extract-document-data comes as xml string element in json output - json

Question

I am trying to enrich my search results with some elements taken from the "matching" documents, using the query option "extract-document-data" like
<options xmlns="http://marklogic.com/appservices/search">
<extract-document-data selected="include">
<extract-path>/language-version/language-version-canonical-model/title</extract-path>
<extract-path>/language-version/language-version-canonical-model/language</extract-path>
</extract-document-data>
(...)
</options>
When I run the search and I ask for Json output (using the header Accept: application/json) I get as mix of json and "strinxml" as a result:
{
"snippet-format": "snippet",
"total": 564,
"start": 1,
"page-length": 10,
"selected": "include",
"results": [
{
"index": 1,
"uri": "ENV/CHEM/NANO(2015)22/ANN5/2",
"path": "fn:doc(\"ENV/CHEM/NANO(2015)22/ANN5/2\")",
(...)
"matches": [
{
"path": "fn:doc(\"ENV/CHEM/NANO(2015)22/ANN5/2\")/ns2:language-version/ns2:language-version-raw-data/*:document/*:page[22]",
(...)
}
],
"extracted": {
"kind": "element",
"content": [
"<language>En</language>",
"<title>ZINC OXIDE DOSSIERANNEX 5</title>",
"<reference>ENV/CHEM/NANO(2015)22/ANN5</reference>",
"<classification>2</classification>",
"<modificationDate>2015-04-16T00:00:00.000+02:00</modificationDate>",
"<subject label_en=\"media\" >media</subject>",
"<subject label_en=\"fish\" ">fish</subject>",
]
}
},
The problem here is with the "extracted" part, as you can see, it looks like the xml elements have been simply copied as string, when I would really expect them to be converted to json.
Does anybody have an idea about this problem?

Answer 1

MarkLogic won’t convert content. So, XML will remain XML when asking for JSON formatted search response. And since you can't really embed XML inside JSON, it gets serialized as a string.
You could try applying a REST transform on your search results, and use something like json:transform-to-json (probably with the custom config) to convert those on the fly. For instance something like this Server-side JavaScript transform:
/* jshint node:true,esnext:true */
/* global xdmp */
var json = require('/MarkLogic/json/json.xqy');
var config = json.config('custom');
function toJson(context, params, content) {
'use strict';
var response = content.toObject();
if (response.results) {
response.results.map(function(result) {
if (result.extracted && result.extracted.content) {
result.extracted.content.map(function(content, index) {
if (content.match(/^</) && !content.match(/^<!/)) {
result.extracted.content[index] = json.transformToJson(xdmp.unquote(content), config);
}
});
}
});
}
return response;
}
exports.transform = toJson;
You could also convert client-side of course.
HTH!

Answer 2

If you're using the Java Client API you can use the correct handle for each result to read the extracted items (see ExtractedResult and ExtractedItem):
SearchHandle results = queryManager.search(query, new SearchHandle());
for (MatchDocumentSummary summary : results.getMatchResults()) {
ExtractedResult extracted = summary.getExtracted();
// here we check to see if this result is XML format, and if so
// we use org.w3c.dom.Document
if ( Format.XML == summary.getFormat() ) {
for (ExtractedItem item : extracted) {
Document extractItem = item.getAs(new DOMHandle()).get();
...
}
// or if the result is JSON we could choose a different handle
} else if ( Format.JSON == summary.getFormat() ) {
for (ExtractedItem item : extracted) {
JsonNode extractItem = item.getAs(JsonNode.class);
...
}
}
}