"""
Created on 13 Jan 2011

@author: charles
"""
import inspect, re, traceback, numbers
from math import trunc

from ebook_converter import human_readable
from ebook_converter.constants import DEBUG
from ebook_converter.ebooks.metadata import title_sort
from ebook_converter.utils.config import tweaks
from ebook_converter.utils.titlecase import titlecase
from ebook_converter.utils.icu import capitalize, strcmp, sort_key
from ebook_converter.utils.date import parse_date, format_date, now, UNDEFINED_DATE
from ebook_converter.utils.localization import calibre_langcode_to_name, canonicalize_lang
from ebook_converter.polyglot.builtins import iteritems, itervalues


__license__ = 'GPL v3'
__copyright__ = '2010, Kovid Goyal <kovid@kovidgoyal.net>'
__docformat__ = 'restructuredtext en'


class FormatterFunctions(object):

    error_function_body = ('def evaluate(self, formatter, kwargs, mi, locals):\n'
                       '\treturn "' +
                            _('Duplicate user function name {0}. '
                              'Change the name or ensure that the functions are identical') + '"')

    def __init__(self):
        self._builtins = {}
        self._functions = {}
        self._functions_from_library = {}

    def register_builtin(self, func_class):
        if not isinstance(func_class, FormatterFunction):
            raise ValueError('Class %s is not an instance of FormatterFunction'%(
                                    func_class.__class__.__name__))
        name = func_class.name
        if name in self._functions:
            raise ValueError('Name %s already used'%name)
        self._builtins[name] = func_class
        self._functions[name] = func_class
        for a in func_class.aliases:
            self._functions[a] = func_class

    def _register_function(self, func_class, replace=False):
        if not isinstance(func_class, FormatterFunction):
            raise ValueError('Class %s is not an instance of FormatterFunction'%(
                                    func_class.__class__.__name__))
        name = func_class.name
        if not replace and name in self._functions:
            raise ValueError('Name %s already used'%name)
        self._functions[name] = func_class

    def register_functions(self, library_uuid, funcs):
        self._functions_from_library[library_uuid] = funcs
        self._register_functions()

    def _register_functions(self):
        for compiled_funcs in itervalues(self._functions_from_library):
            for cls in compiled_funcs:
                f = self._functions.get(cls.name, None)
                replace = False
                if f is not None:
                    existing_body = f.program_text
                    new_body = cls.program_text
                    if new_body != existing_body:
                        # Change the body of the template function to one that will
                        # return an error message. Also change the arg count to
                        # -1 (variable) to avoid template compilation errors
                        replace = True
                        func = [cls.name, '', -1, self.error_function_body.format(cls.name)]
                        cls = compile_user_function(*func)
                    else:
                        continue
                formatter_functions()._register_function(cls, replace=replace)

    def unregister_functions(self, library_uuid):
        if library_uuid in self._functions_from_library:
            for cls in self._functions_from_library[library_uuid]:
                self._functions.pop(cls.name, None)
            self._functions_from_library.pop(library_uuid)
            self._register_functions()

    def get_builtins(self):
        return self._builtins

    def get_builtins_and_aliases(self):
        res = {}
        for f in itervalues(self._builtins):
            res[f.name] = f
            for a in f.aliases:
                res[a] = f
        return res

    def get_functions(self):
        return self._functions

    def reset_to_builtins(self):
        self._functions = {}
        for n,c in self._builtins.items():
            self._functions[n] = c
            for a in c.aliases:
                self._functions[a] = c


_ff = FormatterFunctions()


def formatter_functions():
    global _ff
    return _ff


class FormatterFunction(object):

    doc = _('No documentation provided')
    name = 'no name provided'
    category = 'Unknown'
    arg_count = 0
    aliases = []

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        raise NotImplementedError()

    def eval_(self, formatter, kwargs, mi, locals, *args):
        ret = self.evaluate(formatter, kwargs, mi, locals, *args)
        if isinstance(ret, (bytes, str)):
            return ret
        if isinstance(ret, list):
            return ','.join(ret)
        if isinstance(ret, (numbers.Number, bool)):
            return str(ret)


class BuiltinFormatterFunction(FormatterFunction):

    def __init__(self):
        formatter_functions().register_builtin(self)
        eval_func = inspect.getmembers(self.__class__,
                        lambda x: inspect.ismethod(x) and x.__name__ == 'evaluate')
        try:
            lines = [l[4:] for l in inspect.getsourcelines(eval_func[0][1])[0]]
        except:
            lines = []
        self.program_text = ''.join(lines)


class BuiltinStrcmp(BuiltinFormatterFunction):
    name = 'strcmp'
    arg_count = 5
    category = 'Relational'
    __doc__ = doc = _('strcmp(x, y, lt, eq, gt) -- does a case-insensitive comparison of x '
            'and y as strings. Returns lt if x < y. Returns eq if x == y. '
            'Otherwise returns gt.')

