Elasticsearch 之（31）fielddata原理初探
1、《Elasticsearch 之（6）kibana嵌套聚合，下钻分析，聚合分析》提到 对于分词的field执行aggregation，发现报错
GET /test_index/test_type/_search
{
"aggs": {
"group_by_test_field": {
"terms": {
"field": "test_field"
}
}
}
}{
"error": {
"root_cause": [
{
"type": "illegal_argument_exception",
"reason": "Fielddata is disabled on text fields by default. Set fielddata=true on [test_field] in order to load fielddata in memory by uninverting the inverted index. Note that this can however use significant memory."
}
],
"type": "search_phase_execution_exception",
"reason": "all shards failed",
"phase": "query",
"grouped": true,
"failed_shards": [
{
"shard": 0,
"index": "test_index",
"node": "4onsTYVZTjGvIj9_spWz2w",
"reason": {
"type": "illegal_argument_exception",
"reason": "Fielddata is disabled on text fields by default. Set fielddata=true on [test_field] in order to load fielddata in memory by uninverting the inverted index. Note that this can however use significant memory."
}
}
],
"caused_by": {
"type": "illegal_argument_exception",
"reason": "Fielddata is disabled on text fields by default. Set fielddata=true on [test_field] in order to load fielddata in memory by uninverting the inverted index. Note that this can however use significant memory."
}
},
"status": 400
}对分词的field，直接执行聚合操作，会报错，大概意思是说，你必须要打开fielddata，然后将正排索引数据加载到内存中，才可以对分词的field执行聚合操作，而且会消耗很大的内存
2、如果要对分词的field执行聚合操作，必须将fielddata设置为true
POST /test_index/_mapping/test_type
{
"properties": {
"test_field": {
"type": "text",
"fielddata": true
}
}
}
{
"test_index": {
"mappings": {
"test_type": {
"properties": {
"test_field": {
"type": "text",
"fields": {
"keyword": {
"type": "keyword",
"ignore_above": 256
}
},
"fielddata": true
}
}
}
}
}
}GET /test_index/test_type/_search
{
"size": 0,
"aggs": {
"group_by_test_field": {
"terms": {
"field": "test_field"
}
}
}
}
{
"took": 23,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 2,
"max_score": 0,
"hits": []
},
"aggregations": {
"group_by_test_field": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": "test",
"doc_count": 2
}
]
}
}
}3、使用内置field不分词，对string field进行聚合
如果对不分词的field执行聚合操作，直接就可以执行，不需要设置fieldata=true（keyword在256字符内忽略分词）
GET /test_index/test_type/_search
{
"size": 0,
"aggs": {
"group_by_test_field": {
"terms": {
"field": "test_field.keyword"
}
}
}
}{
"took": 3,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 2,
"max_score": 0,
"hits": []
},
"aggregations": {
"group_by_test_field": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": "test",
"doc_count": 2
}
]
}
}
}4、分词field+fielddata的工作原理
doc value --> 不分词的所有field，可以执行聚合操作 --> 如果你的某个field不分词，那么在index-time，就会自动生成doc value --> 针对这些不分词的field执行聚合操作的时候，自动就会用doc value来执行
分词field，是没有doc value的。在index-time，如果某个field是分词的，那么是不会给它建立doc value正排索引的，因为分词后，占用的空间过于大，所以默认是不支持分词field进行聚合的
分词field默认没有doc value，所以直接对分词field执行聚合操作，是会报错的
对于分词field，必须打开和使用fielddata，完全存在于纯内存中，结构和doc value类似。如果是ngram或者是大量term，那么必将占用大量的内存。
如果一定要对分词的field执行聚合，那么必须将fielddata=true，然后es就会在执行聚合操作的时候，现场将field对应的数据，建立一份fielddata正排索引，fielddata正排索引的结构跟doc value是类似的，但是只会讲fielddata正排索引加载到内存中来，然后基于内存中的fielddata正排索引执行分词field的聚合操作
如果直接对分词field执行聚合，报错，才会让我们开启fielddata=true，告诉我们，会将fielddata uninverted index，正排索引，加载到内存，会耗费内存空间
为什么fielddata必须在内存？因为大家自己思考一下，分词的字符串，需要按照term进行聚合，需要执行更加复杂的算法和操作，如果基于磁盘和os cache，那么性能会很差
