path: root/cmd2/parsing.py

#
# -*- coding: utf-8 -*-
"""Statement parsing classes for cmd2"""

import re
import shlex
from typing import (
    Dict,
    Iterable,
    List,
    Optional,
    Tuple,
    Union,
)

import attr

from . import (
    constants,
    utils,
)
from .exceptions import (
    Cmd2ShlexError,
)


def shlex_split(str_to_split: str) -> List[str]:
    """
    A wrapper around shlex.split() that uses cmd2's preferred arguments.
    This allows other classes to easily call split() the same way StatementParser does.

    :param str_to_split: the string being split
    :return: A list of tokens
    """
    return shlex.split(str_to_split, comments=False, posix=False)


@attr.s(frozen=True)
class MacroArg:
    """
    Information used to replace or unescape arguments in a macro value when the macro is resolved
    Normal argument syntax:    {5}
    Escaped argument syntax:  {{5}}
    """

    # The starting index of this argument in the macro value
    start_index = attr.ib(validator=attr.validators.instance_of(int))

    # The number string that appears between the braces
    # This is a string instead of an int because we support unicode digits and must be able
    # to reproduce this string later
    number_str = attr.ib(validator=attr.validators.instance_of(str))

    # Tells if this argument is escaped and therefore needs to be unescaped
    is_escaped = attr.ib(validator=attr.validators.instance_of(bool))

    # Pattern used to find normal argument
    # Digits surrounded by exactly 1 brace on a side and 1 or more braces on the opposite side
    # Match strings like: {5}, {{{{{4}, {2}}}}}
    macro_normal_arg_pattern = re.compile(r'(?<!{){\d+}|{\d+}(?!})')

    # Pattern used to find escaped arguments
    # Digits surrounded by 2 or more braces on both sides
    # Match strings like: {{5}}, {{{{{4}}, {{2}}}}}
    macro_escaped_arg_pattern = re.compile(r'{{2}\d+}{2}')

    # Finds a string of digits
    digit_pattern = re.compile(r'\d+')


@attr.s(frozen=True)
class Macro:
    """Defines a cmd2 macro"""

    # Name of the macro
    name = attr.ib(validator=attr.validators.instance_of(str))

    # The string the macro resolves to
    value = attr.ib(validator=attr.validators.instance_of(str))

    # The minimum number of args the user has to pass to this macro
    minimum_arg_count = attr.ib(validator=attr.validators.instance_of(int))

    # Used to fill in argument placeholders in the macro
    arg_list = attr.ib(default=attr.Factory(list), validator=attr.validators.instance_of(list))


@attr.s(frozen=True)
class Statement(str):
    """String subclass with additional attributes to store the results of parsing.

    The ``cmd`` module in the standard library passes commands around as a
    string. To retain backwards compatibility, ``cmd2`` does the same. However,
    we need a place to capture the additional output of the command parsing, so
    we add our own attributes to this subclass.

    Instances of this class should not be created by anything other than the
    :meth:`cmd2.parsing.StatementParser.parse` method, nor should any of the
    attributes be modified once the object is created.

    The string portion of the class contains the arguments, but not the
    command, nor the output redirection clauses.

    Tips:

    1. `argparse <https://docs.python.org/3/library/argparse.html>`_ is your
       friend for anything complex. ``cmd2`` has the decorator
       (:func:`~cmd2.decorators.with_argparser`) which you can
       use to make your command method receive a namespace of parsed arguments,
       whether positional or denoted with switches.

    2. For commands with simple positional arguments, use
       :attr:`~cmd2.Statement.args` or :attr:`~cmd2.Statement.arg_list`

    3. If you don't want to have to worry about quoted arguments, see
       :attr:`argv` for a trick which strips quotes off for you.
    """

    # the arguments, but not the command, nor the output redirection clauses.
    args = attr.ib(default='', validator=attr.validators.instance_of(str))

    # string containing exactly what we input by the user
    raw = attr.ib(default='', validator=attr.validators.instance_of(str))

    # the command, i.e. the first whitespace delimited word
    command = attr.ib(default='', validator=attr.validators.instance_of(str))

    # list of arguments to the command, not including any output redirection or terminators; quoted args remain quoted
    arg_list = attr.ib(default=attr.Factory(list), validator=attr.validators.instance_of(list))

    # if the command is a multiline command, the name of the command, otherwise empty
    multiline_command = attr.ib(default='', validator=attr.validators.instance_of(str))

    # the character which terminated the multiline command, if there was one
    terminator = attr.ib(default='', validator=attr.validators.instance_of(str))

    # characters appearing after the terminator but before output redirection, if any
    suffix = attr.ib(default='', validator=attr.validators.instance_of(str))

    # if output was piped to a shell command, the shell command as a string
    pipe_to = attr.ib(default='', validator=attr.validators.instance_of(str))

    # if output was redirected, the redirection token, i.e. '>>'
    output = attr.ib(default='', validator=attr.validators.instance_of(str))

    # if output was redirected, the destination file token (quotes preserved)
    output_to = attr.ib(default='', validator=attr.validators.instance_of(str))

    def __new__(cls, value: object, *pos_args, **kw_args):
        """Create a new instance of Statement.

