BooleanQuery与TermInSetQuery分析

booleanQuery:
"must" : [
        { "term" : { "like" : "cooking" } },
        { "term" : { "property" : "bike" } }
      ]
termInsetQuery:
{ "terms": {"like": [ "cooking", "fishing", "swimming"]}}

BooleanQuery
BooleanWeight
TermQuery.scorer
查询单个term, 参考查询分析.
https://www.cnblogs.com/vsop/p/12152207.html

将子查询结果按照must, should, filter, must_not分类.
根据should, must, filter数量进行：
pure conjunction.
pure disjunction.
conjunction-disjunction mix.

BlockMaxConjunctionScorer
对多个postings求交集, advance时, 使用skipper.
approximations: 每个子查询的结果集.
BlockMaxConjunctionScorer$DocIdSetIterator.
如果每个approximation都有doc, 则命中该doc.
score: 该doc在每个子查询的score求和。
moveToNextBlock：target > upTo(当前block最大doc)时，move to next block。
advanceShallow(target)：每个子结果都move到包含target的block。
BlockImpactsDocsEnum 跳表实现
advanceShallow(target)：advance到包含target的block上.
SlowImpactsEnum 对于结果数少的postings，不需要skipping

DisjunctionScorer
使用heap, 对多个postings求并集.
DisiPriorityQueue
取堆top, updateTop, 如果newTop等于top, 继续updateTop.  
DisjunctionSumScorer, DisjunctionMaxScorer
postings合并后, 评分阶段, 一个doc有多个term时, 用来评分, sum或max。

TermInSetQuery
当terms数量不超过16时, rewrite成should boolean query.
否则使用DocIdSetBuilder构建bit set.
TermInSetQuery$ConstantScoreWeight.rewrite
DocIdSetBuilder
add(iter):
当前已有bitset, 如果iter也是FixedBitSet(bitmap, 而非roaring bitmap), 两个bitset求OR; 否则遍历iter, set到bitset.  
否则遍历iter add到buffer或者达到threshold转成bitset。

参考：
lucene8.7.0