PHP - XSD schema validation error with unicode regular expression

Question

Description:

When I try to validate XML file with given XSD schema containing Unicode regular expression, the function DOMDocument::schemaValidate return a validation error. The XSD schema is W3C well formed and the validation pass with the other validation tools. The problem doesn't occur if the XSD pattern is format like this (without square brackets):

<xsd:pattern value="\P{Ll}+"/>

PHP Version: 5.2.14
LibXml Version: 2.7.7

The previous pattern [\P{Ll}]+ works correctly with preg_match function.

---

Test script:

PHP Validation Code:

function libxml_display_errors()
{
   $errors = libxml_get_errors();

   print_r($errors);

   libxml_clear_errors();
}

libxml_use_internal_errors(true);

$dom = new DOMDocument();
$dom->load('test.xml');

if ( !$dom->schemaValidate('test.xsd') ) {
  echo "XML Error\n";
  libxml_display_errors();
} else {
  echo "XML ok\n";
}

---

XSD Schema:

<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema" elementFormDefault="qualified" attributeFormDefault="unqualified">
    <xsd:simpleType name="noLowerCase">
        <xsd:restriction base="xsd:string">
            <xsd:pattern value="[\P{Ll}]+"/>
        </xsd:restriction>
    </xsd:simpleType>
    <xsd:complexType name="DatiUtenteType">
        <xsd:sequence>
            <xsd:element name="Cognome" type="noLowerCase"/>
            <xsd:element name="Nome" type="noLowerCase"/>
        </xsd:sequence>
    </xsd:complexType>
    <xsd:complexType name="DataExchangeFisso">
        <xsd:sequence>
            <xsd:element name="DatiUtente" type="DatiUtenteType"/>
        </xsd:sequence>
    </xsd:complexType>
    <xsd:element name="ListOfDataExchange">
        <xsd:complexType>
            <xsd:sequence>
                <xsd:element name="DataExchangeFisso" type="DataExchangeFisso" minOccurs="0" maxOccurs="unbounded"/>
            </xsd:sequence>
        </xsd:complexType>
    </xsd:element>
</xsd:schema>

---

XML File:

<?xml version="1.0" encoding="UTF-8"?>
<ListOfDataExchange>
  <DataExchangeFisso>
    <DatiUtente>
      <Cognome>FOO</Cognome>
      <Nome>BAR</Nome>
    </DatiUtente>
  </DataExchangeFisso>
</ListOfDataExchange>

---

Expected result:

XML ok

---

Actual result:

XML Error
Array
(
    [0] => LibXMLError Object
        (
            [level] => 2
            [code] => 1839
            [column] => 0
            [message] => Element 'Cognome': [facet 'pattern'] The value 'FOO' is not accepted by the pattern '[\P{Ll}]+'.
            [file] => /var/www/html/test.xml
            [line] => 5
        )

    [1] => LibXMLError Object
        (
            [level] => 2
            [code] => 1824
            [column] => 0
            [message] => Element 'Cognome': 'FOO' is not a valid value of the atomic type 'noLowerCase'.
            [file] => /var/www/html/test.xml
            [line] => 5
        )

    [2] => LibXMLError Object
        (
            [level] => 2
            [code] => 1839
            [column] => 0
            [message] => Element 'Nome': [facet 'pattern'] The value 'BAR' is not accepted by the pattern '[\P{Ll}]+'.
            [file] => /var/www/html/test.xml
            [line] => 6
        )

    [3] => LibXMLError Object
        (
            [level] => 2
            [code] => 1824
            [column] => 0
            [message] => Element 'Nome': 'BAR' is not a valid value of the atomic type 'noLowerCase'.
            [file] => /var/www/html/test.xml
            [line] => 6
        )
)

Answer 1

This is not a complete answer to your question, but probably some clarification:

The regular expressions in XSD, even it might be similar to those with preg_match, are a different thing. So assuming that something has to work with XSD because it does work with preg_match is a guess, but not a strict test.

The category^Docs Letters lowercase with it's property Ll is defined by Unicode, the XSD library should support it.

Probably it's a problem with the negativity of the category, because it only says what not, but not what in a non-negative character class.

Try:

[^\p{Ll}]+

Answer 2

The schema is good, your schema processor has a bug or non-conformance. It's actually not unusual for schema processors to implement regex dialects somewhat different from the dialect defined in the XSD specification: lazy implementors just pass the regex straight through to their chosen library.