Add the ability to autocorrect a user's command

news/4193.feature

@@ -0,0 +1,5 @@
+Add ability to suggest commands and also to autocorrect mistyped commands.
+
+A new "autocorrect" configuration key is now supported. It can be set to a
+numerical value. If it is set, pip will wait for the specified seconds and then
+autocorrect your command and continue, if a replacement is similar enough.

pip/__init__.py

@@ -3,11 +3,11 @@
 
 import locale
 import logging
-import os
 import optparse
-import warnings
-
+import os
 import sys
+import time
+import warnings
 
 # 2016-06-17 [email protected]: urllib3 1.14 added optional support for socks,
 # but if invoked (i.e. imported), it will issue a warning to stderr if socks
@@ -40,12 +40,12 @@
         else:
             securetransport.inject_into_urllib3()
 
-from pip.exceptions import CommandError, PipError
+from pip.exceptions import CommandError, ConfigurationError, PipError
 from pip.utils import get_installed_distributions, get_prog
 from pip.utils import deprecation
 from pip.vcs import git, mercurial, subversion, bazaar  # noqa
 from pip.baseparser import ConfigOptionParser, UpdatingDefaultsHelpFormatter
-from pip.commands import get_summaries, get_similar_commands
+from pip.commands import get_summaries, get_closest_command
 from pip.commands import commands_dict
 from pip._vendor.requests.packages.urllib3.exceptions import (
     InsecureRequestWarning,
@@ -199,18 +199,57 @@ def parseopts(args):
     # the subcommand name
     cmd_name = args_else[0]
 
+    # all the args without the subcommand
+    cmd_args = args[:]
+    cmd_args.remove(cmd_name)
+
+    # Perform diagnosis of what the user typed
     if cmd_name not in commands_dict:
-        guess = get_similar_commands(cmd_name)
+        # these were manually chosen
+        suggest_cutoff = 0.6
+        autocorrect_cutoff = 0.8
 
-        msg = ['unknown command "%s"' % cmd_name]
-        if guess:
-            msg.append('maybe you meant "%s"' % guess)
+        # Determine if user wants autocorrect.
+        try:
+            autocorrect_delay = parser.config.get_value("global.autocorrect")
+        except ConfigurationError:
+            autocorrect_delay = None
 
-        raise CommandError(' - '.join(msg))
+        try:
+            autocorrect_delay = float(autocorrect_delay)
+        except ValueError:
+            raise ConfigurationError(
+                "autocorrect needs to be a numerical value"
+            )
 
-    # all the args without the subcommand
-    cmd_args = args[:]
-    cmd_args.remove(cmd_name)
+        guess, score = get_closest_command(cmd_name)
+
+        # Decide what message has to be shown to user
+        msg = 'pip does not have a command "%s"' % cmd_name
+        if score > suggest_cutoff:
+            msg += ' - did you mean "%s"?' % guess
+
+        allowed_to_autocorrect = (
+            autocorrect_delay is not None and score > autocorrect_cutoff
+        )
+        if not allowed_to_autocorrect:
+            raise CommandError(msg)
+
+        # Show a message to the user that pip is assuming.
+        msg = (
+            'You called a pip command named "%s" which does not exist.\n'
+            'Assuming you meant "%s", pip will continue in %.1f seconds...'
+        )
+
+        logger.warning(msg, cmd_name, guess, autocorrect_delay)
+        try:
+            time.sleep(autocorrect_delay)
+        except KeyboardInterrupt:
+            logger.critical('Operation cancelled by user')
+            sys.exit(pip.status_codes.ERROR)
+
+        # Assume and proceed.
+        cmd_name = guess
 
     return cmd_name, cmd_args
 

pip/commands/__init__.py

@@ -3,6 +3,8 @@
 """
 from __future__ import absolute_import
 
+from difflib import SequenceMatcher
+
 from pip.commands.completion import CompletionCommand
 from pip.commands.configuration import ConfigurationCommand
 from pip.commands.download import DownloadCommand
@@ -48,20 +50,6 @@ def get_summaries(ordered=True):
         yield (name, command_class.summary)
 
 
-def get_similar_commands(name):
-    """Command name auto-correct."""
-    from difflib import get_close_matches
-
-    name = name.lower()
-
-    close_commands = get_close_matches(name, commands_dict.keys())
-
-    if close_commands:
-        return close_commands[0]
-    else:
-        return False
-
-
 def _sort_commands(cmddict, order):
     def keyfn(key):
         try:
@@ -71,3 +59,50 @@ def keyfn(key):
             return 0xff
 
     return sorted(cmddict.items(), key=keyfn)
+
+
+def _get_closest_match(word, possibilities):
+    """Get the closest match of word in possibilities.
+
+    Returns a tuple of (name, similarity) where possibility has the
+    highest similarity, where 0 <= similarity <= 1.
+    Returns (None, 0) as a fallback, if no possibility matches.
+
+    If more than one possibilities have the highest similarity, the first
+    matched is returned.
+    """
+    guess, best_score = None, 0
+
+    matcher = SequenceMatcher()
+    matcher.set_seq2(word)
+
+    for trial in possibilities:
+        matcher.set_seq1(trial)
+
+        # These are upper limits.
+        if matcher.real_quick_ratio() < best_score:
+            continue
+        if matcher.quick_ratio() < best_score:
+            continue
+
+        # Select the first best match
+        score = matcher.ratio()
+        if score > best_score:
+            guess = trial
+            best_score = score
+
+    return guess, best_score
+
+
+def get_closest_command(name):
+    """Command name auto-correction
+
+    If there are any commands with a similarity greater than cutoff, returns
+    (command_name, similarity) of the command_name with highest similarity.
+
+    If there is no such command, returns (None, 0).
+
+    If more than one commands have the highest similarity, the alphabetically
+    first is returned.
+    """
+    return _get_closest_match(name.lower(), sorted(commands_dict.keys()))

pip/commands/help.py

@@ -13,7 +13,7 @@ class HelpCommand(Command):
     ignore_require_venv = True
 
     def run(self, options, args):
-        from pip.commands import commands_dict, get_similar_commands
+        from pip.commands import commands_dict, get_closest_command
 
         try:
             # 'pip help' with no args is handled by pip.__init__.parseopt()
@@ -22,10 +22,11 @@ def run(self, options, args):
             return SUCCESS
 
         if cmd_name not in commands_dict:
-            guess = get_similar_commands(cmd_name)
+            guess, score = get_closest_command(cmd_name)
+            suggest_cut_off = 0.6
 
             msg = ['unknown command "%s"' % cmd_name]
-            if guess:
+            if guess and score > suggest_cut_off:
                 msg.append('maybe you meant "%s"' % guess)
 
             raise CommandError(' - '.join(msg))

pip/configuration.py

@@ -24,7 +24,6 @@
 )
 from pip.utils import ensure_dir, enum
 
-
 logger = logging.getLogger(__name__)