History

Viktor Lofgren b08e302dd5 (lexicon) Optimize lexicon by using Murmur3_128's hash function		2023-08-01 15:02:13 +02:00
..
src	(lexicon) Optimize lexicon by using Murmur3_128's hash function	2023-08-01 15:02:13 +02:00
build.gradle	(lexicon) Optimize lexicon by using Murmur3_128's hash function	2023-08-01 15:02:13 +02:00
readme.md	Update readme.md	2023-03-22 17:15:36 +01:00

readme.md

Lexicon

The lexicon contains a mapping for words to identifiers.

To ease index construction, it makes calculations easier if the domain of word identifiers is dense, that is, there is no gaps between ids; if there are 100 words, they're indexed 0-99 and not 5, 23, 107, 9999, 819235 etc. The lexicon exists to create such a mapping.

This lexicon is populated from a journal. The actual word data isn't mapped, but rather a 64 bit hash. As a result of the birthday paradox, colissions will be rare up until about to 2³² words.

The lexicon is constructed by processes/loading-process and read when services-core/index-service interprets queries.

Central Classes

KeywordLexicon
KeywordLexiconJournal
DictionaryMap comes in two versions
- OnHeapDictionaryMap - basically just a fastutil Long2IntOpenHashMap
- OffHeapDictionaryHashMap - a heavily modified trove TLongIntHashMap that uses off heap memory