    def evaluate(self, formatter, kwargs, mi, locals, x, y, lt, eq, gt):
        v = strcmp(x, y)
        if v < 0:
            return lt
        if v == 0:
            return eq
        return gt


class BuiltinCmp(BuiltinFormatterFunction):
    name = 'cmp'
    category = 'Relational'
    arg_count = 5
    __doc__ = doc =   _('cmp(x, y, lt, eq, gt) -- compares x and y after converting both to '
            'numbers. Returns lt if x < y. Returns eq if x == y. Otherwise returns gt.')

    def evaluate(self, formatter, kwargs, mi, locals, x, y, lt, eq, gt):
        x = float(x if x and x != 'None' else 0)
        y = float(y if y and y != 'None' else 0)
        if x < y:
            return lt
        if x == y:
            return eq
        return gt


class BuiltinFirstMatchingCmp(BuiltinFormatterFunction):
    name = 'first_matching_cmp'
    category = 'Relational'
    arg_count = -1
    __doc__ = doc =   _('first_matching_cmp(val, cmp1, result1, cmp2, r2, ..., else_result) -- '
            'compares "val < cmpN" in sequence, returning resultN for '
            'the first comparison that succeeds. Returns else_result '
            'if no comparison succeeds. Example: '
            'first_matching_cmp(10,5,"small",10,"middle",15,"large","giant") '
            'returns "large". The same example with a first value of 16 returns "giant".')

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        if (len(args) % 2) != 0:
            raise ValueError(_('first_matching_cmp requires an even number of arguments'))
        val = float(args[0] if args[0] and args[0] != 'None' else 0)
        for i in range(1, len(args) - 1, 2):
            c = float(args[i] if args[i] and args[i] != 'None' else 0)
            if val < c:
                return args[i+1]
        return args[len(args)-1]


class BuiltinStrcat(BuiltinFormatterFunction):
    name = 'strcat'
    arg_count = -1
    category = 'String manipulation'
    __doc__ = doc = _('strcat(a, b, ...) -- can take any number of arguments. Returns a '
            'string formed by concatenating all the arguments')

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        i = 0
        res = ''
        for i in range(0, len(args)):
            res += args[i]
        return res


class BuiltinStrlen(BuiltinFormatterFunction):
    name = 'strlen'
    arg_count = 1
    category = 'String manipulation'
    __doc__ = doc = _('strlen(a) -- Returns the length of the string passed as '
            'the argument')

    def evaluate(self, formatter, kwargs, mi, locals, a):
        try:
            return len(a)
        except:
            return -1


class BuiltinAdd(BuiltinFormatterFunction):
    name = 'add'
    arg_count = 2
    category = 'Arithmetic'
    __doc__ = doc = _('add(x, y) -- returns x + y. Throws an exception if either x or y are not numbers.')

    def evaluate(self, formatter, kwargs, mi, locals, x, y):
        x = float(x if x and x != 'None' else 0)
        y = float(y if y and y != 'None' else 0)
        return str(x + y)


class BuiltinSubtract(BuiltinFormatterFunction):
    name = 'subtract'
    arg_count = 2
    category = 'Arithmetic'
    __doc__ = doc = _('subtract(x, y) -- returns x - y. Throws an exception if either x or y are not numbers.')

    def evaluate(self, formatter, kwargs, mi, locals, x, y):
        x = float(x if x and x != 'None' else 0)
        y = float(y if y and y != 'None' else 0)
        return str(x - y)


class BuiltinMultiply(BuiltinFormatterFunction):
    name = 'multiply'
    arg_count = 2
    category = 'Arithmetic'
    __doc__ = doc = _('multiply(x, y) -- returns x * y. Throws an exception if either x or y are not numbers.')

