索引

如何索引和搜索 JSON 文档

除了索引 Redis 哈希之外,Redis Stack 还可以索引 JSON 文档。

先决条件

在索引和搜索 JSON 文档之前,您需要一个具有以下任一功能的数据库:

  • Redis Stack,自动包含 JSON 以及搜索和查询功能
  • Redis v6.x 或更高版本,安装并启用以下模块:
    • RediSearch v2.2 或更高版本
    • RedisJSON v2.0 或更高版本

使用 JSON 架构创建索引

使用命令创建索引时FT.CREATE,请包含ON JSON关键字来索引数据库中存储的任何现有和未来的 JSON 文档。

要定义SCHEMA,您可以提供JSONPath表达式。每个 JSONPath 表达式的结果都已编入索引,并与名为attribute(以前称为field)的逻辑名称相关联。您可以在查询中使用这些属性。

笔记:
注意:attribute是可选的FT.CREATE

使用以下语法创建 JSON 索引:

FT.CREATE {index_name} ON JSON SCHEMA {json_path} AS {attribute} {type}

例如,此命令创建一个索引,该索引对代表库存项目的每个 JSON 文档的名称、描述、价格和图像向量嵌入进行索引:

127.0.0.1:6379> FT.CREATE itemIdx ON JSON PREFIX 1 item: SCHEMA $.name AS name TEXT $.description as description TEXT $.price AS price NUMERIC $.embedding AS embedding VECTOR FLAT 6 DIM 4 DISTANCE_METRIC L2 TYPE FLOAT32

有关 JSON 索引限制的更多详细信息,请参阅索引限制SCHEMA

添加 JSON 文档

创建索引后,Redis Stack 会自动为数据库中存储的任何现有、已修改或新创建的 JSON 文档编制索引。对于现有文档,索引编制在后台异步运行,因此文档可能需要一些时间才能使用。已修改和新创建的文档会同步编制索引,因此在添加或修改命令完成时,文档将可用。

您可以使用任何 JSON 写入命令(例如JSON.SETJSON.ARRAPPEND)来创建或修改 JSON 文档。

以下示例使用这些 JSON 文档来表示单个库存项目。

第 1 项 JSON 文档:

{
  "name": "Noise-cancelling Bluetooth headphones",
  "description": "Wireless Bluetooth headphones with noise-cancelling technology",
  "connection": {
    "wireless": true,
    "type": "Bluetooth"
  },
  "price": 99.98,
  "stock": 25,
  "colors": [
    "black",
    "silver"
  ],
  "embedding": [0.87, -0.15, 0.55, 0.03]
}

第 2 项 JSON 文档:

{
  "name": "Wireless earbuds",
  "description": "Wireless Bluetooth in-ear headphones",
  "connection": {
    "wireless": true,
    "type": "Bluetooth"
  },
  "price": 64.99,
  "stock": 17,
  "colors": [
    "black",
    "white"
  ],
  "embedding": [-0.7, -0.51, 0.88, 0.14]
}

用于JSON.SET将这些文档存储在数据库中:

127.0.0.1:6379> JSON.SET item:1 $ '{"name":"Noise-cancelling Bluetooth headphones","description":"Wireless Bluetooth headphones with noise-cancelling technology","connection":{"wireless":true,"type":"Bluetooth"},"price":99.98,"stock":25,"colors":["black","silver"],"embedding":[0.87,-0.15,0.55,0.03]}'
"OK"
127.0.0.1:6379> JSON.SET item:2 $ '{"name":"Wireless earbuds","description":"Wireless Bluetooth in-ear headphones","connection":{"wireless":true,"type":"Bluetooth"},"price":64.99,"stock":17,"colors":["black","white"],"embedding":[-0.7,-0.51,0.88,0.14]}'
"OK"

由于在这种情况下索引是同步的,因此命令返回后文档将立即在索引中可用JSON.SET。任何与索引内容匹配的后续查询都将返回该文档。

搜索索引

要在索引中搜索 JSON 文档,请使用命令FT.SEARCH。您可以搜索 中定义的任何属性SCHEMA

例如,使用此查询搜索名称中带有单词“earbuds”的商品:

127.0.0.1:6379> FT.SEARCH itemIdx '@name:(earbuds)'
1) "1"
2) "item:2"
3) 1) "$"
   2) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"connection\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"embedding\":[-0.7,-0.51,0.88,0.14]}"