        We must override __new__ because we are subclassing `str` which is
        immutable and takes a different number of arguments as Statement.

        NOTE:  attrs takes care of initializing other members in the __init__ it
        generates.
        """
        stmt = super().__new__(cls, value)
        return stmt

    @property
    def command_and_args(self) -> str:
        """Combine command and args with a space separating them.

        Quoted arguments remain quoted. Output redirection and piping are
        excluded, as are any command terminators.
        """
        if self.command and self.args:
            rtn = '{} {}'.format(self.command, self.args)
        elif self.command:
            # there were no arguments to the command
            rtn = self.command
        else:
            rtn = ''
        return rtn

    @property
    def post_command(self) -> str:
        """A string containing any ending terminator, suffix, and redirection chars"""
        rtn = ''
        if self.terminator:
            rtn += self.terminator

        if self.suffix:
            rtn += ' ' + self.suffix

        if self.pipe_to:
            rtn += ' | ' + self.pipe_to

        if self.output:
            rtn += ' ' + self.output
            if self.output_to:
                rtn += ' ' + self.output_to

        return rtn

    @property
    def expanded_command_line(self) -> str:
        """Concatenate :meth:`~cmd2.Statement.command_and_args`
        and :meth:`~cmd2.Statement.post_command`"""
        return self.command_and_args + self.post_command

    @property
    def argv(self) -> List[str]:
        """a list of arguments a-la ``sys.argv``.

        The first element of the list is the command after shortcut and macro
        expansion. Subsequent elements of the list contain any additional
        arguments, with quotes removed, just like bash would. This is very
        useful if you are going to use ``argparse.parse_args()``.

        If you want to strip quotes from the input, you can use ``argv[1:]``.
        """
        if self.command:
            rtn = [utils.strip_quotes(self.command)]
            for cur_token in self.arg_list:
                rtn.append(utils.strip_quotes(cur_token))
        else:
            rtn = []

        return rtn


class StatementParser:
    """Parse user input as a string into discrete command components."""

    def __init__(
        self,
        terminators: Optional[Iterable[str]] = None,
        multiline_commands: Optional[Iterable[str]] = None,
        aliases: Optional[Dict[str, str]] = None,
        shortcuts: Optional[Dict[str, str]] = None,
    ) -> None:
        """Initialize an instance of StatementParser.

        The following will get converted to an immutable tuple before storing internally:
        terminators, multiline commands, and shortcuts.

        :param terminators: iterable containing strings which should terminate commands
        :param multiline_commands: iterable containing the names of commands that accept multiline input
        :param aliases: dictionary containing aliases
        :param shortcuts: dictionary containing shortcuts
        """
        if terminators is None:
            self.terminators = (constants.MULTILINE_TERMINATOR,)
        else:
            self.terminators = tuple(terminators)
        if multiline_commands is None:
            self.multiline_commands = tuple()
        else:
            self.multiline_commands = tuple(multiline_commands)
        if aliases is None:
            self.aliases = dict()
        else:
            self.aliases = aliases

        if shortcuts is None:
            shortcuts = constants.DEFAULT_SHORTCUTS

        # Sort the shortcuts in descending order by name length because the longest match
        # should take precedence. (e.g., @@file should match '@@' and not '@'.
        self.shortcuts = tuple(sorted(shortcuts.items(), key=lambda x: len(x[0]), reverse=True))

