#!/usr/bin/env python3
#
# wmllint -- check WML for conformance to the most recent dialect
#
# By Eric S. Raymond April 2007.
#
# All conversion logic for lifting WML and maps from older versions of the
# markup to newer ones should live here.  This includes resource path changes
# and renames, also map format conversions.
#
# While the script is at it, it checks for various incorrect and dodgy WML
# constructs, including:
#   * unbalanced tags
#   * strings that need a translation mark and should not have them
#   * strings that have a translation mark and should not
#   * translatable strings containing macro references
#   * filter references by description= not matched by an actual unit
#   * abilities or traits without matching special notes, or vice-versa
#   * consistency between recruit= and recruitment_pattern= instances
#   * double space after punctuation in translatable strings.
#   * unknown races or movement types in units
#
# Takes any number of directories as arguments.  Each directory is converted.
# If no directories are specified, acts on the current directory.
#
# The recommended procedure is this:
# 1. Run it with --dryrun first to see what it will do.
# 2. If the messages look good, run without --dryrun; the old content
#    will be left in backup files with a -bak extension.
# 3. Eyeball the changes with the --diff option.
# 4. Use wmlscope, with a directory list including the Wesnoth mainline WML
#    as first argument, to check that you have no unresolved references.
# 5. Test the conversion.
# 6. Use either --clean to remove the -bak files or --revert to
#    undo the conversion.
#
# This script will barf on 1.2.x maps with custom terrains.  Also, if you
# have a single subdirectory that mixes old-style and new-style
# terrain coding it might get confused.
#
# Standalone terrain mask files *must* have a .mask extension on their name
# or they'll have an incorrect usage=map generated into them.
#
# Note: You can shut wmllint up about custom terrains by having a comment
# on the same line that includes the string "wmllint: ignore".
#
# You can also prevent description insertions with "wmllint: no-icon".
#
# You can force otherwise undeclared characters to be recognized with
# a magic comment containing the string "wmllint: recognize".
# The rest of the line is stripped and treated as the name of a character
# who should be recognized in descriptions.  This will be useful,
# for example, if your scenario follows a continue so there are
# characters present who were not explicitly recalled.  It may
# also be useful if you have wrapped unit-creation or recall markup in macros
# and wmllint cannot recognize it.
#
# Similarly, it is possible to explicitly declare a unit's usage class
# with a magic comment that looks like thisL
#    wmllint: usage of <unit> is <class>
# Note that <unit> must be a string wrapped in ASCII doublequotes.  This
# declaration will be useful if you are declaring units with macros that
# include a substitutable formal in the unit name; there are examples in UtBS.
#
# You can disable stack-based malformation checks with a comment
# containing "wmllint: validate-off" and re-enable with "wmllint: validate-on".
#
# You can prevent file conversions with a comment containing
# "wmllint: noconvert" on the same line as the filename.
#
# You can skip checks on unbalanced WML (e.g. in a macro definition) by
# bracketing it with "wmllint: unbalanced-on" and "wmllint: unbalanced-off".
# Note that this will also disable stack-based validation on the span
# of lines they enclose.

import sys, os, re, getopt, string, copy, difflib, time, codecs
from wesnoth.wmltools3 import *
from wesnoth.wmliterator3 import *

filemoves = {
    # Older includes all previous to 1.3.1.
    "older" : (
        # Moved before 1.1.9
        ("portraits/core/", "portraits/"),
        # File naming error made repeatedly in NR and elsewhere.
        ("human-loyalists/human-", "human-loyalists/"),
        # These are picked to cover as many as possible of the broken
        # references in UMC on the campaign server.  Some things we
        # don't try to fix include:
        # - attack/staff.png may map to one of several staves.
        # - magic.wav may map to one of several sounds depending on the unit.
        # Some other assumptions that are sound in current UMC as of April 2007
        # but theoretically dubious are marked with *.
        ("../music/defeat.ogg",     "defeat.ogg"),
        ("../music/victory.ogg",    "victory.ogg"),
        ("AMLA TOUGH",              "AMLA_TOUGH"),
        ("AMLA_TOUGH_2",            "AMLA_TOUGH 2"),
        ("AMLA_TOUGH_3",            "AMLA_TOUGH 3"),
        ("SOUND_LIST:DAGGER_SWISH", "SOUND_LIST:SWORD_SWISH"),
        ("arrow-hit.wav",           "bow.ogg"),
        ("arrow-miss.wav",          "bow-miss.ogg"),
        ("attacks/animal-fangs.png","attacks/fangs-animal.png"),
        ("attacks/crossbow.png",    "attacks/human-crossbow.png"),        #*
        ("attacks/dagger.png",      "attacks/human-dagger.png"),          #*
        ("attacks/darkstaff.png",   "attacks/staff-necromantic.png"),
        ("attacks/human-fist.png",  "attacks/fist-human.png"),
        ("attacks/human-mace.png",  "attacks/mace.png"),
        ("attacks/human-sabre.png", "attacks/sabre-human.png"),
        ("attacks/icebolt.png",     "attacks/iceball.png"),    # Is this right?
        ("attacks/lightingbolt.png","attacks/lightning.png"),
        ("attacks/missile.png",     "attacks/magic-missile.png"),
        ("attacks/morning_star.png","attacks/morning-star.png"),
        ("attacks/plaguestaff.png", "attacks/staff-plague.png"),
        ("attacks/slam.png",        "attacks/slam-drake.png"),
        ("attacks/staff-magical.png","attacks/staff-magic.png"),
        ("attacks/sword-paladin.png","attacks/sword-holy.png"),
        ("attacks/sword.png",       "attacks/human-sword.png"),                #*
        ("attacks/sword_holy.png",  "attacks/sword-holy.png"),
        ("attacks/throwing-dagger-human.png", "attacks/dagger-thrown-human.png"),
        ("bow-hit.ogg",             "bow.ogg"),
        ("bow-hit.wav",             "bow.ogg"),
        ("bowman-attack-sword.png", "bowman-sword-1.png"),
        ("bowman-attack1.png",      "bowman-ranged-1.png"),
        ("bowman-attack2.png",      "bowman-ranged-2.png"),
        ("creepy.ogg",              "underground.ogg"),
        ("dwarves/warrior.png",     "dwarves/fighter.png"),
        ("eagle.wav",               "gryphon-shriek-1.ogg"),
        ("elfland.ogg",             "elf-land.ogg"),
        ("elvish-fighter.png",      "elves-wood/fighter.png"),
        ("elvish-hero.png",         "elves-wood/hero.png"),
        ("fist.wav",                "fist.ogg"),
        ("flame-miss.ogg",          "flame-big-miss.ogg"),
        ("flame.ogg",               "flame-big.ogg"),
        ("gameplay2.ogg",           "gameplay02.ogg"),                    # Changes in 1.3.2
        ("goblin-hit2.ogg",         "goblin-hit-2.ogg"),
        ("hatchet-miss-1.ogg",      "hatchet-miss.wav"),
        ("heal.ogg",                "heal.wav"),
        ("hiss-big.ogg",            "hiss-big.wav"),
        ("human-dagger.png",        "dagger-human.png"),
        ("human-male-die.ogg",      "human-die-1.ogg"),
        ("human-male-hit.ogg",      "human-hit-1.ogg"),
        ("human-male-weak-die.ogg", "human-old-die-1.ogg"),
        ("human-male-weak-hit.ogg", "human-old-hit-1.ogg"),
        ("human-sword.png",         "sword-human.png"),
        ("items/castle-ruins.png",  "scenery/castle-ruins.png"),
        ("items/fire.png",          "scenery/fire1.png"),
        ("items/fire1.png",         "scenery/fire1.png"),
        ("items/fire2.png",         "scenery/fire2.png"),
        ("items/fire3.png",         "scenery/fire3.png"),
        ("items/fire4.png",         "scenery/fire4.png"),
        ("items/hero-icon.png",     "misc/hero-icon.png"),
        ("items/leanto.png",        "scenery/leanto.png"),
        ("items/lighthouse.png",    "scenery/lighthouse.png"),
        ("items/monolith1.png",     "scenery/monolith1.png"),
        ("items/monolith2.png",     "scenery/monolith2.png"),
        ("items/monolith3.png",     "scenery/monolith3.png"),
        ("items/monolith4.png",     "scenery/monolith4.png"),
        ("items/ring1.png",         "items/ring-silver.png"),  # Is this right?
        ("items/ring2.png",         "items/ring-gold.png"),    # Is this right?
        ("items/rock1.png",         "scenery/rock1.png"),
        ("items/rock2.png",         "scenery/rock2.png"),
        ("items/rock3.png",         "scenery/rock3.png"),
        ("items/rock4.png",         "scenery/rock4.png"),
        ("items/signpost.png",      "scenery/signpost.png"),
        ("items/slab.png",          "scenery/slab-1.png"),
        ("items/well.png",          "scenery/well.png"),
        ("knife.ogg",               "dagger-swish.wav"),       # Is this right?
        ("knife.wav",               "dagger-swish.wav"),       # Is this right?
        ("lightning.wav",           "lightning.ogg"),
        ("longbowman-ranged-1.png", "longbowman-bow-attack1.png"),
        ("longbowman-ranged-2.png", "longbowman-bow-attack2.png"),
        ("longbowman-ranged-3.png", "longbowman-bow-attack3.png"),
        ("longbowman-ranged-4.png", "longbowman-bow-attack4.png"),
        ("misc/chest.png",          "items/chest.png"),
        ("misc/dwarven-doors.png",  "scenery/dwarven-doors-closed.png"),
        ("misc/mine.png",           "scenery/mine-abandoned.png"),
        ("misc/nest-empty.png",     "scenery/nest-empty.png"),
        ("misc/rocks.png",          "scenery/rubble.png"),
        ("misc/snowbits.png",       "scenery/snowbits.png"),
        ("misc/temple.png",         "scenery/temple1.png"),
        ("miss.wav",                "miss-1.ogg"),
        ("orc-die.wav",             "orc-die-1.ogg"),
        ("orc-hit.wav",             "orc-hit-1.ogg"),
        ("ork-die-2.ogg",           "orc-die-2.ogg"),
        ("pistol.wav",              "gunshot.wav"),
        ("spear-miss-1.ogg",        "spear-miss.ogg"),
        ("spearman-attack-south-1.png", "spearman-attack-s-1.png"),
        ("spearman-attack-south-2.png", "spearman-attack-s-2.png"),
        ("spearman-attack-south-3.png", "spearman-attack-s-3.png"),
        ("squishy-miss-1.ogg",      "squishy-miss.wav"),
        ("sword-swish.wav",         "sword-1.ogg"),
        ("sword.wav",               "sword-1.ogg"),
        ("terrain/flag-1.png",      "flags/flag-1.png"),
        ("terrain/flag-2.png",      "flags/flag-2.png"),
        ("terrain/flag-3.png",      "flags/flag-3.png"),
        ("terrain/flag-4.png",      "flags/flag-4.png"),
        ("terrain/rocks.png",       "scenery/rock2.png"),
        ("terrain/signpost.png",    "scenery/signpost.png"),
        ("terrain/village-cave-tile.png","terrain/village/cave-tile.png"),
        ("terrain/village-dwarven-tile.png","terrain/village/dwarven-tile.png"),
        ("terrain/village-elven4.png","terrain/village/elven4.png"),
        ("terrain/village-human-snow.png", "terrain/village/human-snow.png"),
        ("terrain/village-human.png","terrain/village/human.png"),
        ("terrain/village-human4.png","terrain/village/human4.png"),
        ("throwing-dagger-swish.wav","dagger-swish.wav"),      # Is this right?
        ("units/undead/ghost-attack.png", "units/undead/ghost-attack-2.png"),
        ("units/undead/ghost-attack1.png", "units/undead/ghost-attack-1.png"),
        ("wolf-attack.wav",         "wolf-bite.ogg"),
        ("wolf-cry.wav",            "wolf-die.wav"),
        ("wose-attack.wav",         "wose-attack.ogg"),
        (r"wose\.attack.ogg",       "wose-attack.ogg"),
    ),
    "1.3.1" : (
        # Peasant images moved to a new directory
        ("human-loyalists/peasant.png", "human-peasants/peasant.png"),
        ("human-loyalists/peasant-attack.png", "human-peasants/peasant-attack.png"),
        ("human-loyalists/peasant-attack2.png", "human-peasants/peasant-attack2.png"),
        ("human-loyalists/peasant-ranged.png", "human-peasants/peasant-ranged.png"),
        ("human-loyalists/peasant-idle-1.png", "human-peasants/peasant-idle-1.png"),
        ("human-loyalists/peasant-idle-2.png", "human-peasants/peasant-idle-2.png"),
        ("human-loyalists/peasant-idle-3.png", "human-peasants/peasant-idle-3.png"),
        ("human-loyalists/peasant-idle-4.png", "human-peasants/peasant-idle-4.png"),
        ("human-loyalists/peasant-idle-5.png", "human-peasants/peasant-idle-5.png"),
        ("human-loyalists/peasant-idle-6.png", "human-peasants/peasant-idle-6.png"),
        ("human-loyalists/peasant-idle-7.png", "human-peasants/peasant-idle-7.png"),
        # All Great Mage attacks were renamed
        ("great-mage-attack-magic1.png", "great-mage-attack-magic-1.png"),
        ("great-mage-attack-magic2.png", "great-mage-attack-magic-2.png"),
        ("great-mage+female-attack-magic1.png", "great-mage+female-attack-magic-1.png"),
        ("great-mage+female-attack-magic2.png", "great-mage+female-attack-magic-2.png"),
        ("great-mage-attack-staff1.png", "great-mage-attack-staff-1.png"),
        ("great-mage-attack-staff2.png", "great-mage-attack-staff-2.png"),
        ("great-mage+female-attack-staff1.png", "great-mage+female-attack-staff-1.png"),
        ("great-mage+female-attack-staff2.png", "great-mage+female-attack-staff-2.png"),
        # All Arch Mage attacks were renamed
        ("arch-mage-attack-magic1.png", "arch-mage-attack-magic-1.png"),
        ("arch-mage-attack-magic2.png", "arch-mage-attack-magic-2.png"),
        ("arch-mage+female-attack-magic1.png", "arch-mage+female-attack-magic-1.png"),
        ("arch-mage+female-attack-magic2.png", "arch-mage+female-attack-magic-2.png"),
        ("arch-mage-attack-staff1.png", "arch-mage-attack-staff-1.png"),
        ("arch-mage-attack-staff2.png", "arch-mage-attack-staff-2.png"),
        ("arch-mage+female-attack-staff1.png", "arch-mage+female-attack-staff-1.png"),
        ("arch-mage+female-attack-staff2.png", "arch-mage+female-attack-staff-2.png"),
        # All Red Mage attacks were renamed
        ("red-mage-attack-magic1.png", "red-mage-attack-magic-1.png"),
        ("red-mage-attack-magic2.png", "red-mage-attack-magic-2.png"),
        ("red-mage+female-attack-magic1.png", "red-mage+female-attack-magic-1.png"),
        ("red-mage+female-attack-magic2.png", "red-mage+female-attack-magic-2.png"),
        ("red-mage-attack-staff1.png", "red-mage-attack-staff-1.png"),
        ("red-mage-attack-staff2.png", "red-mage-attack-staff-2.png"),
        ("red-mage+female-attack-staff1.png", "red-mage+female-attack-staff-1.png"),
        ("red-mage+female-attack-staff2.png", "red-mage+female-attack-staff-2.png"),
        # Timothy Pinkham supplied titles for two of his music files.
        # Zhaytee supplied a title for wesnoth-1.ogg
        # gameplay03.ogg, and and wesnoth-[25].ogg already had titles.
        ("gameplay01.ogg", "knolls.ogg"),
        ("gameplay02.ogg", "wanderer.ogg"),
        ("gameplay03.ogg", "battle.ogg"),
        ("wesnoth-1.ogg", "revelation.ogg"),
        ("wesnoth-2.ogg", "loyalists.ogg"),
        ("wesnoth-5.ogg", "northerners.ogg"),
        # And the holy->arcane change
        ("type=holy", "type=arcane"),
        ("holy=", "arcane="),
    ),
    "1.3.2" : (
        ("misc/item-holywater.png", "items/holywater.png"),
        ("orc-small-hit.wav",       "orc-small-hit-1.ogg"),
    ),
    "1.3.3" : (
        ("sounds/dragonstick-hit.ogg", "sounds/dragonstick-hit-1.ogg"),
        ("sounds/dragonstick-miss.ogg", "sounds/dragonstick-miss.wav"),
    ),
    "1.3.4" : (
        # This release changed from numeric to string palette IDs
        ("RC(magenta>1)", "RC(magenta>red)"),
        ("RC(magenta>2)", "RC(magenta>blue)"),
        ("RC(magenta>3)", "RC(magenta>green)"),
        ("RC(magenta>4)", "RC(magenta>purple)"),
        ("RC(magenta>5)", "RC(magenta>black)"),
        ("RC(magenta>6)", "RC(magenta>brown)"),
        ("RC(magenta>7)", "RC(magenta>orange)"),
        ("RC(magenta>8)", "RC(magenta>white)"),
        ("RC(magenta>9)", "RC(magenta>teal)"),
        ("colour=1", "colour=red"),
        ("colour=2", "colour=blue"),
        ("colour=3", "colour=green"),
        ("colour=4", "colour=purple"),
        ("colour=5", "colour=black"),
        ("colour=6", "colour=brown"),
        ("colour=7", "colour=orange"),
        ("colour=8", "colour=white"),
        ("colour=9", "colour=teal"),
    ),
    # 1.35 was an aborted release
    "1.3.6" : (
        ("Soul Shooter", "Banebow"),
        ("Halbardier" , "Halberdier"),
    ),
    "1.3.9" : (
        ("Outlaw Ranger", "Ranger"),
    ),
    "1.3.14" : (
        ("{AMLA_TOUGH 3}", "{AMLA_DEFAULT}"),
    ),
    # An empty sentinel value at end is required.
    "trunk" : (),
}

