我有一个相当复杂的映射，用于存储产品，并且在每个文档中，它包含每个客户预先计算的价格的嵌套数组。
索引中每个产品可能有多个版本（具有唯一代码）。替代产品按常见的xrefs_hash分组。我正在编写的查询需要为每个客户选择最佳产品（即xrefs_hash上的汇总/崩溃），然后根据prices.weight嵌套字段的值选择最佳产品。 / p>
prices.weight字段是一个浮点数，我们已经根据商店的客户设置（他们希望如何区分自己的商品的优先级）预先计算出该浮点数。根据这些设置（存储在prices.pricing_hash中的值创建的哈希值，以便在多个客户共享相同设置的情况下，我们可以存储一组定价。
该索引包含多达300,000个产品，一旦计算并插入所有价格，最终可以生成约100,000,000个文档。
映射看起来像这样（为简便起见，简称为：）
'mappings' => [
'_source' => [
'enabled' => true,
],
'dynamic' => false,
'properties' => [
'dealer_item_id' => [
'type' => 'integer',
],
'code' => [
'type' => 'text',
'analyzer' => 'custom_code_analyzer',
'fields' => [
'raw' => [
'type' => 'keyword',
],
],
],
'xrefs' => [
'type' => 'text',
'analyzer' => 'custom_code_analyzer',
'fields' => [
'raw' => [
'type' => 'keyword',
],
],
],
'xrefs_hash' => [
'type' => 'keyword',
],
'title' => [
'type' => 'text',
'analyzer' => 'custom_english_analyzer',
'fields' => [
'ngram_title' => [
'type' => 'text',
'analyzer' => 'custom_title_analyzer',
],
'raw' => [
'type' => 'keyword',
],
],
],
...
'prices' => [
'type' => 'nested',
'dynamic' => false,
'properties' => [
'pricing_hash' => [
'type' => 'keyword',
'index' => true,
],
'unit_price' => [
'type' => 'float',
'index' => true,
],
'pricebreaks' => [
'type' => 'object',
'dynamic' => false,
'properties' => [
'quantity' => [
'type' => 'integer',
'index' => false,
],
'price' => [
'type' => 'integer',
'index' => false,
],
],
],
'weight' => [
'type' => 'float',
'index' => true,
],
],
],
],
],
示例文档：
{
"dealer_item_id": 122023,
"code": "ABC123A",
"xrefs": [
"ABC123A",
"ABC123B",
],
"title": "Product A",
"xrefs_hash": "16d5415674c8365f63329b11ffc88da109590cec",
"prices": [
{
"pricebreaks": [
{
"quantity": 1,
"price": 9.75,
"contract": false
}
],
"weight": 0.20512820512820512,
"pricing_hash": "aabe06b7",
"unit_price": 9.75,
},
{
"pricebreaks": [
{
"quantity": 1,
"price": 9.75,
"contract": false
}
],
"weight": 0.20512820512820512,
"pricing_hash": "73643f3b",
"unit_price": 9.75,
}
]
},
{
"dealer_item_id": 124293,
"code": "ABC1234B",
"xrefs": [
"ABC123A",
"ABC123B",
],
"title": "Product B",
"xrefs_hash": "16d5415674c8365f63329b11ffc88da109590cec",
"prices": [
{
"contract_item": false,
"pricebreaks": [
{
"quantity": 1,
"price": 7.39,
"contract": false
}
],
"weight": 0.33829499323410017,
"pricing_hash": "aabe06b7",
"unit_price": 7.39,
},
{
"pricebreaks": [
{
"quantity": 1,
"price": 9.75,
"contract": false
}
],
"weight": 0.20512820512820512,
"pricing_hash": "73643f3b",
"unit_price": 9.75,
}
]
},
查询示例：
{
"track_total_hits": 100000,
"query": {
"bool": {
"filter": {
"bool": {
"must": [
{
"nested": {
"path": "prices",
"score_mode": "none",
"inner_hits": {
"_source": {
"include": [
"prices"
]
}
},
"query": {
"bool": {
"must": [
{
"term": {
"prices.pricing_hash": "aabe06b7"
}
}
]
}
}
}
},
{
"term": {
"code.raw": "RX58022"
}
}
],
"must_not": [
{
"term": {
"disabled": true
}
}
]
}
}
}
},
"_source": {
"includes": [
"code",
"dealer_item_id",
"title",
"xrefs"
]
},
"collapse": {
"field": "xrefs_hash",
"inner_hits": {
"name": "best_xrefs",
"sort": {
"prices.weight": "desc"
},
"size": 1
}
},
"aggregations": {
"xrefs_count": {
"cardinality": {
"field": "xrefs_hash",
"precision_threshold": 40000
}
}
}
}
我尝试使用折叠查询来选择最佳产品，但这似乎不支持按嵌套prices.weight字段进行排序。
我也尝试过基于xrefs_hash进行拼版，但这似乎无法在类别级别进行分页。
上面的示例查询几乎可以正常工作，但不会以正确的顺序返回折叠后的结果。检查查询时，似乎是用Infinity代替折叠排序，如果文档不包含排序字段，ES显然会这样做。
xref_hash值返回1个文档prices.weight值并与客户的pricing_hash相匹配的特定文档