XML parsing with BigQuery UDF

2020-11-01

Simple BigQuery User Defined Function (UDF) that wil parse a well-formed XML bigquery STRING column into JSON.

The specific library used in this example to parse XML to JSON is xml-js

This is an extension of a StackOverfow post I came across recently. The specific extension is to further parse the output using standard BigQuer JSON functions to extract specific fields.


You can find the source here


this is not officially supported by google

Setup

You can either build the UDF or use the one in this repo xml_udf.js

To use the provided one, simply make a copy of it and place it into a GCS bucket in your project

To build, you’ll need nodejs

cd webpack-demo/
npm init -y
npm install webpack webpack-cli --save-dev
npm install --save xml-js

The actual code that does the UDF is at webpack-demo/src/index.ts

var convert = require('xml-js');

module.exports.XML2JSON = function (xml) {
    return convert.xml2json(xml, {compact: true, spaces: 4});
}

module.exports.JSON2XML = function (json) {
    var options = {compact: true, ignoreComment: true, spaces: 4};
    return convert.json2xml(json, options);
}

Then build the webpack

npx webpack --config webpack.config.js 

Edit the file xml_udf.js and add the contents of the webpack

---> contents of webpack-demo/dist/main.js <---

function frmXML(d) {
    c = xmllib.XML2JSON(d);
    return c;
}

Upload the file and finally, use the UDF

CREATE TEMP FUNCTION myFunc(a STRING)
  RETURNS STRING  
  LANGUAGE js AS
"""  
      return  frmXML(a);
"""    
OPTIONS (
  library=["gs://mineral-minutia-820/xml_udf.js"]
);

SELECT JSON_EXTRACT(myFunc(a),"$.html.body.li._text")  FROM UNNEST(["<html><title>Title of Page</title><body><li>hi sal </li></body></html>"]) as a;

Note the output of the UDF is JSON which i’ve parse out to extract just one field (hi sal):

images/bq_udf.png

Alternative XML Parsing

You can also use the following XML Parsers in your own UDF:

XSLT Support

This sample does not utilize XLST. Trust me, you don’t want to go there…but if you do, see fiduswriter/xslt-processor

This site supports webmentions. Send me a mention via this form.