Make available the XPath path of the node where a SyntaxError occurred during Schema validation

[ryanong]

What problem is this PR intended to solve?

When validating an XML document with a schema it is difficult to impossible to know where the error happened in the document. With the path on the error, we can identify where in the document the error occurred.

Have you included adequate test coverage?

I believe so. I updated the test in the schema to check to see if path errors were being returned correctly.

Does this change affect the behavior of either the C or the Java implementations?

This change the C implementation gently. Setting the path during error creation.
I looked if xerces-j could do it, but I couldn't find anything on the SAXException that would help.

[flavorjones]

@ryanong Thanks. Looks like some rubocop failures and appveyor is showing test failures.

[ryanong]

It looks like path only returns when validating an in memory document and not a file. Should I document this in the path description?

[flavorjones]

I'd like some DOM parser tests for SyntaxError#path, too. Specifically I'd like separate tests for XML and HTML4 to make sure those parsers' behavior is captured.

Don't worry about JRuby, I can spike on support and if I can't make it work I'll make the decision about whether to just document it as "libxml only".

[flavorjones]

It looks like path only returns when validating an in memory document and not a file. Should I document this in the path description?

That's strange, if you can commit a test for each (one passing, one failing) I will investigate to check if that's intended libxml2 behavior.

[flavorjones]

I'd like some DOM parser tests for SyntaxError#path, too. Specifically I'd like separate tests for XML and HTML4 to make sure those parsers' behavior is captured.

Can you please also exercise in-memory and from-file parsing for these two parsers as well, just to capture the current behavior. Thanks.

[ryanong]

It appears that html doesn't raise Nokogiri::XML::SyntaxError It uses Nokogiri::SyntaxError

I'll add some tests for doc parsing for xml and html

[ryanong]

I think node is returning null because validate_file uses xmlSchemaValidateFile which uses sax parsing and doesn't have a node. There might be a way to get the path from the node->ctxt which is a xmlParserCtxt in sax parsing mode https://gnome.pages.gitlab.gnome.org/libxml2/devhelp/libxml2-tree.html#xmlParserCtxt

But this is way over might head

[flavorjones]

@ryanong Can we reset for a minute and talk about your original motivation for this change? What does the path get you that the line/column info won't suffice? Is there another simpler way we can solve this problem?

[ryanong]

I'm attempting to validate an XML file with an XSD and then associate the errors with the inputs that create the file.

users = [
  { id: "1", name: "bob" },
  { id: "2", name: "boboboboboboboboboobobobbob"},
]
doc = build_doc(users)
# Nokogiri::XML
# <users>
#   <user id="1">
#     <name>bob</name>
#   </user>
#   <user id="2">
#     <name>boboboboboboboboboobobobbob</name>
#   </user>
# </users>

xsd_errors = validate(doc)
# [Nokigiri::XML::SyntaxError.new(path: "/users/user[2]/name")] # name has too many characters 
mapped_errors = []
xsd_errors.each do |error|
  id = doc.at_xpath(error.path).xpath("ancestor::user")["id"]
  user = users.find { |u| u[:id] == id }
  mapped_errors << { error: error, parent: user }
end

That way I can return each error to associated user.

Is there a way to know what node there is at a given line a column? I imagine I could do that by walking the entire tree and indexing it and matching against that but this feels cleaner

[flavorjones]

Ah, I see. Sorry, it's been a long week for me and I'm just now understanding fully what you're trying to do and why it's not working for input files. All of this makes sense.

I'm fine with documenting that SyntaxError#path is populated only when there is an actual DOM object to reference ("best efforts") and that it doesn't work in JRuby.

In other places I've used a string like this to document differences between xerces/neko and libxml2/gumbo:

⚠ HTML5 functionality is not available when running JRuby.

so something like that is probably fine for this.

[flavorjones]

@ryanong Thank you again! Will be in the v1.17 release (hopefully soon).

[flavorjones]

Changelog updated in 1e06cd1