lsp: Use pygls Documents for get_initial_doctree

The check for whether we should parse a document in
`get_initial_doctree` now relies on `get_location_type` rather than
checking the file extension. This means that files that don't have the
traditional `*.rst` extension but are treated by the client as
reStructuredText documents should now work as expected.

Additionally, by reading the document's source from the `Document`
maintained by `pygls` rather than reading from disk, document symbol
requests are more accurate taking into account any unsaved changes in
the user's editor.

This also removes the need for direct file system access - poentially
opening the door for running the vanilla docutils server in Pyodide?!

lib/esbonio/changes/434.enhancement.rst

@@ -0,0 +1 @@
+Document symbol requests made for unsaved files now use the language client's version rather than the version on disk.

lib/esbonio/changes/434.fix.rst

@@ -0,0 +1 @@
+The server should now handle document symbol requests for files that are treated as reStructuredText files by a language client but don't have an ``*.rst`` extension.

lib/esbonio/esbonio/lsp/rst/__init__.py

@@ -575,12 +575,17 @@ def get_initial_doctree(self, uri: str) -> Optional[Any]:
         uri
            Returns the doctree that corresponds with the given uri.
         """
-        filename = pathlib.Path(Uri.to_fs_path(uri))
-        try:
-            return read_initial_doctree(filename, self.logger)
-        except FileNotFoundError:
-            self.logger.debug(traceback.format_exc())
+
+        doc = self.workspace.get_document(uri)
+        loctype = self.get_location_type(doc, Position(line=1, character=0))
+
+        if loctype != "rst":
             return None
+
+        self.logger.debug("Getting initial doctree for: '%s'", Uri.to_fs_path(uri))
+
+        try:
+            return read_initial_doctree(doc, self.logger)
         except Exception:
             self.logger.error(traceback.format_exc())
             return None

lib/esbonio/esbonio/lsp/rst/io.py

@@ -1,23 +1,24 @@
 import logging
-import pathlib
 import typing
 from typing import Any
 from typing import Callable
 from typing import IO
 from typing import Optional
 from typing import Type
 
+import pygls.uris as uri
 from docutils import nodes
 from docutils.core import Publisher
-from docutils.io import FileInput
 from docutils.io import NullOutput
+from docutils.io import StringInput
 from docutils.parsers.rst import Directive
 from docutils.parsers.rst import directives
 from docutils.parsers.rst import Parser
 from docutils.parsers.rst import roles
 from docutils.readers.standalone import Reader
 from docutils.utils import Reporter
 from docutils.writers import Writer
+from pygls.workspace import Document
 from sphinx.environment import default_settings
 
 from esbonio.lsp.util.patterns import DIRECTIVE
@@ -174,9 +175,9 @@ def new_document(self) -> nodes.document:
 
 
 def read_initial_doctree(
-    filename: pathlib.Path, logger: logging.Logger
+    document: Document, logger: logging.Logger
 ) -> Optional[nodes.document]:
-    """Parse the given reStructuredText file into its "initial" doctree.
+    """Parse the given reStructuredText document into its "initial" doctree.
 
     An "initial" doctree can be thought of as the abstract syntax tree of a
     reStructuredText document. This method disables all role and directives
@@ -185,29 +186,26 @@ def read_initial_doctree(
 
     Parameters
     ----------
-    filename
-       Returns the doctree that corresponds with the given file.
+    document
+       The document containing the reStructuredText source.
 
     logger
        Logger to log debug info to.
     """
 
     parser = Parser()
-
-    if filename.suffix.replace(".", "") not in parser.supported:
-        return None
-
-    logger.debug("Getting initial doctree for: '%s'", filename)
     with disable_roles_and_directives():
         publisher = Publisher(
             reader=InitialDoctreeReader(logger),
             parser=parser,
             writer=DummyWriter(),
-            source_class=FileInput,
+            source_class=StringInput,
             destination=NullOutput(),
         )
         publisher.process_programmatic_settings(None, default_settings, None)
-        publisher.set_source(source_path=str(filename))
+        publisher.set_source(
+            source=document.source, source_path=uri.to_fs_path(document.uri)
+        )
         publisher.publish()
 
         return publisher.document