# Turn all the filemove string substitution pairs into nearly equivalent
# regexp-substitution pairs, forbidding the match from being preceded
# by a dash.  This prevents, e.g., "miss.ogg" false-matching on "big-miss.ogg".
for (key, value) in filemoves.items():
    filemoves[key] = [(re.compile("(?<!-)"+old), new) for (old, new) in value]

# 1.2.x to 1.3.2 terrain conversion
conversion1 = {
    " " : "_s",
    "&" : "Mm^Xm",
    "'" : "Uu^Ii",
    "/" : "Ww^Bw/",
    "1" : "1 _K",
    "2" : "2 _K",
    "3" : "3 _K",
    "4" : "4 _K",
    "5" : "5 _K",
    "6" : "6 _K",
    "7" : "7 _K",
    "8" : "8 _K",
    "9" : "9 _K",
    "?" : "Gg^Fet",
    "A" : "Ha^Vhha",
    "B" : "Dd^Vda",
    "C" : "Ch",
    "D" : "Uu^Vu",
    "E" : "Rd",
    "F" : "Aa^Fpa",
    "G" : "Gs",
    "H" : "Ha",
    "I" : "Dd",
    "J" : "Hd",
    "K" : "_K",
    "L" : "Gs^Vht",
    "M" : "Md",
    "N" : "Chr",
    "P" : "Dd^Do",
    "Q" : "Chw",
    "R" : "Rr",
    "S" : "Aa",
    "T" : "Gs^Ft",
    "U" : "Dd^Vdt",
    "V" : "Aa^Vha",
    "W" : "Xu",
    "X" : "Qxu",
    "Y" : "Ss^Vhs",
    "Z" : "Ww^Vm",
    "[" : "Uh",
    "\\": "Ww^Bw\\",
    "]" : "Uu^Uf",
    "a" : "Hh^Vhh",
    "b" : "Mm^Vhh",
    "c" : "Ww",
    "d" : "Ds",
    "e" : "Aa^Vea",
    "f" : "Gs^Fp",
    "g" : "Gg",
    "h" : "Hh",
    "i" : "Ai",
    "k" : "Wwf",
    "l" : "Ql",
    "m" : "Mm",
    "n" : "Ce",
    "o" : "Cud",
    "p" : "Uu^Vud",
    "q" : "Chs",
    "r" : "Re",
    "s" : "Wo",
    "t" : "Gg^Ve",
    "u" : "Uu",
    "v" : "Gg^Vh",
    "w" : "Ss",
    "|" : "Ww^Bw|",
    "~" : "_f",
}
max_len = max(len(elem) for elem in conversion1.values())
width = max_len+2

def neighborhood(x, y, map_):
    "Returns list of original location+adjacent locations from a hex map"
    odd = (x) % 2
    adj = [map_[y][x]];
    if x > 0:
        adj.append(map_[y][x-1])
    if x < len(map_[y])-1:
        adj.append(map_[y][x+1])
    if y > 0:
        adj.append(map_[y-1][x])
    if y < len(map_)-1:
        adj.append(map_[y+1][x])
    if x > 0 and y > 0 and not odd:
        adj.append(map_[y-1][x-1])
    if x < len(map_[y])-1 and y > 0 and not odd:
        adj.append(map_[y-1][x+1])
    if x > 0 and y < len(map_)-1 and odd:
        adj.append(map_[y+1][x-1])
    if x < len(map_[y])-1 and y < len(map_)-1 and odd:
        adj.append(map_[y+1][x+1])
    return adj

def maptransform1(filename, baseline, inmap, y):
    "Transform a map line from 1.2.x to 1.3.x format."
    global lock_terrain_coding
    # The one truly ugly piece of implementation.
    # We're relying here on maps being seen before scenario files.
    # We notice whether the maps are oldstyle (single-letter codes)
    # or newstyle (multiletter comma-seeparated fields) and retain that
    # information to help with ambiguous cases later on.  We're also relying
    # on terrain coding to be consistent within a single subdirectory.
    if len(inmap[y][0]) > 1:
        lock_terrain_coding = "newstyle"
    else:
        format = "%%%d.%ds" % (width, max_len)
        for (x, field) in enumerate(inmap[y]):
            if field in conversion1:
                lock_terrain_coding = "oldstyle"
                inmap[y][x] = format % conversion1[field]
            else:
                raise maptransform_error(filename, baseline+y+1,
                                         "unrecognized map element %s at (%s, %s)" % (repr(field), x, y))