        # commands have to be a word, so make a regular expression
        # that matches the first word in the line. This regex has three
        # parts:
        #     - the '\A\s*' matches the beginning of the string (even
        #       if contains multiple lines) and gobbles up any leading
        #       whitespace
        #     - the first parenthesis enclosed group matches one
        #       or more non-whitespace characters with a non-greedy match
        #       (that's what the '+?' part does). The non-greedy match
        #       ensures that this first group doesn't include anything
        #       matched by the second group
        #     - the second parenthesis group must be dynamically created
        #       because it needs to match either whitespace, something in
        #       REDIRECTION_CHARS, one of the terminators, or the end of
        #       the string (\Z matches the end of the string even if it
        #       contains multiple lines)
        #
        invalid_command_chars = []
        invalid_command_chars.extend(constants.QUOTES)
        invalid_command_chars.extend(constants.REDIRECTION_CHARS)
        invalid_command_chars.extend(self.terminators)
        # escape each item so it will for sure get treated as a literal
        second_group_items = [re.escape(x) for x in invalid_command_chars]
        # add the whitespace and end of string, not escaped because they
        # are not literals
        second_group_items.extend([r'\s', r'\Z'])
        # join them up with a pipe
        second_group = '|'.join(second_group_items)
        # build the regular expression
        expr = r'\A\s*(\S*?)({})'.format(second_group)
        self._command_pattern = re.compile(expr)

    def is_valid_command(self, word: str, *, is_subcommand: bool = False) -> Tuple[bool, str]:
        """Determine whether a word is a valid name for a command.

        Commands cannot include redirection characters, whitespace,
        or termination characters. They also cannot start with a
        shortcut.

        :param word: the word to check as a command
        :param is_subcommand: Flag whether this command name is a subcommand name
        :return: a tuple of a boolean and an error string

        If word is not a valid command, return ``False`` and an error string
        suitable for inclusion in an error message of your choice::

            checkit = '>'
            valid, errmsg = statement_parser.is_valid_command(checkit)
            if not valid:
                errmsg = "alias: {}".format(errmsg)
        """
        valid = False

        if not isinstance(word, str):
            return False, 'must be a string. Received {} instead'.format(str(type(word)))

        if not word:
            return False, 'cannot be an empty string'

        if word.startswith(constants.COMMENT_CHAR):
            return False, 'cannot start with the comment character'

        if not is_subcommand:
            for (shortcut, _) in self.shortcuts:
                if word.startswith(shortcut):
                    # Build an error string with all shortcuts listed
                    errmsg = 'cannot start with a shortcut: '
                    errmsg += ', '.join(shortcut for (shortcut, _) in self.shortcuts)
                    return False, errmsg

        errmsg = 'cannot contain: whitespace, quotes, '
        errchars = []
        errchars.extend(constants.REDIRECTION_CHARS)
        errchars.extend(self.terminators)
        errmsg += ', '.join([shlex.quote(x) for x in errchars])

        match = self._command_pattern.search(word)
        if match:
            if word == match.group(1):
                valid = True
                errmsg = ''
        return valid, errmsg

    def tokenize(self, line: str) -> List[str]:
        """
        Lex a string into a list of tokens. Shortcuts and aliases are expanded and
        comments are removed.

        :param line: the command line being lexed
        :return: A list of tokens
        :raises: Cmd2ShlexError if a shlex error occurs (e.g. No closing quotation)
        """

        # expand shortcuts and aliases
        line = self._expand(line)

        # check if this line is a comment
        if line.lstrip().startswith(constants.COMMENT_CHAR):
            return []

        # split on whitespace
        try:
            tokens = shlex_split(line)
        except ValueError as ex:
            raise Cmd2ShlexError(ex)

        # custom lexing
        tokens = self.split_on_punctuation(tokens)
        return tokens

    def parse(self, line: str) -> Statement:
        """
        Tokenize the input and parse it into a :class:`~cmd2.Statement` object,
        stripping comments, expanding aliases and shortcuts, and extracting output
        redirection directives.

        :param line: the command line being parsed
        :return: a new :class:`~cmd2.Statement` object
        :raises: Cmd2ShlexError if a shlex error occurs (e.g. No closing quotation)
        """

        # handle the special case/hardcoded terminator of a blank line
        # we have to do this before we tokenize because tokenizing
        # destroys all unquoted whitespace in the input
        terminator = ''
        if line[-1:] == constants.LINE_FEED:
            terminator = constants.LINE_FEED

        command = ''
        args = ''
        arg_list = []

        # lex the input into a list of tokens
        tokens = self.tokenize(line)

        # of the valid terminators, find the first one to occur in the input
        terminator_pos = len(tokens) + 1
        for pos, cur_token in enumerate(tokens):
            for test_terminator in self.terminators:
                if cur_token.startswith(test_terminator):
                    terminator_pos = pos
                    terminator = test_terminator
                    # break the inner loop, and we want to break the
                    # outer loop too
                    break
            else:
                # this else clause is only run if the inner loop
                # didn't execute a break. If it didn't, then
                # continue to the next iteration of the outer loop
                continue
            # inner loop was broken, break the outer
            break

        if terminator:
            if terminator == constants.LINE_FEED:
                terminator_pos = len(tokens) + 1

