Newlines and whitespace in XML literals are removed since v1.0.0

[spacid]

In versions 1.0.0 (and 1.1.0), the XML triplifier removes newlines and whitespace within element text nodes, even when trim-strings=false is provided as property. This behavior worked correctly in versions 0.8.2 (and 0.9.0).

To reproduce

Using the following XML file:

test.xml

<root>
    <description>Line 1
    Line 2</description>
</root>

With the following SPARQL query:

test.sparql

PREFIX fx: <http://sparql.xyz/facade-x/ns/>
CONSTRUCT {
  ?s ?p ?o
} WHERE {
  SERVICE <x-sparql-anything:location=test.xml,trim-strings=false> {
    ?s ?p ?o .
    FILTER(isLiteral(?o))
  }
}

A small Python script to test the behavior:

import subprocess

JARS = {
    "v0.8.2": "sparql-anything-0.8.2.jar",
    "v0.9.0": "sparql-anything-0.9.0.jar",
    "v1.0.0": "sparql-anything-v1.0.0.jar",
    "v1.1.0": "sparql-anything-v1.1.0.jar",
}


def run_sparql_anything(version: str, jar_path: str):
    print(f"{version}\n-------")
    cmd = ["java", "-jar", jar_path, "-q", "test.sparql", "-f", "TTL"]
    try:
        result = subprocess.run(cmd, capture_output=True, text=True, check=True)
        print(result.stdout)
    except Exception as e:
        print(f"Error in {version}: {e}")


if __name__ == "__main__":
    for version, jar_path in JARS.items():
        run_sparql_anything(version, jar_path)

The output:

v0.8.2
-------
@prefix fx: <http://sparql.xyz/facade-x/ns/> .

[ <http://www.w3.org/1999/02/22-rdf-syntax-ns#_1>
          "Line 1\n    Line 2" ] .

v0.9.0
-------
@prefix fx: <http://sparql.xyz/facade-x/ns/> .

[ <http://www.w3.org/1999/02/22-rdf-syntax-ns#_1>
          "Line 1\n    Line 2" ] .

v1.0.0
-------
PREFIX fx: <http://sparql.xyz/facade-x/ns/>

[ <http://www.w3.org/1999/02/22-rdf-syntax-ns#_1>
          "Line 1Line 2" ] .

v1.1.0
-------
PREFIX fx: <http://sparql.xyz/facade-x/ns/>

[ <http://www.w3.org/1999/02/22-rdf-syntax-ns#_1>
          "Line 1Line 2" ] .

[enridaga]

This is a bug, unrelated to the trim-strings option (which remove spaces and newlines at the start or the end of the string). This bug depends on how the code collects XML events of type char stream, by trimming each block. Working on this, I also noticed a number of tests that assumed string values to be trimmed, which is not correct (trim-strings defaults to false, and happens later anyway).