Double-quote characters in regex patterns create invalid dataclasses

[ScottKelly]

Unexpected Behavior

If there is a pattern restriction in an XML file similar to this

	<xs:simpleType name="mytype">
		<xs:restriction base="xs:token">
			<xs:minLength value="1"/>
			<xs:maxLength value="256"/>
			<xs:pattern value="([^\\ \? &gt; &lt; \* / &quot; : |]{1,256})"/>
		</xs:restriction>
	</xs:simpleType>

xsdata will generate a field with an unescaped quote character in the metadata dict which looks like the following:

    mytype: Optional[str] = field(
        default=None,
        metadata={
            "type": "Attribute",
            "sa": Column(String),
            "required": True,
            "min_length": 1,
            "max_length": 256,
            "pattern": r"([^\\ \? > < \* / " : |]{1,256})",
        }
    )

This terminates the regex string early and causes a syntax error

Expected Behavior

The expected behavior should produce this:

    mytype: Optional[str] = field(
        default=None,
        metadata={
            "type": "Attribute",
            "sa": Column(String),
            "required": True,
            "min_length": 1,
            "max_length": 256,
            "pattern": r"([^\\ \? > < \* / \" : |]{1,256})",
        }
    )

which has the double-quote escaped in the regex pattern.

The double-quote in the regex pattern in the XML is valid so the issue is the dataclass file is generated using double-quotes around strings but doesn't have any logic to escape double-quotes in regex patterns. However, double-quotes are properly escaped for all other strings. This issue is only because of the special if for patterns located here

https://github.com/tefra/xsdata/blob/master/xsdata/formats/dataclass/filters.py#L433

##Proposed Change
Using the existing behavior of text.escape_string wouldn't be valid for regexes since we don't want to escape other characters. We only need to escape double-quotes in the regex. Also, we shouldn't escape already escaped double-quotes.

I can create a PR where I add a simple regex sub like this:

        if key == "pattern":
            value = re.sub(r'([^\\])\"', r'\1\\"', data)
            return f'r"{value}"'

so we only escape double-quotes if they aren't already escaped.

Let me know if you like that solution and I'll create a PR.

[tefra]

Thanks for reporting @ScottKelly, please go ahead I will accept your solution

Down the road I want to add some sort of translation between xsd patterns and python regexes but that's not something I am looking forward to...