Simpler mechanism to extend input transformations

IPython/core/interactiveshell.py

@@ -339,6 +339,15 b' class InteractiveShell(SingletonConfigurable):'
                                          ())
 
     @property
+    def input_transformers_cleanup(self):
+        return self.input_transformer_manager.cleanup_transforms
+
+    input_transformers_post = List([],
+        help="A list of string input transformers, to be applied after IPython's "
+             "own input transformations."
+    )
+
+    @property
     def input_splitter(self):
         """Make this available for compatibility
 
@@ -2717,21 +2726,10 b' class InteractiveShell(SingletonConfigurable):'
         preprocessing_exc_tuple = None
         try:
             # Static input transformations
-            cell = self.input_transformer_manager.transform_cell(raw_cell)
-        except SyntaxError:
-            preprocessing_exc_tuple = sys.exc_info()
-            cell = raw_cell  # cell has to exist so it can be stored/logged
-        else:
-            if len(cell.splitlines()) == 1:
-                # Dynamic transformations - only applied for single line commands
-                with self.builtin_trap:
-                    try:
-                        # use prefilter_lines to handle trailing newlines
-                        # restore trailing newline for ast.parse
-                        cell = self.prefilter_manager.prefilter_lines(cell) + '\n'
+            cell = self.transform_cell(raw_cell)
                     except Exception:
-                        # don't allow prefilter errors to crash IPython
                         preprocessing_exc_tuple = sys.exc_info()
+            cell = raw_cell  # cell has to exist so it can be stored/logged
 
         # Store raw and processed history
         if store_history:
@@ -2803,6 +2801,24 b' class InteractiveShell(SingletonConfigurable):'
 
         return result
     
+    def transform_cell(self, raw_cell):
+        # Static input transformations
+        cell = self.input_transformer_manager.transform_cell(raw_cell)
+
+        if len(cell.splitlines()) == 1:
+            # Dynamic transformations - only applied for single line commands
+            with self.builtin_trap:
+                # use prefilter_lines to handle trailing newlines
+                # restore trailing newline for ast.parse
+                cell = self.prefilter_manager.prefilter_lines(cell) + '\n'
+
+        lines = cell.splitlines(keepends=True)
+        for transform in self.input_transformers_post:
+            lines = transform(lines)
+        cell = ''.join(lines)
+
+        return cell
+    
     def transform_ast(self, node):
         """Apply the AST transformations from self.ast_transformers
         

docs/source/config/inputtransforms.rst

@@ -24,34 +24,28 b' end of this stage, it must be valid Python syntax.'
    redesigned. Any third party code extending input transformation will need to
    be rewritten. The new API is, hopefully, simpler.
 
-String based transformations are managed by
-:class:`IPython.core.inputtransformer2.TransformerManager`, which is attached to
-the :class:`~IPython.core.interactiveshell.InteractiveShell` instance as
-``input_transformer_manager``. This passes the
-data through a series of individual transformers. There are two kinds of
-transformers stored in three groups:
-
-* ``cleanup_transforms`` and ``line_transforms`` are lists of functions. Each
-  function is called with a list of input lines (which include trailing
-  newlines), and they return a list in the same format. ``cleanup_transforms``
-  are run first; they strip prompts and leading indentation from input.
-  The only default transform in ``line_transforms`` processes cell magics.
-* ``token_transformers`` is a list of :class:`IPython.core.inputtransformer2.TokenTransformBase`
-  subclasses (not instances). They recognise special syntax like
-  ``%line magics`` and ``help?``, and transform them to Python syntax. The
-  interface for these is more complex; see below.
+String based transformations are functions which accept a list of strings:
+each string is a single line of the input cell, including its line ending.
+The transformation function should return output in the same structure.
+
+These transformations are in two groups, accessible as attributes of
+the :class:`~IPython.core.interactiveshell.InteractiveShell` instance.
+Each group is a list of transformation functions.
+
+* ``input_transformers_cleanup`` run first on input, to do things like stripping
+  prompts and leading indents from copied code. It may not be possible at this
+  stage to parse the input as valid Python code.
+* Then IPython runs its own transformations to handle its special syntax, like
+  ``%magics`` and ``!system`` commands. This part does not expose extension
+  points.
+* ``input_transformers_post`` run as the last step, to do things like converting
+  float literals into decimal objects. These may attempt to parse the input as
+  Python code.  
 