# Mostly 1.3.1 -> 1.3.2 terrain conversions.
# One 1.3.13 -> 1.3.14 conversion -- old windmill terrain to stock villages.
# We don't version-check these because the source patterns are vanishingly
# unlikely to turn up by accident.
conversion2 = {
    re.compile(r"(?<!\^)Bww([|/\\])") : "Ww^Bw\\1",
    re.compile(r"(?<!\^)Bwo([|/\\])") : "Wo^Bw\\1",
    re.compile(r"(?<!\^)Bss([|/\\])") : "Ss^Bw\\1",
    re.compile(r"(?<!\^)Dc\b") : "Dd^Dc",
    re.compile(r"(?<!\^)Dr\b") : "Dd^Dr",
    re.compile(r"(?<!\^)Do\b") : "Dd^Do",
    re.compile(r"(?<!\^)Fa\b") : "Aa^Fpa",
    re.compile(r"(?<!\^)Fet\b") : "Gg^Fet",
    re.compile(r"(?<!\^)Ff\b") : "Gs^Fp",
    re.compile(r"(?<!\^)Ft\b") : "Gs^Ft",
    re.compile(r"(?<!\^)Rfvs\b") : "Re^Gvs",
    re.compile(r"(?<!\^)Uf\b") : "Uu^Uf",
    re.compile(r"(?<!\^)Uui\b") : "Uu^Ii",
    re.compile(r"(?<!\^)Uhi\b") : "Uh^Ii",
    re.compile(r"(?<!\^)Vda\b") : "Dd^Vda",
    re.compile(r"(?<!\^)Vdt\b") : "Dd^Vdt",
    re.compile(r"(?<!\^)Vea\b") : "Aa^Vea",
    re.compile(r"(?<!\^)Veg\b") : "Gg^Ve",
    re.compile(r"(?<!\^)Vha\b") : "Aa^Vha",
    re.compile(r"(?<!\^)Vhg\b") : "Gg^Vh",
    re.compile(r"(?<!\^)Vhh\b") : "Hh^Vhh",
    re.compile(r"(?<!\^)Vhha\b") : "Ha^Vhha",
    re.compile(r"(?<!\^)Vhm\b") : "Mm^Vhh",
    re.compile(r"(?<!\^)Vht\b") : "Gs^Vht",
    re.compile(r"(?<!\^)Vu\b") : "Uu^Vu",
    re.compile(r"(?<!\^)Vud\b") : "Uu^Vud",
    re.compile(r"(?<!\^)Vwm\b") : "Ww^Vm",
    re.compile(r"(?<!\^)Vs\b") : "Ss^Vhs",
    re.compile(r"(?<!\^)Vsm\b") : "Ss^Vm",
    re.compile(r"(?<!\^)Xm\b") : "Mm^Xm",
    re.compile(r"\bGg\^Vwm\b") : "Gg^Vh", # This is the 1.3.13 -> 1.3.14 one
    }

# Global changes meant to be done on all lines
linechanges = (
        ("canrecruit=1", "canrecruit=yes"),
        ("canrecruit=0", "canrecruit=no"),
)

def maptransform2(filename, baseline, inmap, y):
    "Convert a map line between 1.3.x formats."
    for x in range(len(inmap[y])):
        # General conversions
        for (old, new) in conversion2.items():
            inmap[y][x] = old.sub(new, inmap[y][x])
        # Convert keeps according to adjacent hexes
        if "_K" in inmap[y][x]:
            adj = [elem.strip() for elem in neighborhood(x, y, inmap)]

            # print("adjacent: %s" % adj)
            hexcount = {}
            # Intentionally skipping 0 as it is original hex
            for i in range(1, len(adj)):
                if adj[i].startswith("C"): # this is a castle hex
                    # Magic: extract second character of each adjacent castle,
                    # which is its base type.  Count occurrences of each type.
                    basetype = adj[i][1]
                    hexcount[basetype] = hexcount.get(basetype, 0) + 1
            maxc = 0;
            maxk = "h";
            # Note: if two kinds of basetype tie for most instances adjacent,
            # which one dominates will be a pseudorandom artifact of
            # Python's hash function.
            for k in hexcount.keys():
                if hexcount[k] > maxc:
                    maxc = hexcount[k]
                    maxk = k
            #print("Dominated by %s" % maxk)
            inmap[y][x] = inmap[y][x].replace("_K", "K" + maxk)
            # There's only one kind of underground keep at present.
            inmap[y][x] = inmap[y][x].replace("Ku", "Kud")

def validate_stack(stack, filename, lineno):
    "Check the stack for deprecated WML syntax."
    if verbose >= 3:
        print('"%s", line %d: %s' % (filename, lineno+1, stack))
    if stack:
        (tag, attributes) = tagstack[-1]
        ancestors = [elem[0] for elem in tagstack]
        #if tag == "sound" and "attack" in ancestors:
        #    print('"%s", line %d: deprecated [sound] within [attack] tag' % (filename, lineno+1))

def validate_on_pop(tagstack, closer, filename, lineno):
    "Validate the stack at the time a new close tag is seen."
    (tag, attributes) = tagstack[-1]
    ancestors = [elem[0] for elem in tagstack]
    if verbose >= 3:
        print('"%s", line %d: closing %s I see %s with %s' % (filename, lineno, closer, tag, attributes))
    # Detect a malformation that will cause the game to barf while attempting
    # to deserialize an empty unit.
    if closer == "side" and "type" not in attributes and ("no_leader" not in attributes or attributes["no_leader"] != "yes") and "multiplayer" not in ancestors:
        print('"%s", line %d: [side] without type attribute' % (filename, lineno))

def within(tag):
    "Did the specified tag lead one of our enclosing contexts?"
    return tag in [elem[0] for elem in tagstack]

# Sanity checking

# Associations for the ability sanity checks.
# Note: Depends on ABILITY_EXTRA_HEAL not occurring outside ABILITY_CURES.
notepairs = (
    ("movement_type=undeadspirit", "{SPECIAL_NOTES_SPIRIT}"),
    ("type=arcane", "{SPECIAL_NOTES_ARCANE}"),
    ("{ABILITY_HEALS}", "{SPECIAL_NOTES_HEALS}"),
    #("{ABILITY_EXTRA_HEAL}", "{SPECIAL_NOTES_EXTRA_HEAL}"),
    ("{ABILITY_UNPOISON}", "{SPECIAL_NOTES_UNPOISON}"),
    ("{ABILITY_CURES}", "{SPECIAL_NOTES_CURES}"),
    ("{ABILITY_REGENERATES}", "{SPECIAL_NOTES_REGENERATES}"),
    ("{ABILITY_STEADFAST}", "{SPECIAL_NOTES_STEADFAST}"),
    ("{ABILITY_LEADERSHIP_LEVEL_", "{SPECIAL_NOTES_LEADERSHIP}"), # No } deliberately
    ("{ABILITY_SKIRMISHER}", "{SPECIAL_NOTES_SKIRMISHER}"),
    ("{ABILITY_ILLUMINATES}", "{SPECIAL_NOTES_ILLUMINATES}"),
    ("{ABILITY_TELEPORT}", "{SPECIAL_NOTES_TELEPORT}"),
    ("{ABILITY_AMBUSH}", "{SPECIAL_NOTES_AMBUSH}"),
    ("{ABILITY_NIGHTSTALK}", "{SPECIAL_NOTES_NIGHTSTALK}"),
    ("{ABILITY_CONCEALMENT}", "{SPECIAL_NOTES_CONCEALMENT}"),
    ("{ABILITY_SUBMERGE}", "{SPECIAL_NOTES_SUBMERGE}"),
    ("{ABILITY_FEEDING}", "{SPECIAL_NOTES_FEEDING}"),
    ("{WEAPON_SPECIAL_BERSERK}", "{SPECIAL_NOTES_BERSERK}"),
    ("{WEAPON_SPECIAL_BACKSTAB}", "{SPECIAL_NOTES_BACKSTAB}"),
    ("{WEAPON_SPECIAL_PLAGUE", "{SPECIAL_NOTES_PLAGUE}"), # No } deliberately
    ("{WEAPON_SPECIAL_SLOW}", "{SPECIAL_NOTES_SLOW}"),
    ("{WEAPON_SPECIAL_STONE}", "{SPECIAL_NOTES_STONE}"),
    ("{WEAPON_SPECIAL_MARKSMAN}", "{SPECIAL_NOTES_MARKSMAN}"),
    ("{WEAPON_SPECIAL_MAGICAL}", "{SPECIAL_NOTES_MAGICAL}"),
    ("{WEAPON_SPECIAL_SWARM}", "{SPECIAL_NOTES_SWARM}"),
    ("{WEAPON_SPECIAL_CHARGE}", "{SPECIAL_NOTES_CHARGE}"),
    ("{WEAPON_SPECIAL_DRAIN}", "{SPECIAL_NOTES_DRAIN}"),
    ("{WEAPON_SPECIAL_FIRSTSTRIKE}", "{SPECIAL_NOTES_FIRSTSTRIKE}"),
    ("{WEAPON_SPECIAL_POISON}", "{SPECIAL_NOTES_POISON}"),
    )

trait_note = dict(notepairs)
note_trait = {pair[1]: pair[0] for pair in notepairs}

def string_strip(value):
    "String-strip the value"
    if value.startswith('"'):
        value = value[1:]
        if value.endswith('"'):
            value = value[:-1]
    if value.startswith("'"):
        value = value[1:]
        if value.endswith("'"):
            value = value[:-1]
    return value

def parse_attribute(str_):
    "Parse a WML key-value pair from a line."
    if '=' not in str_:
        return None
    where = str_.find("=")
    leader = str_[:where]
    after = str_[where+1:]
    after = after.lstrip()
    if "#" in after:
        where = after.find("#")
        while after[where-1] in (" ", "\t"):
            where -= 1
        value = after[:where+1]
        comment = after[where:]
    else:
        value = after.rstrip()
        comment = ""
    # Return four fields: stripped key, part of line before value,
    # value, trailing whitespace and comment.
    return (leader.strip(), leader+"=", string_strip(value), comment)

# This needs to match the list of usage types in ai_python.cpp
usage_types = ("scout", "fighter", "mixed fighter", "archer", "healer")

# These are accumulated by sanity_check() and examined by sanity_postcheck()
usage = {}
sides = []
movetypes = []
unit_movetypes = []
races = []
unit_races = []

