Refactor sphinx.io; separate FileInput class for each file type

sphinx/application.py

@@ -83,6 +83,7 @@ builtin_extensions = (
     'sphinx.directives.code',
     'sphinx.directives.other',
     'sphinx.directives.patches',
+    'sphinx.io',
     'sphinx.parsers',
     'sphinx.roles',
     'sphinx.transforms.post_transforms',

sphinx/io.py

@@ -120,7 +120,9 @@ def SphinxDummySourceClass(source, *args, **kwargs):
     return source
 
 
-class SphinxFileInput(FileInput):
+class SphinxBaseFileInput(FileInput):
+    """A base class of SphinxFileInput."""
+
     def __init__(self, app, env, *args, **kwds):
         # type: (Sphinx, BuildEnvironment, Any, Any) -> None
         self.app = app
@@ -145,16 +147,7 @@ class SphinxFileInput(FileInput):
         # emit source-read event
         arg = [data]
         self.app.emit('source-read', self.env.docname, arg)
-        data = arg[0]
-
-        parser = self.app.registry.get_source_parser(self.source_path)
-        docinfo, data = split_docinfo(data)
-        if 'restructuredtext' in parser.supported:
-            if self.env.config.rst_epilog:
-                data = data + '\n' + self.env.config.rst_epilog + '\n'
-            if self.env.config.rst_prolog:
-                data = self.env.config.rst_prolog + '\n' + data
-        return docinfo + data
+        return arg[0]
 
     def warn_and_replace(self, error):
         # type: (Any) -> Tuple
@@ -172,11 +165,28 @@ class SphinxFileInput(FileInput):
         return (u'?', error.end)
 
 
+class SphinxFileInput(SphinxBaseFileInput):
+    pass
+
+
+class SphinxRSTFileInput(SphinxBaseFileInput):
+    def read(self):
+        # type: () -> unicode
+        data = SphinxBaseFileInput.read(self)
+        docinfo, data = split_docinfo(data)
+        if self.env.config.rst_epilog:
+            data = data + '\n' + self.env.config.rst_epilog + '\n'
+        if self.env.config.rst_prolog:
+            data = self.env.config.rst_prolog + '\n' + data
+        return docinfo + data
+
+
 def read_doc(app, env, filename):
     # type: (Sphinx, BuildEnvironment, unicode) -> nodes.document
     """Parse a document and convert to doctree."""
+    input_class = app.registry.get_source_input(filename)
     reader = SphinxStandaloneReader()
-    source = SphinxFileInput(app, env, source=None, source_path=filename,
+    source = input_class(app, env, source=None, source_path=filename,
                          encoding=env.config.source_encoding)
     parser = app.registry.create_source_parser(app, filename)
 
@@ -190,3 +200,8 @@ def read_doc(app, env, filename):
     pub.set_source(source, filename)
     pub.publish()
     return pub.document
+
+
+def setup(app):
+    app.registry.add_source_input('*', SphinxFileInput)
+    app.registry.add_source_input('restructuredtext', SphinxRSTFileInput)

sphinx/registry.py

@@ -30,6 +30,7 @@ if False:
     # For type annotation
     from typing import Any, Callable, Dict, Iterator, List, Type  # NOQA
     from docutils import nodes  # NOQA
+    from docutils.io import Input  # NOQA
     from docutils.parsers import Parser  # NOQA
     from sphinx.application import Sphinx  # NOQA
     from sphinx.builders import Builder  # NOQA
@@ -50,6 +51,7 @@ class SphinxComponentRegistry(object):
         self.builders = {}          # type: Dict[unicode, Type[Builder]]
         self.domains = {}           # type: Dict[unicode, Type[Domain]]
         self.source_parsers = {}    # type: Dict[unicode, Parser]
+        self.source_inputs = {}     # type: Dict[unicode, Input]
         self.translators = {}       # type: Dict[unicode, nodes.NodeVisitor]
 
     def add_builder(self, builder):
@@ -190,6 +192,28 @@ class SphinxComponentRegistry(object):
             parser.set_application(app)
         return parser
 
+    def add_source_input(self, filetype, input_class):
+        # type: (unicode, Type[Input]) -> None
+        if filetype in self.source_inputs:
+            raise ExtensionError(__('source_input for %r is already registered') % filetype)
+        self.source_inputs[filetype] = input_class
+
+    def get_source_input(self, filename):
+        # type: (unicode) -> Type[Input]
+        parser = self.get_source_parser(filename)
+        for filetype in parser.supported:
+            if filetype in self.source_inputs:
+                input_class = self.source_inputs[filetype]
+                break
+        else:
+            # use special source_input for unknown file-type '*' (if exists)
+            input_class = self.source_inputs.get('*')
+
+        if input_class is None:
+            raise SphinxError(__('source_input for %s not registered') % filename)
+        else:
+            return input_class
+
     def add_translator(self, name, translator):
         # type: (unicode, Type[nodes.NodeVisitor]) -> None
         self.translators[name] = translator