 These transformers may raise :exc:`SyntaxError` if the input code is invalid, but
 in most cases it is clearer to pass unrecognised code through unmodified and let
 Python's own parser decide whether it is valid.
 
-.. versionchanged:: 2.0
-
-   Added the option to raise :exc:`SyntaxError`.
-
-Line based transformations
---------------------------
-
 For example, imagine we want to obfuscate our code by reversing each line, so
 we'd write ``)5(f =+ a`` instead of ``a += f(5)``. Here's how we could swap it
 back the right way before IPython tries to run it::
@@ -66,86 +60,7 b' back the right way before IPython tries to run it::'
 To start using this::
 
     ip = get_ipython()
-    ip.input_transformer_manager.line_transforms.append(reverse_line_chars)
-
-Token based transformations
----------------------------
-
-These recognise special syntax like ``%magics`` and ``help?``, and transform it
-into valid Python code. Using tokens makes it easy to avoid transforming similar
-patterns inside comments or strings.
-
-The API for a token-based transformation looks like this::
-
-.. class:: MyTokenTransformer
-   
-   .. classmethod:: find(tokens_by_line)
-
-      Takes a list of lists of :class:`tokenize.TokenInfo` objects. Each sublist
-      is the tokens from one Python line, which may span several physical lines,
-      because of line continuations, multiline strings or expressions. If it
-      finds a pattern to transform, it returns an instance of the class.
-      Otherwise, it returns None.
-
-   .. attribute:: start_lineno
-                  start_col
-                  priority
-
-      These attributes are used to select which transformation to run first.
-      ``start_lineno`` is 0-indexed (whereas the locations on
-      :class:`~tokenize.TokenInfo` use 1-indexed line numbers). If there are
-      multiple matches in the same location, the one with the smaller
-      ``priority`` number is used.
-
-   .. method:: transform(lines)
-
-      This should transform the individual recognised pattern that was
-      previously found. As with line-based transforms, it takes a list of
-      lines as strings, and returns a similar list.
-
-Because each transformation may affect the parsing of the code after it,
-``TransformerManager`` takes a careful approach. It calls ``find()`` on all
-available transformers. If any find a match, the transformation which matched
-closest to the start is run. Then it tokenises the transformed code again,
-and starts the process again. This continues until none of the transformers
-return a match. So it's important that the transformation removes the pattern
-which ``find()`` recognises, otherwise it will enter an infinite loop.
-
-For example, here's a transformer which will recognise ``¬`` as a prefix for a
-new kind of special command::
-
-    import tokenize
-    from IPython.core.inputtransformer2 import TokenTransformBase
-
-    class MySpecialCommand(TokenTransformBase):
-        @classmethod
-        def find(cls, tokens_by_line):
-            """Find the first escaped command (¬foo) in the cell.
-            """
-            for line in tokens_by_line:
-                ix = 0
-                # Find the first token that's not INDENT/DEDENT
-                while line[ix].type in {tokenize.INDENT, tokenize.DEDENT}:
-                    ix += 1
-                if line[ix].string == '¬':
-                    return cls(line[ix].start)
-    
-        def transform(self, lines):   
-            indent  = lines[self.start_line][:self.start_col]
-            content = lines[self.start_line][self.start_col+1:]
-    
-            lines_before = lines[:self.start_line]
-            call = "specialcommand(%r)" % content
-            new_line = indent + call + '\n'
-            lines_after = lines[self.start_line + 1:]
-    
-            return lines_before + [new_line] + lines_after
-
-And here's how you'd use it::
-
-    ip = get_ipython()
-    ip.input_transformer_manager.token_transformers.append(MySpecialCommand)
-
+    ip.input_transformers_cleanup.append(reverse_line_chars)
 
 AST transformations
 ===================