# Will be used to move files to _main.cfg
is_main = []

def sanity_check(filename, lines):
    "Perform sanity and consistency checks on input lines."
    modified = False
    unit_id = ""
    # Sanity-check abilities and traits against notes macros.
    # Note: This check is disabled on units derived via [base_unit].
    # Also, build dictionaries of unit movement types and races
    in_unit = False
    in_attack_filter = False
    for i, line in enumerate(lines):
        if "[attack_filter]" in line:
            in_attack_filter = True
            continue
        elif "[/attack_filter]" in line:
            in_attack_filter = False
            continue
        elif "[unit]" in line:
            traits = []
            notes = []
            has_special_notes = False
            derived_unit = False
            in_unit = i+1
            continue
        elif "[/unit]" in line:
            #print('"%s", %d: unit has traits %s and notes %s' %
            #      (filename, in_unit, traits, notes))
            if unit_id and not derived_unit:
                missing_notes = []
                for trait in traits:
                    tn = trait_note[trait]
                    if tn not in notes and tn not in missing_notes:
                        missing_notes.append(tn)
                missing_traits = []
                for note in notes:
                    nt = note_trait[note]
                    if nt not in traits and nt not in missing_traits:
                        missing_traits.append(nt)
                if (notes or traits) and not has_special_notes:
                    missing_notes = ["{SPECIAL_NOTES}"] + missing_notes
                if missing_notes:
                    print('"%s", line %d: unit %s is missing notes +%s' %
                          (filename, in_unit, unit_id, "+".join(missing_notes)))
                if missing_traits:
                    print('"%s", line %d: unit %s is missing traits %s' %
                          (filename, in_unit, unit_id, "+".join(missing_traits)))
                if not (notes or traits) and has_special_notes:
                    print('"%s", line %d: unit %s has superfluous {SPECIAL_NOTES}' %
                          (filename, in_unit, unit_id))
            in_unit = None
            traits = []
            notes = []
            unit_id = ""
            has_special_notes = False
        if in_unit and not in_attack_filter:
            try:
                (key, prefix, value, comment) = parse_attribute(line)
                if key == "id" and not unit_id:
                    if value[0] == "_":
                        value = value[1:].strip()
                    unit_id = value
                elif key == "usage":
                    assert(unit_id)
                    usage[unit_id] = value
                elif key == "movement_type":
                    if '{' not in value:
                        assert(unit_id)
                        unit_movetypes.append((unit_id, filename, i+1, value))
                elif key == "race":
                    if '{' not in value:
                        assert(unit_id)
                        unit_races.append((unit_id, filename, i+1, value))
            except TypeError:
                pass
            if "{SPECIAL_NOTES}" in line:
                has_special_notes = True
            if "[base_unit]" in line:
                derived_unit = True
            for (p, q) in notepairs:
                if p in line:
                    traits.append(p)
                if q in line:
                    notes.append(q)
    # Collect information on defined movement types
    in_movetype = False
    for i, line in enumerate(lines):
        if "[movetype]" in line:
            in_movetype = True
            continue
        elif "[/movetype]" in line:
            in_movetype = False
            continue
        if in_movetype:
            try:
                (key, prefix, value, comment) = parse_attribute(line)
                if key == 'name':
                    movetypes.append(value)
            except TypeError:
                pass
    # Collect information on defined races
    in_race = False
    for i, line in enumerate(lines):
        if "[race]" in line:
            in_race = True
            continue
        elif "[/race]" in line:
            in_race = False
            continue
        if in_race:
            try:
                (key, prefix, value, comment) = parse_attribute(line)
                if key == 'id':
                    races.append(value)
            except TypeError:
                pass
    # Sanity-check recruit and recruitment_pattern.
    # This code has a limitation; if there are multiple instances of
    # recruit and recruitment_pattern (as can happen if these lists
    # vary by EASY/NORMAL/HARD level) this code will only record the
    # last of each for later consistency checking.
    in_side = False
    in_subtag = False
    recruit = []
    in_generator = False
    sidecount = 0
    recruitment_pattern = []
    for i, line in enumerate(lines):
        if "[generator]" in line:
            in_generator = True
            continue
        elif "[/generator]" in line:
            in_generator = False
            continue
        elif "[side]" in line:
            in_side = True
            sidecount += 1
            continue
        elif "[/side]" in line:
            if recruit and recruitment_pattern:
                sides.append((filename, recruit, recruitment_pattern))
            in_side = False
            recruit = []
            recruitment_pattern = []
            continue
        elif in_side and ("[unit]" in line or "[ai]" in line):
            in_subtag = True
            continue
        elif in_side and ("[/side]" in line or "[/ai]" in line):
            in_subtag = False
        if not in_side or in_subtag or '=' not in line:
            continue
        try:
            (key, prefix, value, comment) = parse_attribute(line)
            if key == "recruit" and value:
                recruit = (i+1, [elem.strip() for elem in value.split(",")])
            elif key == "recruitment_pattern" and value:
                recruitment_pattern = (i+1, [elem.strip() for elem in value.split(",")])
                for utype in recruitment_pattern[1]:
                    if not utype in usage_types:
                        print('"%s", line %d: unknown usage class %s' %
                              (filename, i+1, utype))
            elif key == "side":
                try:
                    if not in_generator and sidecount != int(value):
                        print('"%s", line %d: side number %s is out of sequence' %
                              (filename, i+1, value))
                except ValueError:
                    pass # Ignore ill-formed integer literals
        except TypeError:
            pass
    # Consistency-check the description attributes in [side], [unit], [recall],
    # and [message] scopes, also correctness-check translation marks.
    present = []
    in_scenario = False
    in_person = False
    in_objective = False
    in_trait = False
    ignoreable = False
    for i, line in enumerate(lines):
        if "[scenario]" in line:
            in_scenario = True
        elif "[/scenario]" in line:
            in_scenario = False
        elif "[objective]" in line:
            in_objective = True
        elif "[/objective]" in line:
            in_objective = False
        elif "[trait]" in line:
            in_trait = True
        elif "[/trait]" in line:
            in_trait = False
        elif "[kill]" in line or "[object]" in line or "[move_unit_fake]" in line or "[scroll_to_unit]" in line:
            ignoreable = True
        elif "[/kill]" in line or "[/object]" in line or "[/move_unit_fake]" in line or "[/scroll_to_unit]" in line:
            ignoreable = False
        elif "[side]" in line or "[unit]" in line or "[recall]" in line:
            in_person = True
            continue
        elif "[/side]" in line or "[/unit]" in line or "[/recall]" in line:
            in_person = False
        if not in_scenario:
            continue
        m = re.search("# *wmllint: recognize +(.*)", line)
        if m:
            present.append(string_strip(m.group(1)).strip())
        if '=' not in line or ignoreable:
            continue
        try:
            (key, prefix, value, comment) = parse_attribute(line)
            if "wmllint: ignore" in comment:
                continue
            if len(value) == 0:
                continue
            has_tr_mark = value.lstrip().startswith("_")
            if key == 'role':
                present.append(value)
            if has_tr_mark:
                if '{' in value:
                    print('"%s", line %d: macro reference in translatable string' %
                          (filename, i+1))
                if future and re.search("[.,!?]  ", line):
                    print('"%s", line %d: extraneous space in translatable string' %
                          (filename, i+1))
            # Check correctness of translation marks and descriptions
            if key.startswith("#"): # FIXME: parse_attribute is confused.
                pass
            elif key == 'letter': # May be led with _s for void
                pass
            elif key == 'name': # FIXME: check this someday
                pass
            elif key in ("message", "user_description", "story", "note", "text", "summary", "caption", "label", "unit_description") and not value.startswith("$"):
                if not has_tr_mark:
                    print('"%s", line %d: %s needs translation mark' %
                          (filename, i+1, key))
                    value = "_ " + value
                    modified = True
            elif  key == "description":
                if (in_trait or in_objective) and not has_tr_mark:
                    print('"%s", line %d: description in [objectives] needs translation mark' %
                          (filename, i+1))
                    value = "_ " + value
                    modified = True
                elif not (in_trait or in_objective) and has_tr_mark:
                    print('"%s", line %d: description should not have translation mark' %
                          (filename, i+1))
                    value = value.replace("_", "", 1)
                    modified = True
                if in_person:
                    present.append(value)
                elif value in ('narrator', 'unit', 'second_unit') or value[0] in ("$", "{"):
                    continue
                elif not in_objective and value not in present:
                    print('"%s", line %d: unknown \'%s\' referred to by description' %
                          (filename, i+1, value))
            elif has_tr_mark:
                print('"%s", line %d: %s should not have a translation mark' %
                      (filename, i+1, key))
                value = value.replace("_", "", 1)
                modified = True
        except TypeError:
            pass
    # Interpret magic comments for setting the sage pattern of units.
    # This copes with some wacky UtBS units that are defined with
    # variant-spawning macros.  The prototype comment looks like this:
    #wmllint: usage of "Desert Fighter" is fighter
    for i, line in enumerate(lines):
        m = re.match('# *wmllint: usage of "([^"]*)" is +(.*)', line)
        if m:
            usage[m.group(1)] = m.group(2).strip()
    # Check for textdomain strings; should be exactly one, on line 1
    # We will also take the opportunity to check if the file is a top-level main
    textdomains = []
    for i, line in enumerate(lines):
        if ("[campaign]" in line or "[binary_path]" in line or "[textdomain]" in line) and not filename.endswith("_main.cfg") and not filename in is_main:
            is_main.append(filename)
        if "#textdomain" in line:
            textdomains.append(i+1)
    if not textdomains:
        print('"%s", line 1: no textdomain string' % filename)
    elif textdomains[0] == 1: # Multiples are OK if first is on line 1
        pass
    elif len(textdomains) > 1:
        print('"%s", line %d: multiple textdomain strings on lines %s' %
              (filename, textdomains[0], ", ".join(map(str, textdomains))))
    else:
        w = textdomains[0]
        print('"%s", line %d: single textdomain declaration not on line 1.' %
              (filename, w))
        lines = [lines[w-1].lstrip()] + lines[:w-1] + lines[w:]
        modified = True
    return (lines, modified)

def consistency_check():
    "Consistency-check state information picked up by sanity_check"
    utypes = []
    for (filename, (rl, recruit), (pl, recruitment_pattern)) in sides:
        #print("%s: %d=%s, %d=%s" % (filename, rl, recruit, pl, recruitment_pattern))
        for rtype in recruit:
            if rtype not in usage:
                print('"%s", line %d: %s has no usage type' % (filename, rl, rtype))
                continue
            utype = usage[rtype]
            if utype not in recruitment_pattern:
                print('"%s", line %d: %s (%s) doesn\'t match the recruitment pattern (%s) for its side' %
                      (filename, rl, rtype, utype, ", ".join(recruitment_pattern)))
            utypes.append(utype)
        for utype in recruitment_pattern:
            if utype not in utypes:
                print('"%s", line %d: %s doesn\'t match a recruitable type for its side' %
                      (filename, rl, utype))
    if movetypes:
        for (unit_id, filename, line, movetype) in unit_movetypes:
            if movetype not in movetypes:
                print('"%s", line %d: %s has unknown movement type' %
                      (filename, line, unit_id))
    if races:
        for (unit_id, filename, line, race) in unit_races:
            if race not in races:
                print('"%s", line %d: %s has unknown race' %
                      (filename, line, unit_id))

# Syntax transformations

leading_ws = re.compile(r"^\s*")

def leader(s):
    "Return a copy of the leading whitespace in the argument."
    return leading_ws.match(s).group(0)

def outdent(s):
    "Outdent line by one level."
    if s.startswith(baseindent):
        return s[len(baseindent):]
    elif s.endswith("\t"):
        return s[:-1] + baseindent
    else:
        return s