    def evaluate(self, formatter, kwargs, mi, locals, x, y):
        x = float(x if x and x != 'None' else 0)
        y = float(y if y and y != 'None' else 0)
        return str(x * y)


class BuiltinDivide(BuiltinFormatterFunction):
    name = 'divide'
    arg_count = 2
    category = 'Arithmetic'
    __doc__ = doc = _('divide(x, y) -- returns x / y. Throws an exception if either x or y are not numbers.')

    def evaluate(self, formatter, kwargs, mi, locals, x, y):
        x = float(x if x and x != 'None' else 0)
        y = float(y if y and y != 'None' else 0)
        return str(x / y)


class BuiltinTemplate(BuiltinFormatterFunction):
    name = 'template'
    arg_count = 1
    category = 'Recursion'

    __doc__ = doc = _('template(x) -- evaluates x as a template. The evaluation is done '
            'in its own context, meaning that variables are not shared between '
            'the caller and the template evaluation. Because the { and } '
            'characters are special, you must use [[ for the { character and '
            ']] for the } character; they are converted automatically. '
            'For example, template(\'[[title_sort]]\') will evaluate the '
            'template {title_sort} and return its value. Note also that '
            'prefixes and suffixes (the `|prefix|suffix` syntax) cannot be '
            'used in the argument to this function when using template program mode.')

    def evaluate(self, formatter, kwargs, mi, locals, template):
        template = template.replace('[[', '{').replace(']]', '}')
        return formatter.__class__().safe_format(template, kwargs, 'TEMPLATE', mi)


class BuiltinEval(BuiltinFormatterFunction):
    name = 'eval'
    arg_count = 1
    category = 'Recursion'
    __doc__ = doc = _('eval(template) -- evaluates the template, passing the local '
            'variables (those \'assign\'ed to) instead of the book metadata. '
            ' This permits using the template processor to construct complex '
            'results from local variables. Because the { and } '
            'characters are special, you must use [[ for the { character and '
            ']] for the } character; they are converted automatically. '
            'Note also that prefixes and suffixes (the `|prefix|suffix` syntax) '
            'cannot be used in the argument to this function when using '
            'template program mode.')

    def evaluate(self, formatter, kwargs, mi, locals, template):
        from ebook_converter.utils.formatter import EvalFormatter
        template = template.replace('[[', '{').replace(']]', '}')
        return EvalFormatter().safe_format(template, locals, 'EVAL', None)


class BuiltinAssign(BuiltinFormatterFunction):
    name = 'assign'
    arg_count = 2
    category = 'Other'
    __doc__ = doc = _('assign(id, val) -- assigns val to id, then returns val. '
            'id must be an identifier, not an expression')

    def evaluate(self, formatter, kwargs, mi, locals, target, value):
        locals[target] = value
        return value


class BuiltinPrint(BuiltinFormatterFunction):
    name = 'print'
    arg_count = -1
    category = 'Other'
    __doc__ = doc = _('print(a, b, ...) -- prints the arguments to standard output. '
            'Unless you start calibre from the command line (calibre-debug -g), '
            'the output will go to a black hole.')

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        print(args)
        return ''


class BuiltinField(BuiltinFormatterFunction):
    name = 'field'
    arg_count = 1
    category = 'Get values from metadata'
    __doc__ = doc = _('field(name) -- returns the metadata field named by name')

    def evaluate(self, formatter, kwargs, mi, locals, name):
        return formatter.get_value(name, [], kwargs)


class BuiltinRawField(BuiltinFormatterFunction):
    name = 'raw_field'
    arg_count = 1
    category = 'Get values from metadata'
    __doc__ = doc = _('raw_field(name) -- returns the metadata field named by name '
            'without applying any formatting.')

    def evaluate(self, formatter, kwargs, mi, locals, name):
        res = getattr(mi, name, None)
        if isinstance(res, list):
            fm = mi.metadata_for_field(name)
            if fm is None:
                return ', '.join(res)
            return fm['is_multiple']['list_to_ui'].join(res)
        return str(res)


class BuiltinRawList(BuiltinFormatterFunction):
    name = 'raw_list'
    arg_count = 2
    category = 'Get values from metadata'
    __doc__ = doc = _('raw_list(name, separator) -- returns the metadata list '
            'named by name without applying any formatting or sorting and '
            'with items separated by separator.')