            # everything before the first terminator is the command and the args
            (command, args) = self._command_and_args(tokens[:terminator_pos])
            arg_list = tokens[1:terminator_pos]
            # we will set the suffix later
            # remove all the tokens before and including the terminator
            tokens = tokens[terminator_pos + 1 :]
        else:
            (testcommand, testargs) = self._command_and_args(tokens)
            if testcommand in self.multiline_commands:
                # no terminator on this line but we have a multiline command
                # everything else on the line is part of the args
                # because redirectors can only be after a terminator
                command = testcommand
                args = testargs
                arg_list = tokens[1:]
                tokens = []

        pipe_to = ''
        output = ''
        output_to = ''

        # Find which redirector character appears first in the command
        try:
            pipe_index = tokens.index(constants.REDIRECTION_PIPE)
        except ValueError:
            pipe_index = len(tokens)

        try:
            redir_index = tokens.index(constants.REDIRECTION_OUTPUT)
        except ValueError:
            redir_index = len(tokens)

        try:
            append_index = tokens.index(constants.REDIRECTION_APPEND)
        except ValueError:
            append_index = len(tokens)

        # Check if output should be piped to a shell command
        if pipe_index < redir_index and pipe_index < append_index:

            # Get the tokens for the pipe command and expand ~ where needed
            pipe_to_tokens = tokens[pipe_index + 1 :]
            utils.expand_user_in_tokens(pipe_to_tokens)

            # Build the pipe command line string
            pipe_to = ' '.join(pipe_to_tokens)

            # remove all the tokens after the pipe
            tokens = tokens[:pipe_index]

        # Check for output redirect/append
        elif redir_index != append_index:
            if redir_index < append_index:
                output = constants.REDIRECTION_OUTPUT
                output_index = redir_index
            else:
                output = constants.REDIRECTION_APPEND
                output_index = append_index

            # Check if we are redirecting to a file
            if len(tokens) > output_index + 1:
                unquoted_path = utils.strip_quotes(tokens[output_index + 1])
                if unquoted_path:
                    output_to = utils.expand_user(tokens[output_index + 1])

            # remove all the tokens after the output redirect
            tokens = tokens[:output_index]

        if terminator:
            # whatever is left is the suffix
            suffix = ' '.join(tokens)
        else:
            # no terminator, so whatever is left is the command and the args
            suffix = ''
            if not command:
                # command could already have been set, if so, don't set it again
                (command, args) = self._command_and_args(tokens)
                arg_list = tokens[1:]

        # set multiline
        if command in self.multiline_commands:
            multiline_command = command
        else:
            multiline_command = ''

        # build the statement
        statement = Statement(
            args,
            raw=line,
            command=command,
            arg_list=arg_list,
            multiline_command=multiline_command,
            terminator=terminator,
            suffix=suffix,
            pipe_to=pipe_to,
            output=output,
            output_to=output_to,
        )
        return statement

    def parse_command_only(self, rawinput: str) -> Statement:
        """Partially parse input into a :class:`~cmd2.Statement` object.

        The command is identified, and shortcuts and aliases are expanded.
        Multiline commands are identified, but terminators and output
        redirection are not parsed.

        This method is used by tab completion code and therefore must not
        generate an exception if there are unclosed quotes.

        The :class:`~cmd2.Statement` object returned by this method can at most
        contain values in the following attributes:
        :attr:`~cmd2.Statement.args`, :attr:`~cmd2.Statement.raw`,
        :attr:`~cmd2.Statement.command`,
        :attr:`~cmd2.Statement.multiline_command`

        :attr:`~cmd2.Statement.args` will include all output redirection
        clauses and command terminators.

        Different from :meth:`~cmd2.parsing.StatementParser.parse` this method
        does not remove redundant whitespace within args. However, it does
        ensure args has no leading or trailing whitespace.

        :param rawinput: the command line as entered by the user
        :return: a new :class:`~cmd2.Statement` object
        """
        # expand shortcuts and aliases
        line = self._expand(rawinput)

        command = ''
        args = ''
        match = self._command_pattern.search(line)
        if match:
            # we got a match, extract the command
            command = match.group(1)

            # take everything from the end of the first match group to
            # the end of the line as the arguments (stripping leading
            # and trailing spaces)
            args = line[match.end(1) :].strip()
            # if the command is empty that means the input was either empty
            # or something weird like '>'. args should be empty if we couldn't
            # parse a command
            if not command or not args:
                args = ''

        # set multiline
        if command in self.multiline_commands:
            multiline_command = command
        else:
            multiline_command = ''

        # build the statement
        statement = Statement(args, raw=rawinput, command=command, multiline_command=multiline_command)
        return statement

    def get_command_arg_list(
        self, command_name: str, to_parse: Union[Statement, str], preserve_quotes: bool
    ) -> Tuple[Statement, List[str]]:
        """
        Convenience method used by the argument parsing decorators.