此查询搜索描述中包含“蓝牙”和“耳机”的所有项目:

127.0.0.1:6379> FT.SEARCH itemIdx '@description:(bluetooth headphones)'
1) "2"
2) "item:1"
3) 1) "$"
   2) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"], \"embedding\":[0.87,-0.15,0.55,0.03]}"
4) "item:2"
5) 1) "$"
   2) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"connection\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"embedding\":[-0.7,-0.51,0.88,0.14]}"

现在搜索价格在70以下的蓝牙耳机:

127.0.0.1:6379> FT.SEARCH itemIdx '@description:(bluetooth headphones) @price:[0 70]'
1) "1"
2) "item:2"
3) 1) "$"
   2) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"connection\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"embedding\":[-0.7,-0.51,0.88,0.14]}"

最后,搜索与嵌入为 [1.0, 1.0, 1.0, 1.0] 的图像最相似的蓝牙耳机:

127.0.0.1:6379> FT.SEARCH itemIdx '@description:(bluetooth headphones)=>[KNN 2 @embedding $blob]' PARAMS 2 blob \x01\x01\x01\x01 DIALECT 2
1) "2"
2) "item:1"
3) 1) "__embedding_score"
   2) "1.08280003071"
   1) "$"
   2) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"],\"embedding\":[0.87,-0.15,0.55,0.03]}"
2) "item:2"
3) 1) "__embedding_score"
   2) "1.54409992695"
   3) "$"
   4) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"connection\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"embedding\":[-0.7,-0.51,0.88,0.14]}"

有关搜索查询的更多信息,请参阅搜索查询语法

笔记:
FT.SEARCH查询需要attribute修饰符。请勿在查询中使用 JSONPath 表达式,因为查询解析器不完全支持它们。

将 JSON 数组索引为 TAG

使用 JSON数组为具有多值术语的 JSON 字段建立索引的首选方法是使用 JSON 数组。数组的每个值都已建立索引,并且这些值必须是标量。如果要将字符串或布尔值作为 JSON 数组中的标签建立索引,请使用JSONPath通配符运算符。

要索引某个项目的可用颜色列表,请在索引创建期间$.colors.*在定义中指定 JSONPath :SCHEMA

127.0.0.1:6379> FT.CREATE itemIdx2 ON JSON PREFIX 1 item: SCHEMA $.colors.* AS colors TAG $.name AS name TEXT $.description as description TEXT

现在您可以搜索银色耳机:

127.0.0.1:6379> FT.SEARCH itemIdx2 "@colors:{silver} (@name:(headphones)|@description:(headphones))"
1) "1"
2) "item:1"
3) 1) "$"
   2) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"]}"

将 JSON 数组作为文本进行索引

从 RediSearch v2.6.0 开始,可以在字符串数组或指向多个字符串的 JSONPath 上进行全文搜索。

如果要将多个字符串值索引为 TEXT,请使用指向单个字符串数组的 JSONPath,或指向多个字符串值的 JSONPath,并使用 JSONPath 运算符(例如通配符、过滤器、联合、数组切片和/或递归下降)。

要索引某个项目的可用颜色列表,请在索引创建期间$.colors在定义中指定 JSONPath :SCHEMA

127.0.0.1:6379> FT.CREATE itemIdx3 ON JSON PREFIX 1 item: SCHEMA $.colors AS colors TEXT $.name AS name TEXT $.description as description TEXT
127.0.0.1:6379> JSON.SET item:3 $ '{"name":"True Wireless earbuds","description":"True Wireless Bluetooth in-ear headphones","connection":{"wireless":true,"type":"Bluetooth"},"price":74.99,"stock":20,"colors":["red","light blue"]}'
"OK"

现在您可以对浅色耳机进行全文搜索:

127.0.0.1:6379> FT.SEARCH itemIdx3 '@colors:(white|light) (@name|description:(headphones))' RETURN 1 $.colors
1) (integer) 2
2) "item:2"
3) 1) "$.colors"
   2) "[\"black\",\"white\"]"
4) "item:3"
5) 1) "$.colors"
   2) "[\"red\",\"light blue\"]"

限制

  • 当 JSONPath 可能导致多个值而不仅仅是单个数组时(例如,当 JSONPath 包含通配符等时),指定SLOPINORDERFT.SEARCH返回错误,因为与 JSONPath 匹配的值的顺序未明确定义,从而导致可能不一致的结果。

