Implement lint and return LintError directly

Instead of creating and then parsing a string with a regular expression,
we can instead build and return a list of `LintError`s (a typed dict)
directly and skip some indirection.

I am not sure if `lint` is *supposed* to be overridden like this, but I
do remember advocating for it a couple years ago when the change for a
proper data structure for lint results was discussed. It just now
happened to be a convenient way to implement end markers for the
highlighted region.

In this particular instance, it allows us to specify a full region for
marking the error, where we now specify the entire match including the
remainder of the comment instead of just the trigger word. This can be
changed, of course, but I figured it wouldn't be a bad idea initially.

linter.py

@@ -12,7 +12,9 @@
 """This module exports the Annotations plugin class."""
 
 import re
-from SublimeLinter.lint import Linter, ERROR, WARNING
+
+import sublime
+from SublimeLinter.lint import Linter, ERROR, WARNING, util
 
 
 def _escape_words(values):
@@ -37,11 +39,6 @@ class Annotations(Linter):
 
     cmd = None
     line_col_base = (0, 0)
-    regex = (
-        r'^(?P<line>\d+):(?P<col>\d+):'
-        r' (?P<error_type>.+?) \((?P<code>.+)\):'
-        r' (?P<message>.*)'
-    )
 
     # We use this to do the matching
     mark_regex_template = r'(?:(?P<info>{infos})|(?P<warning>{warnings})|(?P<error>{errors})):?\s*(?P<message>.*)'
@@ -54,7 +51,13 @@ class Annotations(Linter):
         'infos': ['NOTE', 'README', 'INFO'],
     }
 
-    def run(self, cmd, code):
+    def lint(self, _code, _view_has_changed):
+        self.logger.info(
+            "%s: linting '%s'",
+            self.name,
+            util.canonical_filename(self.view),
+        )
+
         options = {}
         for option in ('errors', 'warnings', 'infos'):
             words = self.settings.get(option)
@@ -63,7 +66,7 @@ def run(self, cmd, code):
         mark_regex = re.compile(self.mark_regex_template.format_map(options))
 
         output = []
-        regions = self.view.find_by_selector('comment - punctuation.definition.comment')
+        regions = self.view.find_by_selector(self.settings['comment_selector'])
 
         for region in regions:
             region_offset = self.view.rowcol(region.a)
@@ -76,6 +79,7 @@ def run(self, cmd, code):
                 row = region_offset[0] + i
                 # Need to account for region column offset only in first row
                 col = match.start() + (region_offset[1] if i == 0 else 0)
+                match_region = sublime.Region(region.a + match.start(), region.a + match.end())
                 message = match.group('message').strip() or '<no message>'
                 word = match.group('error')
                 if word:
@@ -88,7 +92,19 @@ def run(self, cmd, code):
                         word = match.group('info')
                         error_type = 'info'
 
-                output.append('{row}:{col}: {error_type} ({word}): {message}'
-                              .format(**locals()))
-
-        return '\n'.join(output)
+                output.append(dict(
+                    line=row,
+                    start=col,
+                    region=match_region,
+                    # linter=,
+                    error_type=error_type,
+                    code=word,
+                    msg=message,
+                    filename=util.get_filename(self.view),
+                    # uid=,
+                    # priority=,
+                    # panel_line=,
+                    offending_text=word,
+                ))
+
+        return output