MarkLogic 9.0.8.2中XML转CSV的最优方案咨询及示例需求

Great question! Given your scenario—dealing with dynamic, inconsistent XML documents in MarkLogic 9.0.8.2 and needing to export to CSV with array data merged into single comma-separated columns—TDE (Template-Driven Extraction) is hands down the better choice over XSLT. Here's why, plus a practical example to get you started:

Why TDE beats XSLT for your use case #

XSLT works well for rigid, predictable XML structures, but it becomes unwieldy when dealing with dynamic documents where fields or array lengths vary across records. You'd end up writing endless conditional logic to handle missing elements or variable array counts, which is hard to maintain.

TDE is MarkLogic's native tool designed explicitly for extracting semi-structured data into row-based formats like CSV. It handles inconsistent structures out of the box:

• Automatically inserts empty values for missing fields instead of breaking the export
• Simplifies merging variable-length arrays into single comma-separated columns
• Is easier to extend or modify as your data structure evolves

Step-by-Step TDE Example #

1. Create a TDE Template #

First, define a template that maps your XML elements to CSV columns, including logic to merge arrays. Run this XQuery in the MarkLogic Query Console:

xquery version "1.0-ml";
import module namespace tde = "http://marklogic.com/xdmp/tde" at "/MarkLogic/tde.xqy";

let $article-template := 
  <template xmlns="http://marklogic.com/xdmp/tde">
    <!-- Target each <element1> node inside your <article> -->
    <context>/article/element1</context>
    <rows>
      <row>
        <columns>
          <!-- Extract fixed field e1 -->
          <column>
            <name>e1_id</name>
            <scalar-type>string</scalar-type>
            <val>e1/text()</val>
          </column>
          <!-- Extract year from e2 -->
          <column>
            <name>publication_year</name>
            <scalar-type>int</scalar-type>
            <val>e2/yr/text()</val>
          </column>
          <!-- Merge all <grp1> nodes into a single comma-separated column -->
          <column>
            <name>grp1_combined</name>
            <scalar-type>string</scalar-type>
            <val>string-join(../indexing/grp1/text(), ',')</val>
          </column>
          <!-- Merge all <grp2> nodes (adjust XPath if grp2 has child elements) -->
          <column>
            <name>grp2_combined</name>
            <scalar-type>string</scalar-type>
            <val>string-join(../indexing/grp2/text(), ',')</val>
          </column>
        </columns>
      </row>
    </rows>
  </template>

<!-- Save the template to MarkLogic -->
return tde:template-insert("article-csv-export-template", $article-template)

2. Export to CSV #

Use the tde:export-to-csv function to extract and save your data as a CSV file. Replace your-article-collection with the name of the collection where your XML documents are stored:

xquery version "1.0-ml";
import module namespace tde = "http://marklogic.com/xdmp/tde" at "/MarkLogic/tde.xqy";

let $collection-query := cts:collection-query("your-article-collection")
let $csv-options := map:entry("header", "true") -- Include CSV header row

-- Export and save the CSV to a filesystem path (adjust path as needed)
let $csv-content := tde:export-to-csv($collection-query, (), $csv-options)
return xdmp:save("/data/exports/articles_output.csv", $csv-content)

Key Notes for Your Dynamic Data #

• If some <element1> nodes lack fields like yr or grp1, TDE will automatically leave those CSV columns blank—no extra logic required.
• If <grp1> or <grp2> contain child elements instead of plain text, adjust the XPath in the <val> tag (e.g., string-join(../indexing/grp1/child-node/text(), ',')).
• Use tde:validate($article-template) to check for template errors before running the export.

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