    $..b[*]例如,在 JSON 值上使用 JSONPath,例如

    {
       "a": [
          {"b": ["first first", "first second"]},
          {"c":
             {"b": ["second first", "second second"]}},
          {"b": ["third first", "third second"]}
       ]
    }
    

    可能会按各种顺序匹配值,具体取决于所使用的 JSONPath 库的具体实现。

    由于SLOPINORDER考虑索引值之间的相对顺序,并且结果可能会在未来版本中发生变化,因此将返回错误。

  • 当 JSONPath 导致多个值时:

    • 字符串值已编入索引
    • null值被跳过
    • 任何其他值类型都会导致索引失败
  • SORTBY仅按第一个值排序

  • HIGHLIGHT支持

  • RETURNSchema 属性的 JSONPath 指向多个值,仅返回第一个值(作为 JSON 字符串)

  • 如果 JSONPath 由 而不是 Schema 属性指定RETURN,则返回所有值(作为 JSON 字符串)

处理不同数组槽中的短语:

索引时,使用预定义的增量来增加多个文本值的数组槽之间的位置偏移。此增量控制不同数组槽中短语之间的分隔级别(与SLOP的参数相关FT.SEARCH)。此预定义值由配置参数设置MULTI_TEXT_SLOP(在模块加载时)。默认值为 100。

将 JSON 数组索引为 NUMERIC

从 RediSearch v2.6.1 开始,可以在数值数组或导致多个数值​​的 JSONPath 上进行搜索。

如果要将多个数值索引为 NUMERIC,请使用指向单个数字数组的 JSONPath,或指向多个数字的 JSONPath,并使用 JSONPath 运算符(例如通配符、过滤器、联合、数组切片和/或递归下降)。

max_level例如,将可用的音量(以分贝为单位)添加到项目列表中:

127.0.0.1:6379> JSON.SET item:1 $ '{"name":"Noise-cancelling Bluetooth headphones","description":"Wireless Bluetooth headphones with noise-cancelling technology","connection":{"wireless":true,"type":"Bluetooth"},"price":99.98,"stock":25,"colors":["black","silver"], "max_level":[60, 70, 80, 90, 100]}'
OK

127.0.0.1:6379> JSON.SET item:2 $ '{"name":"Wireless earbuds","description":"Wireless Bluetooth in-ear headphones","connection":{"wireless":true,"type":"Bluetooth"},"price":64.99,"stock":17,"colors":["black","white"], "max_level":[80, 100, 120]}'
OK

127.0.0.1:6379> JSON.SET item:3 $ '{"name":"True Wireless earbuds","description":"True Wireless Bluetooth in-ear headphones","connection":{"wireless":true,"type":"Bluetooth"},"price":74.99,"stock":20,"colors":["red","light blue"], "max_level":[90, 100, 110, 120]}'
OK

要索引数组,请在索引创建期间在定义中max_level指定 JSONPath :$.max_levelSCHEMA

127.0.0.1:6379> FT.CREATE itemIdx4 ON JSON PREFIX 1 item: SCHEMA $.max_level AS dB NUMERIC
OK

您现在可以搜索具有特定最大音量级别的耳机,例如,在 70 到 80 之间(含),返回max_level数组中至少有一个值在请求范围内的项目:

127.0.0.1:6379> FT.SEARCH itemIdx4 '@dB:[70 80]'
1) (integer) 2
2) "item:1"
3) 1) "$"
   2) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"],\"max_level\":[60,70,80,90,100]}"
4) "item:2"
5) 1) "$"
   2) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"max_level\":[80,100,120]}"

您还可以搜索所有值都在特定范围内的项目。例如,所有值都在 [90, 120] 范围内(含):

127.0.0.1:6379> FT.SEARCH itemIdx4 '-@dB:[-inf (90] -@dB:[(120 +inf]'
1) (integer) 1
2) "item:3"
3) 1) "$"
   2) "{\"name\":\"True Wireless earbuds\",\"description\":\"True Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":74.99,\"stock\":20,\"colors\":[\"red\",\"light blue\"],\"max_level\":[90,100,110,120]}"

限制

当 JSONPath 导致多个数值​​时:

  • 数值被索引
  • null值被跳过
  • 任何其他值类型都会导致索引失败

将 JSON 数组作为 GEO 进行索引

从 RediSearch v2.6.1 开始,可以对 geo(地理)值数组或导致多个地理值的 JSONPath 进行搜索。

在 RediSearch v2.6.1 之前,每个 GEO 属性仅支持一个地理值。地理值使用逗号分隔的字符串指定,格式为“经度,纬度”。例如,“15.447083,78.238306”。

在 RediSearch v2.6.1 中,还支持此类地理值的 JSON 数组。

为了索引多个地理值,用户可以使用指向单个地理值数组的 JSONPath,或者指向多个地理值的 JSONPath,使用 JSONPath 运算符(例如通配符、过滤器、联合、数组切片和/或递归下降)。

  • null值被跳过
  • 其他值将导致索引失败(布尔值、数字、对象、数组、格式错误的 GEO 字符串、无效坐标)

例如,将 添加到商品列表中vendor_id,即可以实际购买商品的位置:

127.0.0.1:6379> JSON.SET item:1 $ '{"name":"Noise-cancelling Bluetooth headphones","description":"Wireless Bluetooth headphones with noise-cancelling technology","connection":{"wireless":true,"type":"Bluetooth"},"price":99.98,"stock":25,"colors":["black","silver"], "max_level":[60, 70, 80, 90, 100], "vendor_id": [100,300]}'
OK

127.0.0.1:6379> JSON.SET item:2 $ '{"name":"Wireless earbuds","description":"Wireless Bluetooth in-ear headphones","connection":{"wireless":true,"type":"Bluetooth"},"price":64.99,"stock":17,"colors":["black","white"], "max_level":[80, 100, 120], "vendor_id": [100,200]}'
OK

127.0.0.1:6379> JSON.SET item:3 $ '{"name":"True Wireless earbuds","description":"True Wireless Bluetooth in-ear headphones","connection":{"wireless":true,"type":"Bluetooth"},"price":74.99,"stock":20,"colors":["red","light blue"], "max_level":[90, 100, 110, 120], "vendor_id": [100]}'
OK

现在添加一些供应商及其地理位置:

127.0.0.1:6379> JSON.SET vendor:1 $ '{"id":100, "name":"Kwik-E-Mart", "location":["35.213,31.785", "35.178,31.768", "35.827,31.984"]}'
OK

127.0.0.1:6379> JSON.SET vendor:2 $ '{"id":200, "name":"Cypress Creek", "location":["34.638,31.79", "34.639,31.793"]}'
OK

127.0.0.1:6379> JSON.SET vendor:3 $ '{"id":300, "name":"Barneys", "location":["34.648,31.817", "34.638,31.806", "34.65,31.785"]}'
OK

要索引数字数组,请在索引创建期间在定义中vendor_id指定 JSONPath :$.vendor_idSCHEMA

127.0.0.1:6379> FT.CREATE itemIdx5 ON JSON PREFIX 1 item: SCHEMA $.vendor_id AS vid NUMERIC
OK

要索引地理数组,请在索引创建期间在定义中location指定 JSONPath :$.locationSCHEMA

127.0.0.1:6379> FT.CREATE vendorIdx ON JSON PREFIX 1 vendor: SCHEMA $.location AS loc GEO
OK

现在搜索特定位置附近的供应商。例如,客户位于地理坐标 34.5,31.5,而您想要获取距离我们位置 40 公里范围内的供应商:

127.0.0.1:6379> FT.SEARCH vendorIdx '@loc:[34.5 31.5 40 km]' return 1 $.id
1) (integer) 2
2) "vendor:2"
3) 1) "$.id"
   1) "200"
4) "vendor:3"
5) 1) "$.id"
   1) "300"

现在寻找这些供应商提供的产品:

127.0.0.1:6379> FT.SEARCH itemIdx5 '@vid:[200 300]'
1) (integer) 2
2) "item:2"
3) 1) "$"
   2) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"max_level\":[80,100,120],\"vendor_id\":[100,200]}"
4) "item:1"
5) 1) "$"
   2) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"],\"max_level\":[60,70,80,90,100],\"vendor_id\":[100,300]}"