    def evaluate(self, formatter, kwargs, mi, locals, name, separator):
        res = getattr(mi, name, None)
        if not isinstance(res, list):
            return "%s is not a list" % name
        return separator.join(res)


class BuiltinSubstr(BuiltinFormatterFunction):
    name = 'substr'
    arg_count = 3
    category = 'String manipulation'
    __doc__ = doc = _('substr(str, start, end) -- returns the start\'th through the end\'th '
            'characters of str. The first character in str is the zero\'th '
            'character. If end is negative, then it indicates that many '
            'characters counting from the right. If end is zero, then it '
            'indicates the last character. For example, substr(\'12345\', 1, 0) '
            'returns \'2345\', and substr(\'12345\', 1, -1) returns \'234\'.')

    def evaluate(self, formatter, kwargs, mi, locals, str_, start_, end_):
        return str_[int(start_): len(str_) if int(end_) == 0 else int(end_)]


class BuiltinLookup(BuiltinFormatterFunction):
    name = 'lookup'
    arg_count = -1
    category = 'Iterating over values'
    __doc__ = doc = _('lookup(val, pattern, field, pattern, field, ..., else_field) -- '
            'like switch, except the arguments are field (metadata) names, not '
            'text. The value of the appropriate field will be fetched and used. '
            'Note that because composite columns are fields, you can use this '
            'function in one composite field to use the value of some other '
            'composite field. This is extremely useful when constructing '
            'variable save paths')

    def evaluate(self, formatter, kwargs, mi, locals, val, *args):
        if len(args) == 2:  # here for backwards compatibility
            if val:
                return formatter.vformat('{'+args[0].strip()+'}', [], kwargs)
            else:
                return formatter.vformat('{'+args[1].strip()+'}', [], kwargs)
        if (len(args) % 2) != 1:
            raise ValueError(_('lookup requires either 2 or an odd number of arguments'))
        i = 0
        while i < len(args):
            if i + 1 >= len(args):
                return formatter.vformat('{' + args[i].strip() + '}', [], kwargs)
            if re.search(args[i], val, flags=re.I):
                return formatter.vformat('{'+args[i+1].strip() + '}', [], kwargs)
            i += 2


class BuiltinTest(BuiltinFormatterFunction):
    name = 'test'
    arg_count = 3
    category = 'If-then-else'
    __doc__ = doc = _('test(val, text if not empty, text if empty) -- return `text if not '
            'empty` if val is not empty, otherwise return `text if empty`')

    def evaluate(self, formatter, kwargs, mi, locals, val, value_if_set, value_not_set):
        if val:
            return value_if_set
        else:
            return value_not_set


class BuiltinContains(BuiltinFormatterFunction):
    name = 'contains'
    arg_count = 4
    category = 'If-then-else'
    __doc__ = doc = _('contains(val, pattern, text if match, text if not match) -- checks '
            'if val contains matches for the regular expression `pattern`. '
            'Returns `text if match` if matches are found, otherwise it returns '
            '`text if no match`')

    def evaluate(self, formatter, kwargs, mi, locals,
                 val, test, value_if_present, value_if_not):
        if re.search(test, val, flags=re.I):
            return value_if_present
        else:
            return value_if_not


class BuiltinSwitch(BuiltinFormatterFunction):
    name = 'switch'
    arg_count = -1
    category = 'Iterating over values'
    __doc__ = doc = _('switch(val, pattern, value, pattern, value, ..., else_value) -- '
            'for each `pattern, value` pair, checks if `val` matches '
            'the regular expression `pattern` and if so, returns that '
            '`value`. If no pattern matches, then `else_value` is returned. '
            'You can have as many `pattern, value` pairs as you want')

    def evaluate(self, formatter, kwargs, mi, locals, val, *args):
        if (len(args) % 2) != 1:
            raise ValueError(_('switch requires an odd number of arguments'))
        i = 0
        while i < len(args):
            if i + 1 >= len(args):
                return args[i]
            if re.search(args[i], val, flags=re.I):
                return args[i+1]
            i += 2


class BuiltinStrcatMax(BuiltinFormatterFunction):
    name = 'strcat_max'
    arg_count = -1
    category = 'String manipulation'
    __doc__ = doc = _('strcat_max(max, string1, prefix2, string2, ...) -- '
            'Returns a string formed by concatenating the arguments. The '
            'returned value is initialized to string1. `Prefix, string` '
            'pairs are added to the end of the value as long as the '
            'resulting string length is less than `max`. String1 is returned '
            'even if string1 is longer than max. You can pass as many '
            '`prefix, string` pairs as you wish.')

