重新索引后Elasticsearch不遵循_mapping [英] Elasticsearch doesn't obey _mapping after reindex

问题描述

我在弹性搜索中存储了一个名为 last_post_date 的字段。

I have a field called last_post_date stored in elasticsearch.

它显示为：

...
"last_post_date": "1485281760000",
...

并且映射为：

"last_post_date": {
    "type": "text",
    "fields": {
        "keyword": {
            "type": "keyword",
            "ignore_above": 256
        }
    }
},

现在，我创建一个新索引，其映射为：

Now I create a new index with the mapping as:

...
"last_post_date": {
    "type": "date",
    "format": "epoch_millis",
    "index": true,
    "doc_values": true
},
...

并通过以下方式进行完全重新索引：
http://127.0.0.1:9200/_reindex

And do a full reindex via: http://127.0.0.1:9200/_reindex

问题在于，即使重新索引了字段的值存储为strin在映射显示时用g代替日期：

The problem is that even after reindexing the field's value is stored as string instead of date while the mapping shows:

...
"last_post_date": {
    "type": "date",
    "store": true,
    "format": "epoch_millis"
},
...

为什么知道？

推荐答案

您会在源代码中看到，例如格式为 1485281760000 的类似字符串的时间戳，将不会更改，并且将始终反映您发送给Elasticsearch的完全相同的JSON。

What you see in the source, i.e. a string-like timestamp of the form "1485281760000" will not change and will always reflect the exact same JSON you sent to Elasticsearch.

但是，在被索引时，字符串 1485281760000 确实会被解释并索引为长值 1485281760000 。

However, when being indexed, the string "1485281760000" will be indeed interpreted and indexed as the long value 1485281760000.

如果您要更改JSON源，则将使用较长的时间戳而不是字符串时间戳进行发送，如下所示，那么源将显示您的较长时间戳。

If you were to change the JSON source you're sending with a long timestamp instead of a string timestamp, like below, then the source will show your a long timestamp.

...
"last_post_date": 1485281760000,
...

长话短说：

• 您在 _source 中看到的就是发送给ES的内容


• 将根据您创建的映射来解释和索引源的每个字段，即，字符串时间戳将被强制为长整数，并被解释为格式为 epoch_millis 的日期。


• 如果某些字段值无法根据您的映射正确地解释，索引和强制转换，则会出现错误，并且文档也将完全不被索引。

• what you see in the _source is exactly what you sent to ES
• each field of the source is interpreted and indexed according to the mapping you've created, i.e. the string timestamp will be coerced to a long and interpreted as a date of the format epoch_millis
• if some field value cannot be properly interpreted, indexed, coerced according to your mapping, you'll get an error and the document won't be indexed at all.

这篇关于重新索引后Elasticsearch不遵循_mapping的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！