# minirst.py - minimal reStructuredText parser

#

# Copyright 2009, 2010 Matt Mackall <mpm@selenic.com> and others

#

# This software may be used and distributed according to the terms of the

# GNU General Public License version 2 or any later version.

"""simplified reStructuredText parser.

This parser knows just enough about reStructuredText to parse the

Mercurial docstrings.

It cheats in a major way: nested blocks are not really nested. They

are just indented blocks that look like they are nested. This relies

on the user to keep the right indentation for the blocks.

Remember to update http://mercurial.selenic.com/wiki/HelpStyleGuide

when adding support for new constructs.

"""

import re

import util, encoding

from i18n import _

import cgi

def section(s):

    return "%s\n%s\n\n" % (s, "\"" * encoding.colwidth(s))

def subsection(s):

    return "%s\n%s\n\n" % (s, '=' * encoding.colwidth(s))

def subsubsection(s):

    return "%s\n%s\n\n" % (s, "-" * encoding.colwidth(s))

def subsubsubsection(s):

    return "%s\n%s\n\n" % (s, "." * encoding.colwidth(s))

def replace(text, substs):

    '''

    Apply a list of (find, replace) pairs to a text.

    >>> replace("foo bar", [('f', 'F'), ('b', 'B')])

    'Foo Bar'

    >>> encoding.encoding = 'latin1'

    >>> replace('\\x81\\\\', [('\\\\', '/')])

    '\\x81/'

    >>> encoding.encoding = 'shiftjis'

    >>> replace('\\x81\\\\', [('\\\\', '/')])

    '\\x81\\\\'

    '''

    # some character encodings (cp932 for Japanese, at least) use

    # ASCII characters other than control/alphabet/digit as a part of

    # multi-bytes characters, so direct replacing with such characters

    # on strings in local encoding causes invalid byte sequences.

    utext = text.decode(encoding.encoding)

    for f, t in substs:

        utext = utext.replace(f.decode("ascii"), t.decode("ascii"))

    return utext.encode(encoding.encoding)

_blockre = re.compile(r"\n(?:\s*\n)+")

def findblocks(text):

    """Find continuous blocks of lines in text.

    Returns a list of dictionaries representing the blocks. Each block

    has an 'indent' field and a 'lines' field.

    """

    blocks = []

    for b in _blockre.split(text.lstrip('\n').rstrip()):

        lines = b.splitlines()

        if lines:

            indent = min((len(l) - len(l.lstrip())) for l in lines)

            lines = [l[indent:] for l in lines]

            blocks.append({'indent': indent, 'lines': lines})

    return blocks

def findliteralblocks(blocks):

    """Finds literal blocks and adds a 'type' field to the blocks.

    Literal blocks are given the type 'literal', all other blocks are

    given type the 'paragraph'.

    """

    i = 0

    while i < len(blocks):

        # Searching for a block that looks like this:

        #

        # +------------------------------+

        # | paragraph                    |

        # | (ends with "::")             |

        # +------------------------------+

        #    +---------------------------+

        #    | indented literal block    |

        #    +---------------------------+

        blocks[i]['type'] = 'paragraph'

        if blocks[i]['lines'][-1].endswith('::') and i + 1 < len(blocks):

            indent = blocks[i]['indent']

            adjustment = blocks[i + 1]['indent'] - indent

            if blocks[i]['lines'] == ['::']:

                # Expanded form: remove block

                del blocks[i]

                i -= 1

            elif blocks[i]['lines'][-1].endswith(' ::'):

                # Partially minimized form: remove space and both

                # colons.

                blocks[i]['lines'][-1] = blocks[i]['lines'][-1][:-3]

            elif len(blocks[i]['lines']) == 1 and \

                 blocks[i]['lines'][0].lstrip(' ').startswith('.. ') and \

                 blocks[i]['lines'][0].find(' ', 3) == -1:

                # directive on its own line, not a literal block

                i += 1

                continue

            else:

                # Fully minimized form: remove just one colon.

                blocks[i]['lines'][-1] = blocks[i]['lines'][-1][:-1]

