MarkLogic中XML文档更新：保留规范部分的实现方法

Hey there! As a fellow MarkLogic user who's been in your shoes, I totally get the frustration of losing that canonical data when updating documents with the same URI. Let's break down the best ways to fix this, depending on your use case:

1. Targeted Node Replacement (Most Straightforward for XML) #

If your XML has a clear structure separating canonical data from updatable content—say a <canonical> section and a <data> section like this:

<document>
  <canonical>
    <id>123</id>
    <created-date>2024-01-01</created-date>
  </canonical>
  <data>
    <title>Old Title</title>
    <content>Old Content</content>
  </data>
</document>

You can use MarkLogic's xdmp:node-replace() function to swap out only the updatable part, leaving the canonical section untouched.

Here's a sample XQuery script:

let $target-doc := fn:doc("/path/to/your/document.xml")
let $new-data-section := 
  <data>
    <title>New Title</title>
    <content>Fresh, Updated Content</content>
  </data>
return
  xdmp:node-replace($target-doc//data, $new-data-section)

• This directly targets the <data> node in your existing document and replaces it with the new content—<canonical> stays exactly as it was.
• Pro tip: Add a check to make sure the <data> node exists before replacing it to avoid runtime errors.

2. Merge Updates (Great for Complex Scenarios or REST API) #

If you need more flexibility—like handling cases where the document might not exist yet, or you're working via the REST API—MarkLogic's merge update feature is your friend. It lets you specify exactly which parts of the document to update without overwriting the whole thing.

XQuery Example: #

xdmp:merge(
  "/path/to/your/document.xml",
  <document>
    <data>
      <title>New Title</title>
      <content>Fresh, Updated Content</content>
    </data>
  </document>,
  map:entry("merge-options", map:entry("replace", "data"))
)

The replace parameter tells MarkLogic to only overwrite the <data> node, preserving everything else.

REST API Example (cURL): #

If you're using the REST API, send a PUT request with the merge parameter to specify replacement:

curl -X PUT -H "Content-Type: application/xml" \
  -d '<data><title>New Title</title><content>Fresh Content</content></data>' \
  "http://your-ml-host:8000/v1/documents?uri=/path/to/your/document.xml&merge=replace&path=/document/data"

3. Store Canonical Data as Document Properties (Alternative Approach) #

If your canonical data doesn't need to live inside the XML itself, you can store it as document properties instead. Properties are separate from the document content, so updating the XML won't touch them.

Set Properties: #

xdmp:document-set-properties(
  "/path/to/your/document.xml",
  map:entry("canonical-id", "123"),
  map:entry("canonical-created-date", "2024-01-01")
)

Retrieve Properties Later: #

xdmp:document-properties("/path/to/your/document.xml")

This works best if your canonical data is simple key-value pairs rather than complex XML structures.

Quick Notes to Keep in Mind #

• Make sure your updates are idempotent—running the same update multiple times shouldn't break your data.
• If multiple processes might update the same document, use xdmp:lock-acquire() to prevent race conditions.
• Always test with fn:doc() after updating to verify your canonical data is still intact!

内容的提问来源于stack exchange，提问作者user8420733