def hack_syntax(filename, lines):
    # Syntax transformations go here.  This gets called once per WML file;
    # the name of the file is passed as filename, text of the file as the
    # array of strings in lines.  Modify lines in place as needed, and
    # set modcount to nonzero when you actually change any.
    global versions
    modcount = 0
    # Ensure that every attack has a translatable description.
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        elif "[attack]" in line:
            j = i;
            have_description = False
            while '[/attack]' not in lines[j]:
                if lines[j].strip().startswith("description"):
                    have_description = True
                j += 1
            if not have_description:
                j = i
                while '[/attack]' not in lines[j]:
                    fields = lines[j].strip().split('#')
                    syntactic = fields[0]
                    comment = ""
                    if len(fields) > 1:
                        comment = fields[1]
                    if syntactic.strip().startswith("name"):
                        description = syntactic.split("=")[1].strip()
                        if not description.startswith('"'):
                            description = '"' + description + '"'
                        # Skip the insertion if this is a dummy declaration
                        # or one modifying an attack inherited from a base unit.
                        if "no-icon" not in comment:
                            new_line = leader(lines[j]) + "description=_"+description+'\n'
                            if verbose:
                                print('"%s", line %d: inserting %s' % (filename, i+1, repr(new_line)))
                            lines.insert(j+1, new_line)
                            j += 1
                            modcount += 1
                    j += 1
    # Ensure that every speaker=narrator block without an image uses
    # wesnoth-icon.png as an image.
    need_image = False
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        precomment = line.split("#")[0]
        if "speaker=narrator" in precomment:
            need_image = True
        elif precomment.strip().startswith("image"):
            need_image = False
        elif '[/message]' in precomment:
            if need_image:
                # This line presumes the code has been through wmlindent
                if verbose:
                    print('wmllint: "%s", line %d: inserting "image=wesnoth-icon.png"'%(filename, i+1))
                lines.insert(i, leader(precomment) + baseindent + "image=wesnoth-icon.png\n")
                modcount += 1
            need_image = False
    # Remove get_hit_sound fields (and image_defensive, if also present)
    in_unit = False
    unit_id = ""
    unit_image = None
    unit_sound = None
    get_hit_sound = None
    defend_image = None
    image_defensive = None
    in_defend = False
    defend_frame = None
    has_defense_anim = None
    has_special_notes = False
    image_done = []
    for i, line in enumerate(lines):
        if "[unit]" in line:
            in_unit = i+1
            continue
        elif "[/unit]" in line:
            if has_defense_anim:
                if get_hit_sound:
                    print('wmllint: "%s", lines %d, %d: unit%s has both deprecated get_hit_sound key and a DEFENSE_ANIM' %
                          (filename, get_hit_sound+1, has_defense_anim+1, unit_id))
                if image_defensive:
                    if re.search('{DEFENSE_ANIM[A-Z_]* +["(]*' + defend_image, lines[has_defense_anim]):
                        image_done.append(image_defensive)
                    else:
                        print('wmllint: "%s", lines %d, %d: unit%s has both outdated image_defensive key and a DEFENSE_ANIM' %
                              (filename, image_defensive+1, has_defense_anim+1, unit_id))
                if defend_frame:
                    print('wmllint: "%s", lines %d, %d: unit%s may have both a [defend] animation block and a DEFENSE_ANIM' %
                          (filename, defend_frame+1, has_defense_anim+1, unit_id))
            elif unit_id and get_hit_sound:
                if not defend_image:
                    defend_image = unit_image
                new_anim = "{DEFENSE_ANIM %s %s %s}" % \
                           (defend_image, unit_image, unit_sound)
                (key, prefix, val, comment) = parse_attribute(lines[get_hit_sound])
                print('wmllint: "%s", line %d: unit%s gets %s'%(filename, get_hit_sound+1, unit_id, new_anim))
                lines[get_hit_sound] = leader(lines[get_hit_sound]) \
                                       + new_anim + comment + "\n"
                modcount += 1
                if image_defensive:
                    image_done.append(image_defensive)
                if defend_frame:
                    print('wmllint: "%s", lines %d, %d: unit%s may have both a [defend] animation block and a DEFENSE_ANIM' %
                          (filename, defend_frame+1, get_hit_sound+1, unit_id))
            elif image_defensive:
                print('wmllint: "%s", line %d: unit%s has outdated image_defensive key' %
                      (filename, image_defensive+1, unit_id))
            in_unit = None
            unit_id = ""
            unit_image = None
            unit_sound = None
            get_hit_sound = None
            defend_image = None
            image_defensive = None
            in_defend = False
            defend_frame = None
            has_defense_anim = None
            has_special_notes = False
        if in_unit:
            if "{DEFENSE_ANIM" in line and not has_defense_anim:
                has_defense_anim = i
            if "[defend]" in line:
                in_defend = True
            if "[/defend]" in line or "[attack]" in line:
                in_defend = False
            if in_defend and not defend_frame and "[frame]" in line:
                defend_frame = i
            else:
                fields = parse_attribute(line)
                if fields is None:
                    continue
                (key, prefix, value, comment) = fields
                if key == "id" and not unit_id:
                    unit_id = value
                    if unit_id[0] == "_":
                        unit_id = unit_id[1:].strip()
                    unit_id = " " + unit_id
                elif key == "get_hit_sound":
                    get_hit_sound = i
                    unit_sound = value
                elif key == "image" and not unit_image:
                    unit_image = value
                elif key == "image_defensive" and not image_defensive:
                    defend_image = value
                    image_defensive = i
                elif key == "image" and in_defend and not defend_image:
                    defend_image = value
    # If image_defensive is the same image as the one in DEFENSE_ANIM, it is safe to discard
    image_done.reverse()
    for bye in image_done:
        print('wmllint: "%s", line %d: removing outdated attribute (%s)' %
              (filename, bye+1, lines[bye].strip()))
        del lines[bye]
        modcount += 1
    # Boucman's transformation of animation syntax.
    # Wrap this in try/except because some newer animation syntax chokes it.
    try:
        class anim_frame:
            def __init__(self, attackline, attackname, lineno, female, variation):
                self.attackstart = attackline
                self.name = attackname
                self.animstart = lineno
                self.female = female
                self.variation = variation
                self.animend = None
                self.attackend = None
            def __repr__(self):
                return repr(self.__dict__)
        in_attack = in_animation = in_female = in_variation = False
        animations = []
        attackname = None
        attackline = None
        for i, line in enumerate(lines):
            if "no-syntax-rewrite" in line:
                break
            elif "[female]" in line:
                in_female = True
            elif "[/female]" in line:
                in_female = False
            elif "[variation]" in line:
                variation_index += 1
                in_variation = True
            elif "[/variation]" in line:
                in_variation = False
            elif "[unit]" in line:
                in_attack = in_animation = in_female = in_variation = False
                female_attack_index = -1
                variation_index = 0
                male_attack_start = len(animations)
            elif "[attack]" in line:
                in_attack = True;
                attackname = None
                attackline = i
                if in_female:
                    female_attack_index += 1
            elif "[animation]" in line and in_attack:
                #if verbose:
                #    print('"%s", line %d: [animation] within [attack]' %
                #          (filename, i+1))
                # This weird piece of code is because attacks for female
                # variants don't have names.  Instead, they're supposed
                # to pick up the name of the corresponding male attack,
                # where correspondence is by order of declaration.  The
                # male_attack_start variable copes with the possibility
                # of multiple units per file.
                if attackname == None and in_female:
                    attackname = animations[male_attack_start + female_attack_index].name
                if not attackname:
                    print('"%s", line %d: cannot deduce attack name'%(filename, i+1))
                if in_variation:
                    variation = variation_index
                else:
                    variation = None
                animations.append(anim_frame(attackline, attackname, i, in_female, variation))
                in_animation = True
            elif "[/animation]" in line and in_attack:
                in_animation = False
                if animations and animations[-1].animstart != None and animations[-1].animend == None:
                    animations[-1].animend = i
                else:
                    print('"%s", line %d: [animation] ending here may be ill-formed'%(filename, i+1))
            elif "[/attack]" in line:
                inattack = False;
                attackname = None
                if animations and (animations[-1].attackstart == None or animations[-1].attackend != None):
                    print('"%s", line %d: [attack] ending here may be ill-formed'%(filename, i+1))
                elif animations:
                    # This loop is needed because a single attack tag may
                    # enclose both hit and miss animations.
                    j = len(animations)-1
                    while True:
                        animations[j].attackend = i
                        j -= 1
                        if j < 0 or animations[j].attackend != None:
                            break
            # Only pick up the *first* name field in an attack block;
            # by convention, it will be right after the opening [attack] tag
            elif in_attack and not in_animation and not attackname:
                #print(filename + ":" + repr(i+1) + ";" + repr(line))
                fields = line.strip().split('#')
                syntactic = fields[0]
                comment = ""
                if len(fields) > 1:
                    comment = fields[1]
                if syntactic.strip().startswith("name"):
                    attackname = syntactic.split("=")[1].strip()
        boucmanized = False
        # All animation ranges have been gathered,  We have a list of objects
        # containing the attack information.  Reverse it, because we're
        # going to process them back to front to avoid invalidating the
        # already-collected line numbers.  Then pull out the animation
        # WML and stash it in the frame objects.
        animations.reverse()
        for aframe in animations:
            if verbose:
                print('"%s", line %d: lifting animation block at %d:%d for %s attack (%d:%d)' %
                      (filename, aframe.animstart+1, aframe.animstart+1, aframe.animend+1, aframe.name, aframe.attackstart+1, aframe.attackend+1))
            # Make a copy of the animation block, change its enclosing tags,
            # outdent it, and add the needed filter clause.
            animation = lines[aframe.animstart:aframe.animend+1]
            animation[0] = animation[0].replace("[animation]", "[attack_anim]")
            animation[-1] = animation[-1].replace("[/animation]","[/attack_anim]")
            for i in range(len(animation)):
                animation[i] = outdent(animation[i])
            indent = leader(animation[1])
            animation.insert(1, indent + "[/attack_filter]\n")
            animation.insert(1, indent + baseindent + "name="+aframe.name+"\n")
            animation.insert(1, indent + "[attack_filter]\n")
            # Save it and delete it from its original location
            aframe.wml = "".join(animation)
            lines = lines[:aframe.animstart] + lines[aframe.animend+1:]
            modcount += 1
            boucmanized = True
        # Insert non-variation attacks where they belong
        female_attacks = [a for a in animations if a.female and a.variation == None]
        female_attacks.reverse()
        if female_attacks:
            female_end = -1
            for i, line in enumerate(lines):
                if line.rstrip().endswith("[/female]"):
                    female_end = i
                    break
            assert female_end != -1
            female_wml = "".join(map(lambda x: x.wml, female_attacks))
            lines = lines[:female_end] + [female_wml] + lines[female_end:]
        male_attacks = [a for a in animations if not a.female and a.variation == None]
        male_attacks.reverse()
        if male_attacks:
            male_end = -1
            for i, line in enumerate(lines):
                # Male attacks go either before the [female] tag or just
                # before the closing [/unit]
                if line.rstrip().endswith("[/unit]") or line.rstrip().endswith("[female]"):
                    male_end = i
                    break
            assert male_end != -1
            male_wml = "".join(map(lambda x: x.wml, male_attacks))
            lines = lines[:male_end] + [male_wml] + lines[male_end:]
        # Now insert variation attacks where they belong.
        for animation in animations:
            if animation.variation != None:
                vcount = 0
                for i, line in enumerate(lines):
                    if "[/variation]" in line:
                        vcount += 1
                    if vcount == animation.variation:
                        break
                lines = lines[:i] + [animation.wml] + lines[i:]
    except TypeError:
        pass
    # Garbage-collect any empty [attack] scopes left behind;
    # this is likely to happen with female-variant units.
    nullattack = True
    while nullattack:
        nullattack = False
        for i in range(len(lines)-1):
            if lines[i].strip() == "[attack]" and lines[i+1].strip() == "[/attack]":
                nullattack = True
                break
        if nullattack:
            lines = lines[:i] + lines[i+2:]
    # Lift new_attack animation blocks within [effect] tags
    # Note: This assumes that the animation WML goes last in the [effect] WML
    # with nothing after it, and will fail if that is not true.