            # List items are formatted with a hanging indent. We must

            # correct for this here while we still have the original

            # information on the indentation of the subsequent literal

            # blocks available.

            m = _bulletre.match(blocks[i]['lines'][0])

            if m:

                indent += m.end()

                adjustment -= m.end()

            # Mark the following indented blocks.

            while i + 1 < len(blocks) and blocks[i + 1]['indent'] > indent:

                blocks[i + 1]['type'] = 'literal'

                blocks[i + 1]['indent'] -= adjustment

                i += 1

        i += 1

    return blocks

_bulletre = re.compile(r'(-|[0-9A-Za-z]+\.|\(?[0-9A-Za-z]+\)|\|) ')

_optionre = re.compile(r'^(-([a-zA-Z0-9]), )?(--[a-z0-9-]+)'

                       r'((.*)  +)(.*)$')

_fieldre = re.compile(r':(?![: ])([^:]*)(?<! ):[ ]+(.*)')

_definitionre = re.compile(r'[^ ]')

_tablere = re.compile(r'(=+\s+)*=+')

def splitparagraphs(blocks):

    """Split paragraphs into lists."""

    # Tuples with (list type, item regexp, single line items?). Order

    # matters: definition lists has the least specific regexp and must

    # come last.

    listtypes = [('bullet', _bulletre, True),

                 ('option', _optionre, True),

                 ('field', _fieldre, True),

                 ('definition', _definitionre, False)]

    def match(lines, i, itemre, singleline):

        """Does itemre match an item at line i?

        A list item can be followed by an indented line or another list

        item (but only if singleline is True).

        """

        line1 = lines[i]

        line2 = i + 1 < len(lines) and lines[i + 1] or ''

        if not itemre.match(line1):

            return False

        if singleline:

            return line2 == '' or line2[0] == ' ' or itemre.match(line2)

        else:

            return line2.startswith(' ')

    i = 0

    while i < len(blocks):

        if blocks[i]['type'] == 'paragraph':

            lines = blocks[i]['lines']

            for type, itemre, singleline in listtypes:

                if match(lines, 0, itemre, singleline):

                    items = []

                    for j, line in enumerate(lines):

                        if match(lines, j, itemre, singleline):

                            items.append({'type': type, 'lines': [],

                                          'indent': blocks[i]['indent']})

                        items[-1]['lines'].append(line)

                    blocks[i:i + 1] = items

                    break

        i += 1

    return blocks

_fieldwidth = 14

def updatefieldlists(blocks):

    """Find key for field lists."""

    i = 0

    while i < len(blocks):

        if blocks[i]['type'] != 'field':

            i += 1

            continue

        j = i

        while j < len(blocks) and blocks[j]['type'] == 'field':

            m = _fieldre.match(blocks[j]['lines'][0])

            key, rest = m.groups()

            blocks[j]['lines'][0] = rest

            blocks[j]['key'] = key

            j += 1

        i = j + 1

    return blocks

def updateoptionlists(blocks):

    i = 0

    while i < len(blocks):

        if blocks[i]['type'] != 'option':

            i += 1

            continue

        optstrwidth = 0

        j = i

        while j < len(blocks) and blocks[j]['type'] == 'option':

            m = _optionre.match(blocks[j]['lines'][0])

            shortoption = m.group(2)

            group3 = m.group(3)

            longoption = group3[2:].strip()

            desc = m.group(6).strip()

            longoptionarg = m.group(5).strip()

            blocks[j]['lines'][0] = desc

            noshortop = ''

            if not shortoption:

                noshortop = '   '

            opt = "%s%s" %   (shortoption and "-%s " % shortoption or '',

                            ("%s--%s %s") % (noshortop, longoption,

                                             longoptionarg))

            opt = opt.rstrip()

            blocks[j]['optstr'] = opt

            optstrwidth = max(optstrwidth, encoding.colwidth(opt))

            j += 1

        for block in blocks[i:j]:

            block['optstrwidth'] = optstrwidth

        i = j + 1

    return blocks

def prunecontainers(blocks, keep):

    """Prune unwanted containers.