将 JSON 数组作为 VECTOR 进行索引

从 RediSearch 2.6.0 开始,您可以在索引模式中将 JSONPath 索引为 VECTOR 类型的数值数组。

例如,假设您的 JSON 项目包含一个向量嵌入数组,其中每个向量代表一个产品的图像。要为这些向量编制索引,请$.embedding在索引创建期间在架构定义中指定 JSONPath:

127.0.0.1:6379> FT.CREATE itemIdx5 ON JSON PREFIX 1 item: SCHEMA $.embedding AS embedding VECTOR FLAT 6 DIM 4 DISTANCE_METRIC L2 TYPE FLOAT32
OK
127.0.0.1:6379> JSON.SET item:1 $ '{"name":"Noise-cancelling Bluetooth headphones","description":"Wireless Bluetooth headphones with noise-cancelling technology","price":99.98,"stock":25,"colors":["black","silver"],"embedding":[0.87,-0.15,0.55,0.03]}'
OK
127.0.0.1:6379> JSON.SET item:2 $ '{"name":"Wireless earbuds","description":"Wireless Bluetooth in-ear headphones","price":64.99,"stock":17,"colors":["black","white"],"embedding":[-0.7,-0.51,0.88,0.14]}'
OK

现在,您可以使用向量搜索 KNN 查询来搜索与图像嵌入最相似的两副耳机。(请注意,从方言 2 开始支持向量查询。)例如:

127.0.0.1:6379> FT.SEARCH itemIdx5 '*=>[KNN 2 @embedding $blob AS dist]' SORTBY dist PARAMS 2 blob \x01\x01\x01\x01 DIALECT 2
1) (integer) 2
2) "item:1"
3) 1) "dist"
   2) "1.08280003071"
   3) "$"
   4) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"],\"embedding\":[0.87,-0.15,0.55,0.03]}"
4) "item:2"
5) 1) "dist"
   2) "1.54409992695"
   3) "$"
   4) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"embedding\":[-0.7,-0.51,0.88,0.14]}"

如果要将多个数字数组索引为 VECTOR,请使用JSONPath运算符(例如通配符、过滤器、联合、数组切片和/或递归下降)指向多个数字数组。

例如,假设您的 JSON 项目包含一个向量嵌入数组,其中每个向量代表同一产品的不同图像。要为这些向量编制索引,请$.embeddings[*]在索引创建期间在架构定义中指定 JSONPath:

127.0.0.1:6379> FT.CREATE itemIdx5 ON JSON PREFIX 1 item: SCHEMA $.embeddings[*] AS embeddings VECTOR FLAT 6 DIM 4 DISTANCE_METRIC L2 TYPE FLOAT32
OK
127.0.0.1:6379> JSON.SET item:1 $ '{"name":"Noise-cancelling Bluetooth headphones","description":"Wireless Bluetooth headphones with noise-cancelling technology","price":99.98,"stock":25,"colors":["black","silver"],"embeddings":[[0.87,-0.15,0.55,0.03]]}'
OK
127.0.0.1:6379> JSON.SET item:2 $ '{"name":"Wireless earbuds","description":"Wireless Bluetooth in-ear headphones","price":64.99,"stock":17,"colors":["black","white"],"embeddings":[[-0.7,-0.51,0.88,0.14],[-0.8,-0.15,0.33,-0.01]]}'
OK
重要提示:
与 NUMERIC 类型的情况不同,设置静态路径(例如$.embedding在 VECTOR 类型的架构中)不允许您索引存储在该字段下的多个向量。因此,如果您将其设置$.embedding为索引架构的路径,则在 JSON 中的字段中指定向量数组embedding将导致索引失败。

现在,您可以使用向量搜索 KNN 查询来搜索与图像嵌入最相似的两款耳机。(请注意,从方言 2 开始支持向量查询。)文档与查询向量之间的距离定义为查询向量与与架构中指定的 JSONPath 匹配的向量之间的最小距离。例如:

127.0.0.1:6379> FT.SEARCH itemIdx5 '*=>[KNN 2 @embeddings $blob AS dist]' SORTBY dist PARAMS 2 blob \x01\x01\x01\x01 DIALECT 2
1) (integer) 2
2) "item:2"
3) 1) "dist"
   2) "0.771500051022"
   3) "$"
   4) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"],\"embeddings\":[[-0.7,-0.51,0.88,0.14],[-0.8,-0.15,0.33,-0.01]]}"
4) "item:1"
5) 1) "dist"
   2) "1.08280003071"
   3) "$"
   4) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"],\"embeddings\":[[0.87,-0.15,0.55,0.03]]}"