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        if len(args) < 2:
            raise ValueError(_('strcat_max requires 2 or more arguments'))
        if (len(args) % 2) != 0:
            raise ValueError(_('strcat_max requires an even number of arguments'))
        try:
            max = int(args[0])
        except:
            raise ValueError(_('first argument to strcat_max must be an integer'))

        i = 2
        result = args[1]
        try:
            while i < len(args):
                if (len(result) + len(args[i]) + len(args[i+1])) > max:
                    break
                result = result + args[i] + args[i+1]
                i += 2
        except:
            pass
        return result.strip()


class BuiltinInList(BuiltinFormatterFunction):
    name = 'in_list'
    arg_count = -1
    category = 'List lookup'
    __doc__ = doc = _('in_list(val, separator, pattern, found_val, ..., not_found_val) -- '
            'treat val as a list of items separated by separator, '
            'evaluating the pattern against each value in the list. If the '
            'pattern matches a value, return found_val, otherwise return '
            'not_found_val. The pattern and found_value can be repeated as '
            'many times as desired, permitting returning different values '
            'depending on the search. The patterns are checked in order. The '
            'first match is returned.')

    def evaluate(self, formatter, kwargs, mi, locals, val, sep, *args):
        if (len(args) % 2) != 1:
            raise ValueError(_('in_list requires an odd number of arguments'))
        l = [v.strip() for v in val.split(sep) if v.strip()]
        i = 0
        while i < len(args):
            if i + 1 >= len(args):
                return args[i]
            sf = args[i]
            fv = args[i+1]
            if l:
                for v in l:
                    if re.search(sf, v, flags=re.I):
                        return fv
            i += 2


class BuiltinStrInList(BuiltinFormatterFunction):
    name = 'str_in_list'
    arg_count = -1
    category = 'List lookup'
    __doc__ = doc = _('str_in_list(val, separator, string, found_val, ..., not_found_val) -- '
            'treat val as a list of items separated by separator, '
            'comparing the string against each value in the list. If the '
            'string matches a value (ignoring case) then return found_val, otherwise return '
            'not_found_val. If the string contains separators, then it is '
            'also treated as a list and each value is checked. The string and '
            'found_value can be repeated as many times as desired, permitting '
            'returning different values depending on the search. The strings are '
            'checked in order. The first match is returned.')

    def evaluate(self, formatter, kwargs, mi, locals, val, sep, *args):
        if (len(args) % 2) != 1:
            raise ValueError(_('str_in_list requires an odd number of arguments'))
        l = [v.strip() for v in val.split(sep) if v.strip()]
        i = 0
        while i < len(args):
            if i + 1 >= len(args):
                return args[i]
            sf = args[i]
            fv = args[i+1]
            c = [v.strip() for v in sf.split(sep) if v.strip()]
            if l:
                for v in l:
                    for t in c:
                        if strcmp(t, v) == 0:
                            return fv
            i += 2


class BuiltinIdentifierInList(BuiltinFormatterFunction):
    name = 'identifier_in_list'
    arg_count = 4
    category = 'List lookup'
    __doc__ = doc = _('identifier_in_list(val, id, found_val, not_found_val) -- '
            'treat val as a list of identifiers separated by commas, '
            'comparing the string against each value in the list. An identifier '
            'has the format "identifier:value". The id parameter should be '
            'either "id" or "id:regexp". The first case matches if there is any '
            'identifier with that id. The second case matches if the regexp '
            'matches the identifier\'s value. If there is a match, '
            'return found_val, otherwise return not_found_val.')

    def evaluate(self, formatter, kwargs, mi, locals, val, ident, fv, nfv):
        l = [v.strip() for v in val.split(',') if v.strip()]
        (id, _, regexp) = ident.partition(':')
        if not id:
            return nfv
        id += ':'
        if l:
            for v in l:
                if v.startswith(id):
                    if not regexp or re.search(regexp, v[len(id):], flags=re.I):
                        return fv
        return nfv


class BuiltinRe(BuiltinFormatterFunction):
    name = 're'
    arg_count = 3
    category = 'String manipulation'
    __doc__ = doc = _('re(val, pattern, replacement) -- return val after applying '
            'the regular expression. All instances of `pattern` are replaced '
            'with `replacement`. As in all of calibre, these are '
            'Python-compatible regular expressions')

    def evaluate(self, formatter, kwargs, mi, locals, val, pattern, replacement):
        return re.sub(pattern, replacement, val, flags=re.I)


