be queried for a request when an explicit shards parameter was not used. This method only accepts a single shard key (or null). If you have a comma separated list of shard keys, call getSearchSlices
        Overrides:
        getSearchSlicesSingle in class HashBasedRouter

      • preprocessRouteKey

        protected String preprocessRouteKey​(String shardKey)
        Methods accepting a route key (shard key) can have this input preprocessed by a subclass before further analysis.