        Retrieves just the arguments being passed to their ``do_*`` methods as a list.

        :param command_name: name of the command being run
        :param to_parse: what is being passed to the ``do_*`` method. It can be one of two types:

                             1. An already parsed :class:`~cmd2.Statement`
                             2. An argument string in cases where a ``do_*`` method is
                                explicitly called. Calling ``do_help('alias create')`` would
                                cause ``to_parse`` to be 'alias create'.

                                In this case, the string will be converted to a
                                :class:`~cmd2.Statement` and returned along with
                                the argument list.

        :param preserve_quotes: if ``True``, then quotes will not be stripped from
                                the arguments
        :return: A tuple containing the :class:`~cmd2.Statement` and a list of
                 strings representing the arguments
        """
        # Check if to_parse needs to be converted to a Statement
        if not isinstance(to_parse, Statement):
            to_parse = self.parse(command_name + ' ' + to_parse)

        if preserve_quotes:
            return to_parse, to_parse.arg_list
        else:
            return to_parse, to_parse.argv[1:]

    def _expand(self, line: str) -> str:
        """Expand aliases and shortcuts"""

        # Make a copy of aliases so we can keep track of what aliases have been resolved to avoid an infinite loop
        remaining_aliases = list(self.aliases.keys())
        keep_expanding = bool(remaining_aliases)

        while keep_expanding:
            keep_expanding = False

            # apply our regex to line
            match = self._command_pattern.search(line)
            if match:
                # we got a match, extract the command
                command = match.group(1)

                # Check if this command matches an alias that wasn't already processed
                if command in remaining_aliases:
                    # rebuild line with the expanded alias
                    line = self.aliases[command] + match.group(2) + line[match.end(2) :]
                    remaining_aliases.remove(command)
                    keep_expanding = bool(remaining_aliases)

        # expand shortcuts
        for (shortcut, expansion) in self.shortcuts:
            if line.startswith(shortcut):
                # If the next character after the shortcut isn't a space, then insert one
                shortcut_len = len(shortcut)
                if len(line) == shortcut_len or line[shortcut_len] != ' ':
                    expansion += ' '

                # Expand the shortcut
                line = line.replace(shortcut, expansion, 1)
                break
        return line

    @staticmethod
    def _command_and_args(tokens: List[str]) -> Tuple[str, str]:
        """Given a list of tokens, return a tuple of the command
        and the args as a string.
        """
        command = ''
        args = ''

        if tokens:
            command = tokens[0]

        if len(tokens) > 1:
            args = ' '.join(tokens[1:])

        return command, args

    def split_on_punctuation(self, tokens: List[str]) -> List[str]:
        """Further splits tokens from a command line using punctuation characters.

        Punctuation characters are treated as word breaks when they are in
        unquoted strings. Each run of punctuation characters is treated as a
        single token.

        :param tokens: the tokens as parsed by shlex
        :return: a new list of tokens, further split using punctuation
        """
        punctuation = []
        punctuation.extend(self.terminators)
        punctuation.extend(constants.REDIRECTION_CHARS)

        punctuated_tokens = []

        for cur_initial_token in tokens:

            # Save tokens up to 1 character in length or quoted tokens. No need to parse these.
            if len(cur_initial_token) <= 1 or cur_initial_token[0] in constants.QUOTES:
                punctuated_tokens.append(cur_initial_token)
                continue

            # Iterate over each character in this token
            cur_index = 0
            cur_char = cur_initial_token[cur_index]

            # Keep track of the token we are building
            new_token = ''

            while True:
                if cur_char not in punctuation:

                    # Keep appending to new_token until we hit a punctuation char
                    while cur_char not in punctuation:
                        new_token += cur_char
                        cur_index += 1
                        if cur_index < len(cur_initial_token):
                            cur_char = cur_initial_token[cur_index]
                        else:
                            break

                else:
                    cur_punc = cur_char

                    # Keep appending to new_token until we hit something other than cur_punc
                    while cur_char == cur_punc:
                        new_token += cur_char
                        cur_index += 1
                        if cur_index < len(cur_initial_token):
                            cur_char = cur_initial_token[cur_index]
                        else:
                            break

                # Save the new token
                punctuated_tokens.append(new_token)
                new_token = ''

                # Check if we've viewed all characters
                if cur_index >= len(cur_initial_token):
                    break

        return punctuated_tokens