    in_effect = False
    attackname = None
    converting = False
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        elif "[effect]" in line:
            in_effect = True
        elif "apply_to=new_attack" in line:
            converting = True
        elif "[/effect]" in line:
            converting = in_effect = False
        elif in_effect and not attackname:
            #print(filename + ":" + repr(i+1) + ";" + repr(line))
            fields = line.strip().split('#')
            syntactic = fields[0]
            comment = ""
            if len(fields) > 1:
                comment = fields[1]
            if syntactic.strip().startswith("name"):
                attackname = syntactic.split("=")[1].strip()
        elif converting and "[animation]" in line:
            print('"%s", line %d: converting [animation] in [effect] '%(filename, i+1))
            ws = leader(line)
            outer = outdent(ws)
            assert attackname != None
            lines[i] = """{0}[/effect]
{0}[effect]
{1}apply_to=new_animation
{2}{1}{3}[attack_filter]
{1}{3}{3}name={4}
{1}{3}[/attack_filter]
""".format(outer, ws, line.replace("animation", "attack_anim"), baseindent, attackname)
            modcount += 1
        elif converting and "[/animation]" in line:
            lines[i] = line.replace("animation", "attack_anim")
    # Upconvert ancient ability declarations from 1.x
    level = None
    abilities = []
    specials = []
    lastability = None
    lastspecial = None
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        if "[unit]" in line:
            abilities = []
        if "[attack]" in line:
            specials = []
        elif "[/attack]" in line:
            if specials:
                if verbose:
                    print("Lifting obsolete specials:", " ".join(specials))
                ws = leader(line)
                insertion = ws + baseindent + "[specials]\n"
                for special in specials:
                    if special.startswith("plague("):
                        insertion += ws + baseindent*2 + "{WEAPON_SPECIAL_PLAGUE_TYPE " + special[7:-1] + "}\n"
                    elif special in ("backstab", "berserk", "charge", "drain",
                                     "firststrike", "magical", "marksman", "plague",
                                     "poison", "slow", "stone", "swarm",):
                        insertion += ws + baseindent*2 + "{WEAPON_SPECIAL_" + special.upper() + "}\n"
                    else:
                        print("Don't know how to convert '%s'" % special)
                insertion += ws + baseindent + "[/specials]\n"
                lines[lastspecial] = insertion
                modcount += 1
        elif "[/unit]" in line:
            if abilities:
                if verbose:
                    print("Lifting obsolete abilities:", " ".join(abilities))
                ws = leader(line)
                insertion = ws + baseindent + "[abilities]\n"
                for ability in abilities:
                    if ability == "leadership":
                        if level is None:
                            print("warning: can't convert ancient leadership ability")
                        else:
                            insertion += ws + baseindent*2 + "{ABILITY_LEADERSHIP_LEVEL_"+level+"}\n"
                    elif ability in ("cures", "heals", "nightstalk", "regenerates",
                                     "skirmisher", "steadfast", "illuminates",
                                     "teleport", "ambush",):
                        insertion += ws + baseindent*2 + "{ABILITY_" + ability.upper() + "}\n"
                    else:
                        print("Don't know how to convert '%s'" % ability)
                insertion += ws + baseindent + "[/abilities]\n"
                lines[lastability] = insertion
                modcount += 1
        elif line.count("=") == 1:
            (tag, value) = line.strip().split("=")
            if tag == "level":
                level = value
            if tag == "ability":
                for able in value.split(","):
                    abilities.append(able.strip())
                lastability = i
                lines[i] = ""
            if tag == "special":
                specials.append(value)
                lastspecial = i
                lines[i] = ""
    # Lift [frame] declarations directly within attacks
    in_attack = False
    attackname = None
    soundpath = None
    misspath = None
    soundtime = None
    in_sound = False
    in_frame = False
    postframe = []
    begins = []
    converting = 0
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        elif "[attack]" in line:
            in_attack = True
        elif "[/attack]" in line:
            if converting:
                assert attackname != None
                lines[i] = line.replace("/attack", "/attack_anim")
                print('"%s", line %d: converting frame in [attack]'%(filename, converting+1))
                ws = leader(lines[converting])
                outer = outdent(ws)
                if soundpath:
                    if misspath:
                        if not soundtime:
                            soundtime = "-100"
                            print('"%s", line %d: inserting "-100" for missing time value in SOUND:HIT_AND_MISS' % (filename, converting+1))
                        lines[converting] = ws + "{SOUND:HIT_AND_MISS %s %s %s}\n" \
                                            % (soundpath, misspath, soundtime) + lines[converting]
                    else:
                        for (when, where) in begins:
                            if when == soundtime:
                                lines[where] = leader(lines[where]) + "sound=" + soundpath + "\n" + lines[where]
                                break
                            continue
                        else:
                            lines[converting] += ws + baseindent + "sound=" + soundpath + "\n"
                            print('"%s", line %d: no match found for [sound] time (%s), placing sound path in first frame' %
                                  (filename, where+1, soundtime))
                insertion = """{0}[/attack]
{0}[attack_anim]
{1}[attack_filter]
{1}{2}name={3}
{1}[/attack_filter]
""".format(outer, ws, baseindent, attackname)
                lines[converting] = insertion + lines[converting]
                postframe.reverse()
                for preframe in postframe:
                    lines[converting] = preframe + lines[converting]
                modcount += 1
                converting = 0
            in_attack = False
            attackname = None
            soundpath = None
            misspath = None
            soundtime = None
            in_sound = False
            in_frame = False
            postframe = []
            begins = []
        elif ("[frame]" in line or "[missile_frame]" in line) and in_attack and converting == 0:
            converting = i
            in_frame = True
        elif in_attack:
            fields = line.strip().split('#')
            syntactic = fields[0]
            comment = ""
            if len(fields) > 1:
                comment = fields[1]
            if  not attackname and syntactic.strip().startswith("name"):
                attackname = syntactic.split("=")[1].strip()
            elif  not soundtime and syntactic.strip().startswith("time"):
                soundtime = syntactic.split("=")[1].strip()
            elif  not misspath and syntactic.strip().startswith("sound_miss"):
                misspath = syntactic.split("=")[1].strip()
            elif  not soundpath and syntactic.strip().startswith("sound"):
                soundpath = syntactic.split("=")[1].strip()
            elif  in_frame and syntactic.strip().startswith("begin"):
                begins.append((syntactic.split("=")[1].strip(), i))
            # Ignore sound tags, and their contents, within [attack]
            if "[sound]" in line:
                print('"%s", line %d: [sound] within [attack] discarded (path will be saved)' %
                      (filename, i+1))
                in_sound = True
                modcount += 1
            if "[/sound]" in line:
                lines[i] = ""
                in_sound = False
            if in_sound:
                lines[i] = ""
            # Move post-frame lines up
            if "[frame]" in line or "[missile_frame]" in line:
                in_frame = True
            elif "[/frame]" in line or "[/missile_frame]" in line:
                in_frame = False
            elif converting and not in_frame:
                postframe.append(line)
                lines[i] = ""
    # Upconvert old radius usage
    if upconvert and "1.3.7" in versions and "older" not in versions:
        radius_pos = wmlfind("radius=", WmlIterator(lines, filename))
        while radius_pos is not None:
            scopeIter = radius_pos.iterScope()
            startline = scopeIter.lineno + 1
            wspace = radius_pos.text
            wspace = wspace[:len(wspace)-len(wspace.lstrip())]
            radius_danger = False
            to_indent = []
            no_indent = []
            insideElem = 0
            for i in scopeIter:
                elem = i.element
                if elem in ("[and]", "[or]", "[not]"):
                    radius_danger = True
                    no_indent.extend(txt+'\n' for txt in i.text.splitlines())
                    insideElem += 1
                elif insideElem:
                    if elem in ("[/and]", "[/or]", "[/not]"):
                        insideElem -= 1
                    no_indent.extend(txt+'\n' for txt in i.text.splitlines())
                elif elem in ("variable=", "side=", "count=", "adjacent="):
                    no_indent.extend(txt+'\n' for txt in i.text.splitlines())
                else:
                    to_add = [txt+'\n' for txt in i.text.splitlines()]
                    to_add[0] = baseindent + to_add[0]
                    to_indent.extend(to_add)
            if radius_danger:
                lines = lines[:startline] + [wspace + "[and]\n"] + to_indent +[
                    wspace + "[/and]\n"] + no_indent + lines[scopeIter.lineno:]
                radius_pos.lines = lines
                modcount += 1
                #backup to rescan
                radius_pos.seek(startline-1)
                #pass the inserted content
                radius_pos.seek(startline+len(to_indent)+1)
            radius_pos = wmlfind("radius=", radius_pos)
    # Boucmanize death animations
    if future:
        in_death = None
        frame_commented = in_death_commented = False
        frame_start = frame_end = None
        image = None
        for i, line in enumerate(lines):
            if "no-syntax-rewrite" in line:
                break
            elif "[death]" in line:
                in_death = i
                in_death_commented = line.strip().startswith("#")
            elif "[/death]" in line:
                if frame_start is None:
                    print('"%s", %d: [death] with no frames' % (filename, i), file=sys.stderr)
                    continue
                # Find the image tag
                for inside in range(frame_start, frame_end):
                    if "image=" in lines[inside]:
                        image = lines[inside].strip().split("=")[1]
                        break
                else:
                    print('"%s", line %d: no image in last frame' %
                          (filename, i), file=sys.stderr)
                    continue
                # Modify the death wrapper
                lines[i] = line.replace("death", "animation")
                inner = leader(lines[in_death])+baseindent
                if in_death_commented:
                    inner = "#" + inner
                lines[in_death] = lines[in_death].replace("death", "animation") \
                                  + "%sapply_to=death\n" % inner
                # Add a new last frame to the death animation
                outer = leader(lines[frame_start])
                if frame_commented:
                    outer = "#" + outer
                inner = outer + baseindent
                if frame_commented:
                    inner = "#" + inner
                insertion = """{0}[frame]
{1}duration=600
{1}alpha=1~0
{1}image={2}
{0}[/frame]
""".format(outer, inner, image)
                lines[i] = insertion + lines[i]
                in_death = frame_start = frame_end = None
                frame_commented = in_death_commented = False
                modcount += 1
            elif in_death and "[frame]" in line:
                frame_start = i
                frame_commented = line.strip().startswith("#")
            elif in_death and "[/frame]" in line:
                frame_end = i
    # Check for duplicated attack names -- may be a result of a naive
    # boucman conversion.
    if boucmanized:
        name_pos = wmlfind("name=", WmlIterator(lines, filename))
        duplist = {}
        while name_pos is not None:
            key = lines[name_pos.lineno].strip()
            context = [x.element for x in name_pos.scopes]
            if '[attack]' in context:
                if key not in duplist:
                    duplist[key] = []
                duplist[key].append(name_pos.lineno)
            # Go to next
            name_pos = wmlfind("name=", name_pos)
        for (key, linenos) in duplist.items():
            if len(linenos) > 1:
                print('warning: duplicated attack %s at:' % key, file=sys.stderr)
                for dup in linenos:
                    print('"%s", %d: %s' % (filename, dup, key), file=sys.stderr)
    # Lift obsolete image_short and image_long tags
    expanded = """\
[attack_anim]
apply_to=attack
start_time=-150
[frame]
    duration=300
    image=%s
[/frame]
[attack_filter]
    range=%s
[/attack_filter]
[/attack_anim]\
"""
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        m = re.search(r"(\s+)image_short=(.*)", lines[i])
        if m:
            image_block = expanded.replace("\n", "\n" + m.group(1)) + "\n"
            lines[i] = m.group(1) + image_block % (m.group(2), "melee")
            modcount += 1
        r = re.search(r"(\s+)image_long=(.*)", lines[i])
        if r:
            image_block = expanded.replace("\n", "\n" + r.group(1)) + "\n"
            lines[i] = r.group(1) + image_block % (r.group(2), "ranged")
            modcount += 1
    # In [terrain], letter= to terrain=
    in_terrain = False
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        if "[terrain]" in line:
            in_terrain = True
        if "[/terrain]" in line:
            in_terrain = False
        if in_terrain:
            lines[i] = line.replace("letter", "terrain")
    # Upgrade old UNIT macro to 1.4's LOYAL_UNIT
    for i, line in enumerate(lines):
        if "no-syntax-rewrite" in line:
            break
        if '{UNIT ' in line:
            old = line.strip()
            (new, num) = re.subn(r'{UNIT +(\([^)]*\)|"[^"]*"|[^(" ][^ ]*) +(\([^)]*\)|"[^"]*"|[^(" ][^ ]*) +(\([^)]*\)|_? *"[^"]*"|_? *[^(" ][^ ]*) +([0-9]+) +([0-9]+|[^ ]*\$[^ ]*x[^ ]*|\(?{[A-Z0-9_]*X[A-Z0-9_]*}\)?) +([0-9]+|[^ ]*\$[^ ]*y[^ ]*|\(?{[A-Z0-9_]*Y[A-Z0-9_]*}\)?)}', r'{LOYAL_UNIT \4 \1 \5 \6 \2 \3}', line)
            if num > 0:
                lines[i] = new
                print('"%s", line %d: %s -> %s' % (filename, i+1, old, new.strip()))
            if num == 0:
                print('"%s", line %d: UNIT macro not converted to LOYAL_UNIT (%s)' % (filename, i+1, old))
    # More syntax transformations would go here.
    return (lines, modcount)