class BuiltinReGroup(BuiltinFormatterFunction):
    name = 're_group'
    arg_count = -1
    category = 'String manipulation'
    __doc__ = doc = _('re_group(val, pattern, template_for_group_1, for_group_2, ...) -- '
            'return a string made by applying the regular expression pattern '
            'to the val and replacing each matched instance with the string '
            'computed by replacing each matched group by the value returned '
            'by the corresponding template. The original matched value for the '
            'group is available as $. In template program mode, like for '
            'the template and the eval functions, you use [[ for { and ]] for }.'
            ' The following example in template program mode looks for series '
            'with more than one word and uppercases the first word: '
            "{series:'re_group($, \"(\\S* )(.*)\", \"[[$:uppercase()]]\", \"[[$]]\")'}")

    def evaluate(self, formatter, kwargs, mi, locals, val, pattern, *args):
        from ebook_converter.utils.formatter import EvalFormatter

        def repl(mo):
            res = ''
            if mo and mo.lastindex:
                for dex in range(0, mo.lastindex):
                    gv = mo.group(dex+1)
                    if gv is None:
                        continue
                    if len(args) > dex:
                        template = args[dex].replace('[[', '{').replace(']]', '}')
                        res += EvalFormatter().safe_format(template, {'$': gv},
                                           'EVAL', None, strip_results=False)
                    else:
                        res += gv
            return res
        return re.sub(pattern, repl, val, flags=re.I)


class BuiltinSwapAroundComma(BuiltinFormatterFunction):
    name = 'swap_around_comma'
    arg_count = 1
    category = 'String manipulation'
    __doc__ = doc = _('swap_around_comma(val) -- given a value of the form '
            '"B, A", return "A B". This is most useful for converting names '
            'in LN, FN format to FN LN. If there is no comma, the function '
            'returns val unchanged')

    def evaluate(self, formatter, kwargs, mi, locals, val):
        return re.sub(r'^(.*?),\s*(.*$)', r'\2 \1', val, flags=re.I).strip()


class BuiltinIfempty(BuiltinFormatterFunction):
    name = 'ifempty'
    arg_count = 2
    category = 'If-then-else'
    __doc__ = doc = _('ifempty(val, text if empty) -- return val if val is not empty, '
            'otherwise return `text if empty`')

    def evaluate(self, formatter, kwargs, mi, locals, val, value_if_empty):
        if val:
            return val
        else:
            return value_if_empty


class BuiltinShorten(BuiltinFormatterFunction):
    name = 'shorten'
    arg_count = 4
    category = 'String manipulation'
    __doc__ = doc = _('shorten(val, left chars, middle text, right chars) -- Return a '
            'shortened version of val, consisting of `left chars` '
            'characters from the beginning of val, followed by '
            '`middle text`, followed by `right chars` characters from '
            'the end of the string. `Left chars` and `right chars` must be '
            'integers. For example, assume the title of the book is '
            '`Ancient English Laws in the Times of Ivanhoe`, and you want '
            'it to fit in a space of at most 15 characters. If you use '
            '{title:shorten(9,-,5)}, the result will be `Ancient E-nhoe`. '
            'If the field\'s length is less than left chars + right chars + '
            'the length of `middle text`, then the field will be used '
            'intact. For example, the title `The Dome` would not be changed.')

    def evaluate(self, formatter, kwargs, mi, locals,
                 val, leading, center_string, trailing):
        l = max(0, int(leading))
        t = max(0, int(trailing))
        if len(val) > l + len(center_string) + t:
            return val[0:l] + center_string + ('' if t == 0 else val[-t:])
        else:
            return val


class BuiltinCount(BuiltinFormatterFunction):
    name = 'count'
    arg_count = 2
    category = 'List manipulation'
    __doc__ = doc = _('count(val, separator) -- interprets the value as a list of items '
            'separated by `separator`, returning the number of items in the '
            'list. Most lists use a comma as the separator, but authors '
            'uses an ampersand. Examples: {tags:count(,)}, {authors:count(&)}')

    def evaluate(self, formatter, kwargs, mi, locals, val, sep):
        return str(len([v for v in val.split(sep) if v]))


class BuiltinListitem(BuiltinFormatterFunction):
    name = 'list_item'
    arg_count = 3
    category = 'List lookup'
    __doc__ = doc = _('list_item(val, index, separator) -- interpret the value as a list of '
            'items separated by `separator`, returning the `index`th item. '
            'The first item is number zero. The last item can be returned '
            'using `list_item(-1,separator)`. If the item is not in the list, '
            'then the empty value is returned. The separator has the same '
            'meaning as in the count function.')

