Line breaks not parsed correctly #1497
Description
Line breaks (e.g. the two spaces at the end of a line) are not parsed correctly. Line breaks should emit a <br> tag, but InlineParser emits it as a Text node, not an element node. This results in a correct rendering when using the supplied HtmlRenderer, but makes other NodeVistor implementations useless.
For example, consider this node visitor, which should print an element tree:
class AstPrinter implements md.NodeVisitor {
int indent = '';
void visitElementAfter(md.Element element) {
print('${indent}leaving: ${element.tag}');
indent = indent.substring(0, indent.length-2);
}
bool visitElementBefore(md.Element element) {
indent += ' ';
if (element.isEmpty) {
print('${indent}void: ${element.tag}');
indent = indent.substring(0, indent.length-2);
return false;
} else {
print('${indent}entering: ${element.tag}');
return true;
}
}
void visitText(md.Text node) {
indent += ' ';
List<String> lines = node.text.replaceAll('\r\n', '\n').split('\n');
for (String line in lines) {
print('${indent}TEXT> ${line}');
}
indent = indent.substring(0, indent.length-2);
}
}
The driver code looks like this:
List<String> lines = ['*hello* **world**!', '', '> this is a quote', 'good bye!'];
md.Document doc = new md.Document();
AstPrinter printer = new AstPrinter();
for (md.Node node in doc.parseLines(lines)) {
node.accept(printer);
}
(I've imported markdown as md so it doesn't conflict with core:html.)
The output of this example is:
entering: p
entering: em
TEXT> hello
leaving: em
TEXT>
entering: strong
TEXT> world
leaving: strong
TEXT> !
leaving: p
entering: blockquote
entering: p
TEXT> this is a quote
leaving: p
leaving: blockquote
entering: p
TEXT> good bye!
leaving: p
(Side note, since there are no unit tests for proper AST construction, this AstPrinter would be a handy class to include in the Markdown library for debugging purposes.)
What tree is printed when a linebreak is parsed? Here's a test:
List<String> lines = ['this is a test ', 'this is only a test'];
The result:
entering: p
TEXT> this is a test
TEXT> <br />
TEXT> this is only a test
leaving: p
The <br /> is emitted as a text node, not an element! So a NodeVisitor can't see <br /> elements, which makes AST analysis impractical and also makes rendering to other formats (e.g. PDF) buggy: you'll get a literal <br /> in your PDF instead of the line break that you expected.
Note: there are multiple unit tests covering line breaks, but these tests only compare output in HTML form, not in AST form. AFAICT, there are zero tests on the AST.