# Generic machinery starts here

def is_map(filename):
    "Is this file a map in either old or new style?"
    if isresource(filename) or '{' in filename or '}' in filename:
        return False
    if "map" in os.path.dirname(filename) or filename.endswith(".map"):
        return True
    try:
        with codecs.open(filename, "r", "utf8") as fp:
            lines = fp.readlines()
        has_map_content = False
        for i, line in enumerate(lines):
            lines[i] = line.rstrip("\n\r")
        w = len(lines[0])
        for line in lines:
            if len(line) != w:
                break
        else:
            has_map_content = len(lines) > 1
    except UnicodeDecodeError:
        return False
    except OSError:
        has_map_content = False
    except IndexError:
        has_map_content = False
    return has_map_content

class maptransform_error(Exception):
    "Error object to be thrown by maptransform."
    def __init__(self, infile, inline, type_):
        self.message = '"%s", line %d: %s' % (infile, inline, type_)

tagstack = [] # For tracking tag nesting

def outermap(func, inmap):
    "Apply a transformation based on neighborhood to the outermost ring."
    # Top and bottom rows
    for i in range(len(inmap[0])):
        inmap[0][i] = func(inmap[0][i])
        inmap[len(inmap)-1][i] = func(inmap[len(inmap)-1][i])
    # Leftmost and rightmost columns excluding top and bottom rows
    for i in range(1, len(inmap)-1):
        inmap[i][0] = func(inmap[i][0])
        inmap[i][len(inmap[0])-1] = func(inmap[i][len(inmap[0])-1])

def translator(filename, mapxforms, textxform, versions):
    "Apply mapxform to map lines and textxform to non-map lines."
    global tagstack
    modified = False
    mfile = []
    map_only = not filename.endswith(".cfg")
    terminator = "\n"
    with codecs.open(filename, "r", "utf8") as content:
        lines = content.readlines()
    for line in lines:
        if line.endswith("\n"):
            line = line[:-1]
        if line.endswith("\r"):
            line = line[:-1]
            if not stripcr:
                terminator = '\r\n'
        mfile.append(line)
        if "map_data" in line:
            map_only = False
    lineno = baseline = 0
    cont = False
    validate = True
    unbalanced = False
    newdata = []
    refname = None
    while mfile:
        if not map_only:
            line = mfile.pop(0)
            if verbose >= 3:
                print(line, end=terminator)
            lineno += 1
        # Check for one certain error condition
        if "{" in line and "}" in line:
            refname = line[line.find("{"):line.rfind("}")]
            # Ignore all-caps macro arguments.
            if refname == refname.upper():
                refname = None
            if 'mask=' in line and refname and not refname.endswith(".mask"):
                print('"%s", line %d: fatal error, mask file without .mask extension (%s)' %
                      (filename, lineno+1, refname), file=sys.stderr)
                sys.exit(1)
        # Exclude map_data= lines that are just 1 line without
        # continuation, or which contain {}.  The former are
        # pathological and the parse won't handle them, the latter
        # refer to map files which will be checked separately.
        if map_only or (("map_data=" in line or "mask=" in line)
                        and line.count('"') in (1, 2)
                        and "{" not in line
                        and "}" not in line
                        and not within('time')):
            outmap = []
            add_border = True
            add_usage = True
            have_header = have_delimiter = False
            maskwarn = False
            maptype = None
            if map_only:
                if filename.endswith(".mask"):
                    maptype = "mask"
                else:
                    maptype = "map"
            else:
                if "map_data" in line:
                    maptype = "map"
                elif "mask" in line:
                    maptype = "mask"
            baseline = 0
            cont = True
            if verbose >= 3:
                print("*** Entering map mode.")
            if not map_only:
                fields = line.split('"')
                if fields[1].strip():
                    mfile.insert(0, fields[1])
                if len(fields) == 3:
                    mfile.insert(1, '"')
            # Gather the map header (if any) and data lines
            while cont and mfile:
                line = mfile.pop(0)
                if verbose >= 3:
                    print(line, end=terminator)
                lineno += 1
                # This code supports ignoring comments and header lines
                if len(line) == 0 or line[0] == '#' or '=' in line:
                    if '=' in line:
                        have_header = True
                    if 'border_size' in line:
                        add_border = False
                    if "usage" in line:
                        add_usage = False
                        usage = line.split("=")[1].strip()
                        if usage == 'mask':
                            add_border = False
                            if filename.endswith(".map"):
                                print("warning: usage=mask in file with .map extension")
                        elif usage == 'map':
                            if filename.endswith(".mask"):
                                print("warning: usage=map in file with .mask extension")
                    if len(line) == 0:
                        have_delimiter = True
                    newdata.append(line + terminator)
                    continue
                if '"' in line:
                    cont = False
                    if verbose >= 3:
                        print("*** Exiting map mode.")
                    line = line.split('"')[0]
                if line:
                    if ',' in line:
                        fields = line.split(",")
                    else:
                        fields = [x for x in line]
                    outmap.append(fields)
                    if not maskwarn and maptype == 'map' and "_s" in line:
                        print('"%s", line %d: warning, fog in map file' %
                              (filename, lineno+1), file=sys.stderr)
                        maskwarn = True
            # Checking the outmap length here is a bit of a crock;
            # the one-line map we don't want to mess with is in the
            # NO_MAP macro.
            if len(outmap) == 1:
                add_border = add_usage = False
            # Deduce the map type
            if not map_only:
                if maptype == "map":
                    newdata.append("map_data=\"" + terminator)
                elif maptype == "mask":
                    newdata.append("mask=\"" + terminator)
            original = copy.deepcopy(outmap)
            for transform in mapxforms:
                for y in range(len(outmap)):
                    transform(filename, baseline, outmap, y)
            if maptype == "mask":
                add_border = False
            if add_border:
                if verbose:
                    print("adding border...")
                newdata.append("border_size=1" + terminator)
                have_header = True
                # Start by duplicating the current outermost ring
                outmap = [outmap[0]] + outmap + [outmap[-1]]
                for i in range(len(outmap)):
                    outmap[i] = [outmap[i][0]] + outmap[i] + [outmap[i][-1]]
                # Strip villages out of the edges
                outermap(lambda n: re.sub(r"\^V[a-z]+", "", n), outmap)
                # Strip keeps out of the edges
                outermap(lambda n: re.sub(r"K([a-z]+)", r"C\1", n), outmap)
                # Strip the starting positions out of the edges
                outermap(lambda n: re.sub(r"[1-9] ", r"", n), outmap)
                # Turn big trees on the edges to ordinary forest hexes
                outermap(lambda n: n.replace(r"Gg^Fet", r"Gs^Fp"), outmap)
                modified = True
            if add_usage:
                newdata.append("usage=" + maptype + terminator)
                have_header = True
                modified = True
            if have_header and not have_delimiter:
                newdata.append(terminator)
                modified = True
            for y in range(len(outmap)):
                newdata.append(",".join(outmap[y]) + terminator)
                if not modified and original[y] != outmap[y]:
                    modified = True
            # All lines of the map are processed, add the appropriate trailer
            if not map_only:
                newdata.append("\"" + terminator)
        elif "map_data=" in line and ("{" in line or "}" in line):
            newline = line
            refre = re.compile(r"\{@?([^A-Z].*)\}").search(line)
            if refre:
                mapfile = refre.group(1)
                if not mapfile.endswith(".map") and is_map(mapfile):
                    newline = newline.replace(mapfile, mapfile + ".map")
            newdata.append(newline + terminator)
            if newline != line:
                modified = True
                if verbose > 0:
                    print('wmllint: "%s", line %d: %s -> %s.' %
                          (filename, lineno, line, newline), file=sys.stderr)
        elif "map_data=" in line and line.count('"') > 1:
            print('wmllint: "%s", line %d: one-line map.' %
                  (filename, lineno), file.sys.stderr)
            newdata.append(line + terminator)
        else:
            # Handle text (non-map) lines
            newline = textxform(filename, lineno, line, versions)
            newdata.append(newline + terminator)
            if newline != line:
                modified = True
            # Now do warnings based on the state of the tag stack.
            if not unbalanced:
                fields = newline.split("#")
                trimmed = fields[0]
                destringed = re.sub('"[^"]*"', '', trimmed) # Ignore string literals
                comment = ""
                if len(fields) > 1:
                    comment = fields[1]
                for instance in re.finditer(r"\[\/?\+?([a-z][a-z_]*[a-z])\]", destringed):
                    tag = instance.group(1)
                    attributes = []
                    closer = instance.group(0)[1] == '/'
                    if not closer:
                        tagstack.append((tag, {}))
                    else:
                        if len(tagstack) == 0:
                            print('"%s", line %d: closer [/%s] with tag stack empty.' %
                                  (filename, lineno+1, tag))
                        elif tagstack[-1][0] != tag:
                            print('"%s", line %d: unbalanced [%s] closed with [/%s].' %
                                  (filename, lineno+1, tagstack[-1][0], tag))
                        else:
                            if validate:
                                validate_on_pop(tagstack, tag, filename, lineno)
                            tagstack.pop()
                if tagstack:
                    for instance in re.finditer(r'([a-z][a-z_]*[a-z])\s*=(\w+|"[^"]*")', trimmed):
                        attribute = instance.group(1)
                        value = instance.group(2)
                        tagstack[-1][1][attribute] = value
                    if validate:
                        validate_stack(tagstack, filename, lineno)
            if "wmllint: validate-on" in comment:
                validate = True
            if "wmllint: validate-off" in comment:
                validate = False
            if "wmllint: unbalanced-on" in comment:
                unbalanced = True
            if "wmllint: unbalanced-off" in comment:
                unbalanced = False
    # It's an error if the tag stack is nonempty at the end of any file:
    if tagstack:
        print('"%s", line %d: tag stack nonempty (%s) at end of file.' %
              (filename, lineno, tagstack), file=sys.stderr)
    tagstack = []
    if iswml(filename):
        # Perform semantic sanity checks
        (newdata, modified1) = sanity_check(filename, newdata)
        # OK, now perform WML rewrites
        (newdata, modified2) = hack_syntax(filename, newdata)
        # Run everything together
        filetext = "".join(newdata)
        modified |= modified1 or modified2
        transformed = filetext
        if upconvert:
            # WML syntax changed in 1.3.5.  The transformation cannot
            # conveniently be done line-by-line.
            transformed = re.sub(r"(if]|while])\s*\[or]([\w\W]*?)\[/or]\s*",
                             r"\1\2", filetext);
            modified |= (transformed != filetext)
    else:
        # Map or mask -- just run everything together
        transformed = "".join(newdata)
    # Simple check for unbalanced macro calls
    unclosed = None
    linecount = 1
    startline = None
    depth = quotecount = 0
    for i, char in enumerate(transformed):
        if char == '\n':
            linecount += 1
        elif char == '{':
            if depth == 0:
                unclosed = startline = linecount
                quotecount = 0
            depth += 1
        elif char == '"':
            quotecount += 1
        elif char == '}':
            depth -= 1
            if depth == 0:
                unclosed = None
                if quotecount % 2:
                    print('"%s", line %d: unbalanced quote.' %
                          (filename, startline), file=sys.stderr)
    if unclosed:
        print('"%s", line %d: unbalanced {.' %
              (filename, unclosed), file=sys.stderr)
    # Return None if the transformation functions made no changes.
    if modified:
        return transformed
    else:
        return None

ignore = (".tgz", ".png", ".jpg", "-bak")

def interesting(fn):
    "Is a file interesting for conversion purposes?"
    return fn.endswith(".cfg") or fn.endswith(".map") \
        or ("maps" in fn and fn[-4:] not in ignore) \
        or is_map(fn)

def allcfgfiles(dir_):
    "Get the names of all interesting files under dir."
    datafiles = []
    if not os.path.isdir(dir_):
        if interesting(dir_):
            if not os.path.exists(dir_):
                print("wmllint: %s does not exist" % dir_, file=sys.stderr)
            else:
                datafiles.append(dir_)
    else:
        for root, dirs, files in os.walk(dir_):
            if vcdir in dirs:
                dirs.remove(vcdir)
            for name in files:
                if interesting(os.path.join(root, name)):
                    datafiles.append(os.path.join(root, name))
    datafiles.sort() # So diffs for same campaigns will cluster in reports
    return [os.path.normpath(elem) for elem in datafiles]

def help():
        print("""\
Usage: wmllint [options] [dir]
    Included because wmllint has dropped support for pre-1.4 Wesnoth WML.
    Use it to convert 1.0 and 1.2 material to "1.4", then use regular wmllint
    to port it to a modern Battle of Wesnoth version.  Can also be used to
    validate 1.4 add-ons against remnants of archaic WML.