    The blocks must have a 'type' field, i.e., they should have been

    run through findliteralblocks first.

    """

    pruned = []

    i = 0

    while i + 1 < len(blocks):

        # Searching for a block that looks like this:

        #

        # +-------+---------------------------+

        # | ".. container ::" type            |

        # +---+                               |

        #     | blocks                        |

        #     +-------------------------------+

        if (blocks[i]['type'] == 'paragraph' and

            blocks[i]['lines'][0].startswith('.. container::')):

            indent = blocks[i]['indent']

            adjustment = blocks[i + 1]['indent'] - indent

            containertype = blocks[i]['lines'][0][15:]

            prune = True

            for c in keep:

                if c in containertype.split('.'):

                    prune = False

            if prune:

                pruned.append(containertype)

            # Always delete "..container:: type" block

            del blocks[i]

            j = i

            i -= 1

            while j < len(blocks) and blocks[j]['indent'] > indent:

                if prune:

                    del blocks[j]

                else:

                    blocks[j]['indent'] -= adjustment

                    j += 1

        i += 1

    return blocks, pruned

_sectionre = re.compile(r"""^([-=`:.'"~^_*+#])\1+$""")

def findtables(blocks):

    '''Find simple tables

       Only simple one-line table elements are supported

    '''

    for block in blocks:

        # Searching for a block that looks like this:

        #

        # === ==== ===

        #  A    B   C

        # === ==== ===  <- optional

        #  1    2   3

        #  x    y   z

        # === ==== ===

        if (block['type'] == 'paragraph' and

            len(block['lines']) > 2 and

            _tablere.match(block['lines'][0]) and

            block['lines'][0] == block['lines'][-1]):

            block['type'] = 'table'

            block['header'] = False

            div = block['lines'][0]

            # column markers are ASCII so we can calculate column

            # position in bytes

            columns = [x for x in xrange(len(div))

                       if div[x] == '=' and (x == 0 or div[x - 1] == ' ')]

            rows = []

            for l in block['lines'][1:-1]:

                if l == div:

                    block['header'] = True

                    continue

                row = []

                # we measure columns not in bytes or characters but in

                # colwidth which makes things tricky

                pos = columns[0] # leading whitespace is bytes

                for n, start in enumerate(columns):

                    if n + 1 < len(columns):

                        width = columns[n + 1] - start

                        v = encoding.getcols(l, pos, width) # gather columns

                        pos += len(v) # calculate byte position of end

                        row.append(v.strip())

                    else:

                        row.append(l[pos:].strip())

                rows.append(row)

            block['table'] = rows

    return blocks

def findsections(blocks):

    """Finds sections.

    The blocks must have a 'type' field, i.e., they should have been

    run through findliteralblocks first.

    """

    for block in blocks:

        # Searching for a block that looks like this:

        #

        # +------------------------------+

        # | Section title                |

        # | -------------                |

        # +------------------------------+

        if (block['type'] == 'paragraph' and

            len(block['lines']) == 2 and

            encoding.colwidth(block['lines'][0]) == len(block['lines'][1]) and

            _sectionre.match(block['lines'][1])):

            block['underline'] = block['lines'][1][0]

            block['type'] = 'section'

            del block['lines'][1]

    return blocks

def inlineliterals(blocks):

    substs = [('``', '"')]

    for b in blocks:

        if b['type'] in ('paragraph', 'section'):

            b['lines'] = [replace(l, substs) for l in b['lines']]

    return blocks

def hgrole(blocks):

    substs = [(':hg:`', '"hg '), ('`', '"')]

    for b in blocks:

        if b['type'] in ('paragraph', 'section'):

            # Turn :hg:`command` into "hg command". This also works

            # when there is a line break in the command and relies on

            # the fact that we have no stray back-quotes in the input

            # (run the blocks through inlineliterals first).

            b['lines'] = [replace(l, substs) for l in b['lines']]

    return blocks

def addmargins(blocks):

    """Adds empty blocks for vertical spacing.

    This groups bullets, options, and definitions together with no vertical

    space between them, and adds an empty block between all other blocks.