请注意,0.771500051022是查询向量与[-0.8,-0.15,0.33,-0.01]嵌入数组中的第二个元素之间的 L2 距离,并且它低于查询向量与[-0.7,-0.51,0.88,0.14]嵌入数组中的第一个元素之间的 L2 距离。

有关向量相似性语法的更多信息,请参阅向量场

索引 JSON 对象

您无法索引 JSON 对象。如果 JSONPath 表达式返回一个对象,它将被忽略。

要对 JSON 对象的内容进行索引,您需要在单独的属性中对对象内的各个元素进行索引。

例如,要索引connectionJSON 对象,请在创建索引时将$.connection.wireless和字段定义为单独的属性:$.connection.type

127.0.0.1:6379> FT.CREATE itemIdx3 ON JSON SCHEMA $.connection.wireless AS wireless TAG $.connection.type AS connectionType TEXT
"OK"

创建新索引后,您可以搜索无线标签设置为的项目true

127.0.0.1:6379> FT.SEARCH itemIdx3 '@wireless:{true}'
1) "2"
2) "item:2"
3) 1) "$"
   2) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"connection\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"]}"
4) "item:1"
5) 1) "$"
   2) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"]}"

您还可以搜索具有蓝牙连接类型的商品:

127.0.0.1:6379> FT.SEARCH itemIdx3 '@connectionType:(bluetooth)'
1) "2"
2) "item:1"
3) 1) "$"
   2) "{\"name\":\"Noise-cancelling Bluetooth headphones\",\"description\":\"Wireless Bluetooth headphones with noise-cancelling technology\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":99.98,\"stock\":25,\"colors\":[\"black\",\"silver\"]}"
4) "item:2"
5) 1) "$"
   2) "{\"name\":\"Wireless earbuds\",\"description\":\"Wireless Bluetooth in-ear headphones\",\"connection\":{\"wireless\":true,\"type\":\"Bluetooth\"},\"price\":64.99,\"stock\":17,\"colors\":[\"black\",\"white\"]}"

场投影

FT.SEARCH默认返回整个 JSON 文档。如果希望将返回的搜索结果限制到特定的属性,可以使用字段投影。

返回特定属性

运行搜索查询时,您可以使用RETURN关键字指定要在搜索结果中包含的属性。您还需要指定要返回的字段数。

例如,此查询仅返回每组耳机的name和:price

127.0.0.1:6379> FT.SEARCH itemIdx '@description:(headphones)' RETURN 2 name price
1) "2"
2) "item:1"
3) 1) "name"
   2) "Noise-cancelling Bluetooth headphones"
   3) "price"
   4) "99.98"
4) "item:2"
5) 1) "name"
   2) "Wireless earbuds"
   3) "price"
   4) "64.99"

带有 JSONPath 的项目

您可以在语句中使用JSONPath表达式RETURN来提取 JSON 文档的任何部分,甚至是索引中未定义的字段SCHEMA

例如,以下查询使用 JSONPath 表达式$.stock除了返回名称和价格属性之外,还返回每件商品的库存。

127.0.0.1:6379> FT.SEARCH itemIdx '@description:(headphones)' RETURN 3 name price $.stock
1) "2"
2) "item:1"
3) 1) "name"
   2) "Noise-cancelling Bluetooth headphones"
   3) "price"
   4) "99.98"
   5) "$.stock"
   6) "25"
4) "item:2"
5) 1) "name"
   2) "Wireless earbuds"
   3) "price"
   4) "64.99"
   5) "$.stock"
   6) "17"

请注意,返回的属性名称是 JSONPath 表达式本身:"$.stock"

您可以使用AS选项为返回的属性指定别名:

127.0.0.1:6379> FT.SEARCH itemIdx '@description:(headphones)' RETURN 5 name price $.stock AS stock
1) "2"
2) "item:1"
3) 1) "name"
   2) "Noise-cancelling Bluetooth headphones"
   3) "price"
   4) "99.98"
   5) "stock"
   6) "25"
4) "item:2"
5) 1) "name"
   2) "Wireless earbuds"
   3) "price"
   4) "64.99"
   5) "stock"
   6) "17"

此查询将字段作为别名"stock"而不是 JSONPath 表达式返回"$.stock"