    Takes any number of directories as arguments.  Each directory is converted.
    If no directories are specified, acts on the current directory.
    Options may be any of these:
    -h, --help                 Emit this help message and quit.
    -d, --dryrun               List changes but don't perform them.
    -n, --nolift               Don't perform version-lifting.
    -o, --oldversion           Specify version to begin with.
                               Note: wmllint-1.4 already defaults to "older",
                               which covers everything before 1.3.1.
    -v, --verbose              -v        lists changes.
                               -v -v     names each file before it's processed.
                               -v -v -v  shows verbose parse details.
    -c, --clean                Clean up -bak files.
    -D, --diff                 Display diffs between converted and unconverted files.
    -r, --revert               Revert the conversion from the -bak files.
    -s, --stripcr              Convert DOS-style CR/LF to Unix-style LF.
                               Note: does not work on Windows, due to Python
                               defaulting to universal newlines support there.
    --future                   Enable experimental WML conversions.
    More about wmllint can be found in the introduction in the file itself, and
    at https://wiki.wesnoth.org/Maintenance_tools.""", file=sys.stderr)

if __name__ == '__main__':
    global versions
    try:
        (options, arguments) = getopt.getopt(sys.argv[1:], "cdDfhno:rsv", [
            "clean",
            "diffs",
            "dryrun",
            "future",
            "help",
            "nolift",
            "oldversion=",
            "revert",
            "stripcr",
            "verbose",
            ])
    except getopt.GetoptError:
        help()
        sys.exit(1)
    clean = False
    diffs = False
    dryrun = False
    future = False
    oldversion = 'older'
    revert = False
    stripcr = False
    upconvert = True
    verbose = 0
    for (switch, val) in options:
        if switch in ('-h', '--help'):
            help()
            sys.exit(0)
        elif switch in ('-c', '--clean'):
            clean = True
        elif switch in ('-d', '--dryrun'):
            dryrun = True
            verbose = max(1, verbose)
        elif switch in ('-D', '--diffs'):
            diffs = True
        elif switch in ('-f', '--future'):
            future = True
        elif switch in ('-n', '--nolift'):
            upconvert = False
        elif switch in ('-o', '--oldversion'):
            oldversion = val
        elif switch in ('-r', '--revert'):
            revert = True
        elif switch in ('-s', '--stripcr'):
            stripcr = True
        elif switch in ('-v', '--verbose'):
            verbose += 1
    if clean and revert:
        print("wmllint: can't do clean and revert together.", file=sys.stderr)
        sys.exit(1)

    # Compute the series of version upgrades to perform, and describe it.
    versions = sorted([key for key in filemoves.keys()])
    # Relies on 'older' sorting before trunk
    versions = [versions[-2]] + versions[:-2] + [versions[-1]] # Move 'older' to front
    if oldversion in versions:
        versions = versions[versions.index(oldversion):]
    else:
        print("wmllint: unrecognized version.", file=sys.stderr)
        sys.exit(1)
    if not dryrun and not clean and not revert and not diffs and len(versions) > 1:
        explain = "Upgrades for:"
        for i in range(len(versions)-1):
            explain += " %s -> %s," % (versions[i],  versions[i+1])
        print(explain[:-1] + ".")
    fileconversions = [filemoves[x] for x in versions[:-1]]

    def hasdigit(str_):
        for c in str_:
            if c in "0123456789":
                return True
        return False

    def texttransform(filename, lineno, line, versions):
        "Resource-name transformation on text lines."
        if not upconvert:
            return line
        transformed = line
        # First, do resource-file moves
        if "wmllint: noconvert" not in line:
            for step in fileconversions:
                for (old, new) in step:
                    transformed = old.sub(new, transformed)
        # Handle terrain_liked=, terrain=, valid_terrain=, letter=
        spaceless = transformed.replace(" ", "").replace("\t", "")
        if spaceless and spaceless[0] != "#" and ("terrain_liked=" in spaceless or "terrain=" in spaceless or 'letter=' in spaceless) and "wmllint:ignore" not in spaceless:
            (key, pre, value, post) = parse_attribute(transformed)
            # We have to cope with the following cases...
            # Old style:
            #    terrain_liked=ghM
            #    terrain_liked=BEITU
            #    valid_terrain=gfh
            #    terrain=AaBbDeLptUVvYZ
            #    terrain=r
            #    terrain={LETTERS}
            #    terrain=""
            #    terrain=s,c,w,k
            # New style:
            #    terrain=Mm
            #    terrain=Gs^Fp
            #    terrain=Hh, Gg^Vh, Mm
            # The sticky part is that, while it never happens in the current
            # corpus, terrain=Mm (capital letter followed by small) could be
            # interpreted either way.
            #
            # There are some unambiguous tests:
            oldstyle = (len(value) == 1 or len(value) > 6) and not ',' in value
            newstyle = len(value) > 1 \
                       and value[0].isupper() and value[1].islower() \
                       and (',' in value \
                            or len(value) == 2 \
                            or (len(value) >= 3 and value[2] == "^"))
            # See maptransform1() for explanation of this ugly hack.
            oldstyle = oldstyle or lock_terrain_coding == "oldstyle"
            newstyle = newstyle or lock_terrain_coding == "newstyle"
            # Maybe we lose...
            if not oldstyle and not newstyle:
                print('"%s", line %d: leaving ambiguous terrain value %s alone.' %
                      (filename, lineno, value))
            else:
                if oldstyle and not lock_terrain_coding == "newstyle":
                    # 1.2.x to 1.3.2 conversions
                    newterrains = ""
                    inmacro = False
                    for c in value:
                        if not inmacro:
                            if c == '{':
                                inmacro = True
                                newterrains += c
                            elif c == ',':
                                pass
                            elif c.isspace():
                                newterrains += c
                            elif c in conversion1:
                                newterrains += conversion1[c] + ","
                            else:
                                print("%s, line %d: custom terrain %s ignored." %
                                      (filename, lineno+1, c))
                        else: # inmacro == True
                            if c == '}':
                                inmacro = False
                            newterrains += c
                    if newterrains.endswith(","):
                        newterrains = newterrains[:-1]
                    transformed = pre + newterrains + post
                if newstyle:
                    if len(value) == 2:
                        # 1.3.1 to 1.3.2 conversion
                        for (old, new) in conversion2.items():
                            transformed = old.sub(new, transformed)
        # Check for things marked translated that aren't strings
        if "_" in transformed and not "wmllint: ignore" in transformed:
            m = re.search(r'[=(]\s*_\s+("?)', transformed)
            if m and not m.group(1):
                msg = '"%s", line %d: translatability mark before non-string' % \
                      (filename, lineno)
                print(msg, file=sys.stderr)
        # Perform unconditional line changes
        for (old, new) in linechanges:
            transformed = transformed.replace(old, new)
        # Report the changes
        if verbose > 0 and transformed != line:
            msg = "%s, line %d: %s -> %s" % \
                  (filename, lineno, line.strip(), transformed.strip())
            print(msg)
        return transformed

    if upconvert and "1.3.1" in versions and "older" not in versions:
        maptransforms = [maptransform2]
    else:
        maptransforms = [maptransform1, maptransform2]

    # In certain situations, Windows' command prompt appends a double quote
    # to the command line parameters. The first block takes care of this issue.
    # Also, Windows does not expand wildcards. The second turns on globbing.
    if arguments and sys.platform == 'win32':
        wildcard = False
        for i,arg in enumerate(arguments):
            if arg.endswith('"'):
                arguments[i] = arg[:-1]
            if '*' in arg:
                wildcard = True
        if wildcard:
            from glob import glob
            for arg in arguments:
                wildarg = glob(arg)
                arguments.remove(arg)
                for newarg in wildarg:
                    arguments.append(newarg)

    if not arguments:
        arguments = ["."]

    for dir_ in arguments:
        ofp = None
        if "older" in versions:
            lock_terrain_coding = None
        else:
            lock_terrain_coding = "newstyle"
        for fn in allcfgfiles(dir_):
            if verbose >= 2:
                print(fn + ":")
            backup = fn + "-bak"
            if clean or revert:
                # Do housekeeping
                if os.path.exists(backup):
                    if clean:
                        print("wmllint: removing %s" % backup)
                        if not dryrun:
                            os.remove(backup)
                    elif revert:
                        print("wmllint: reverting %s" % backup)
                        if not dryrun:
                            if sys.platform == 'win32':
                                os.remove(fn)
                            os.rename(backup, fn)
            elif diffs:
                # Display diffs
                if os.path.exists(backup):
                    fromdate = time.ctime(os.stat(backup).st_mtime)
                    todate = time.ctime(os.stat(fn).st_mtime)
                    with codecs.open(backup, 'U', "utf8") as fromlines, \
                         codecs.open(fn, 'U', "utf8") as tolines:
                        diff = difflib.unified_diff(fromlines.readlines(),
                                                    tolines.readlines(),
                                                    backup, fn, fromdate,
                                                    todate, n=3)
                    sys.stdout.writelines(diff)
            else:
                # Do file conversions
                try:
                    changed = translator(fn, maptransforms, texttransform, versions)
                    if changed:
                        print("wmllint: converting", fn)
                        if not dryrun:
                            if sys.platform == 'win32' and os.path.exists(backup):
                                os.remove(backup)
                            os.rename(fn, backup)
                            with codecs.open(fn, "w", "utf8") as ofp:
                                ofp.write(changed)
                except maptransform_error as e:
                    print("wmllint: " + e.message, file=sys.stderr)
                except:
                    print("wmllint: internal error on %s" % fn, file=sys.stderr)
                    (exc_type, exc_value, exc_traceback) = sys.exc_info()
                    raise exc_type(exc_value).with_traceback(exc_traceback)
            # Time for map and main file renames
            # FIXME: We should make some effort to rename mask files.
            if not revert and not diffs:
                if not fn.endswith(".map") and not fn.endswith(".mask") and is_map(fn):
                    print('wmllint: renaming "%s" to "%s"' % (fn, fn + ".map"))
                    if not dryrun:
                        os.rename(fn, fn + ".map")
                elif fn in is_main and os.path.isdir(fn.replace('.cfg', '')):
                    main = fn.replace('.cfg', '/_main.cfg')
                    if os.path.exists(main):
                        print('wmllint: both "%s" and "%s" topfiles exist' % (fn, main))
                    else:
                        print('wmllint: renaming "%s" to "%s"' % (fn, main))
                        if not dryrun:
                            os.rename(fn, main)
                            if os.path.exists(backup):
                                os.rename(backup, main + '-bak')
    # Constency-check everything we got from the file scans
    consistency_check()

# wmllint ends here