突出显示搜索词

您可以在任何索引属性中突出显示相关的搜索词TEXT

对于FT.SEARCH,您必须明确设置要在RETURNHIGHLIGHT参数后突出显示的属性。

使用可选TAGS关键字来指定围绕(或突出显示)匹配搜索词的字符串。

例如,在项目名称和描述中使用粗体 HTML 标签突出显示单词“蓝牙”:

127.0.0.1:6379> FT.SEARCH itemIdx '(@name:(bluetooth))|(@description:(bluetooth))' RETURN 3 name description price HIGHLIGHT FIELDS 2 name description TAGS '<b>' '</b>'
1) "2"
2) "item:1"
3) 1) "name"
   2) "Noise-cancelling <b>Bluetooth</b> headphones"
   3) "description"
   4) "Wireless <b>Bluetooth</b> headphones with noise-cancelling technology"
   5) "price"
   6) "99.98"
4) "item:2"
5) 1) "name"
   2) "Wireless earbuds"
   3) "description"
   4) "Wireless <b>Bluetooth</b> in-ear headphones"
   5) "price"
   6) "64.99"

使用 JSONPath 进行聚合

您可以使用聚合来生成统计数据或构建方面查询。

LOAD选项接受JSONPath表达式。您可以在管道中使用任何值,即使该值未被索引。

此示例使用聚合来计算每件商品 10% 的价格折扣,并按从最便宜到最昂贵的顺序对商品进行排序:

127.0.0.1:6379> FT.AGGREGATE itemIdx '*' LOAD 4 name $.price AS originalPrice APPLY '@originalPrice - (@originalPrice * 0.10)' AS salePrice SORTBY 2 @salePrice ASC
1) "2"
2) 1) "name"
   2) "Wireless earbuds"
   3) "originalPrice"
   4) "64.99"
   5) "salePrice"
   6) "58.491"
3) 1) "name"
   2) "Noise-cancelling Bluetooth headphones"
   3) "originalPrice"
   4) "99.98"
   5) "salePrice"
   6) "89.982"
笔记:
FT.AGGREGATE查询需要attribute修饰符。除了使用选项外,请勿在查询中使用 JSONPath 表达式LOAD,因为查询解析器不完全支持它们。

索引缺失值或空值

INDEXMISSING从 v2.10 开始,您可以使用选项与查询函数 withFT.CREATE结合使用来搜索缺失的属性,即给定文档中不存在的属性。您还可以使用选项 with搜索没有值(即空)的现有属性。两种查询类型都需要 DIALECT 2。示例如下:ismissingFT.SEARCHINDEXEMPTYFT.CREATE

JSON.SET key:1 $ '{"propA": "foo"}'
JSON.SET key:2 $ '{"propA": "bar", "propB":"abc"}'
FT.CREATE idx ON JSON PREFIX 1 key: SCHEMA $.propA AS propA TAG $.propB AS propB TAG INDEXMISSING

> FT.SEARCH idx 'ismissing(@propB)' DIALECT 2
1) "1"
2) "key:1"
3) 1) "$"
   2) "{\"propA\":\"foo\"}"
JSON.SET key:1 $ '{"propA": "foo", "propB":""}'
JSON.SET key:2 $ '{"propA": "bar", "propB":"abc"}'
FT.CREATE idx ON JSON PREFIX 1 key: SCHEMA $.propA AS propA TAG $.propB AS propB TAG INDEXEMPTY

> FT.SEARCH idx '@propB:{""}' DIALECT 2
1) "1"
2) "key:1"
3) 1) "$"
   2) "{\"propA\":\"foo\",\"propB\":\"\"}"

索引限制

架构映射

在创建索引时,需要将JSON元素映射到SCHEMA字段,如下所示:

  • 字符串为TEXTTAGGEO
  • 数字为NUMERIC
  • 布尔值为TAG
  • JSON 数组
    • 字符串数组为TAGTEXT
    • 数字数组为NUMERICVECTOR
    • 地理坐标数组为GEO
    • null此类数组中的值将被忽略。
  • 您无法为 JSON 对象编制索引。请将各个元素作为单独的属性进行编制索引。
  • null值将被忽略。

可排序标签

如果您为 JSON 文档创建索引,并且 JSONPath 指向一个数组或多个值,则排序只会考虑第一个值。

给此页面评分
返回顶部 ↑