    def evaluate(self, formatter, kwargs, mi, locals, val, index, sep):
        if not val:
            return ''
        index = int(index)
        val = val.split(sep)
        try:
            return val[index].strip()
        except:
            return ''


class BuiltinSelect(BuiltinFormatterFunction):
    name = 'select'
    arg_count = 2
    category = 'List lookup'
    __doc__ = doc = _('select(val, key) -- interpret the value as a comma-separated list '
            'of items, with the items being "id:value". Find the pair with the '
            'id equal to key, and return the corresponding value.'
            )

    def evaluate(self, formatter, kwargs, mi, locals, val, key):
        if not val:
            return ''
        vals = [v.strip() for v in val.split(',')]
        for v in vals:
            if v.startswith(key+':'):
                return v[len(key)+1:]
        return ''


class BuiltinApproximateFormats(BuiltinFormatterFunction):
    name = 'approximate_formats'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('approximate_formats() -- return a comma-separated '
                  'list of formats that at one point were associated with the '
                  'book. There is no guarantee that this list is correct, '
                  'although it probably is. '
                  'This function can be called in template program mode using '
                  'the template "{:\'approximate_formats()\'}". '
                  'Note that format names are always uppercase, as in EPUB. '
                  'This function works only in the GUI. If you want to use these values '
                  'in save-to-disk or send-to-device templates then you '
                  'must make a custom "Column built from other columns", use '
                  'the function in that column\'s template, and use that '
                  'column\'s value in your save/send templates'
            )

    def evaluate(self, formatter, kwargs, mi, locals):
        if hasattr(mi, '_proxy_metadata'):
            fmt_data = mi._proxy_metadata.db_approx_formats
            if not fmt_data:
                return ''
            data = sorted(fmt_data)
            return ','.join(v.upper() for v in data)
        return _('This function can be used only in the GUI')


class BuiltinFormatsModtimes(BuiltinFormatterFunction):
    name = 'formats_modtimes'
    arg_count = 1
    category = 'Get values from metadata'
    __doc__ = doc = _('formats_modtimes(date_format) -- return a comma-separated '
                  'list of colon-separated items representing modification times '
                  'for the formats of a book. The date_format parameter '
                  'specifies how the date is to be formatted. See the '
                  'format_date function for details. You can use the select '
                  'function to get the mod time for a specific '
                  'format. Note that format names are always uppercase, '
                  'as in EPUB.'
            )

    def evaluate(self, formatter, kwargs, mi, locals, fmt):
        fmt_data = mi.get('format_metadata', {})
        try:
            data = sorted(fmt_data.items(), key=lambda x:x[1]['mtime'], reverse=True)
            return ','.join(k.upper()+':'+format_date(v['mtime'], fmt)
                        for k,v in data)
        except:
            return ''


class BuiltinFormatsSizes(BuiltinFormatterFunction):
    name = 'formats_sizes'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('formats_sizes() -- return a comma-separated list of '
                      'colon-separated items representing sizes in bytes '
                      'of the formats of a book. You can use the select '
                      'function to get the size for a specific '
                      'format. Note that format names are always uppercase, '
                      'as in EPUB.'
            )

    def evaluate(self, formatter, kwargs, mi, locals):
        fmt_data = mi.get('format_metadata', {})
        try:
            return ','.join(k.upper()+':'+str(v['size']) for k,v in iteritems(fmt_data))
        except:
            return ''


class BuiltinFormatsPaths(BuiltinFormatterFunction):
    name = 'formats_paths'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('formats_paths() -- return a comma-separated list of '
                      'colon-separated items representing full path to '
                      'the formats of a book. You can use the select '
                      'function to get the path for a specific '
                      'format. Note that format names are always uppercase, '
                      'as in EPUB.')

    def evaluate(self, formatter, kwargs, mi, locals):
        fmt_data = mi.get('format_metadata', {})
        try:
            return ','.join(k.upper()+':'+str(v['path']) for k,v in iteritems(fmt_data))
        except:
            return ''


class BuiltinHumanReadable(BuiltinFormatterFunction):
    name = 'human_readable'
    arg_count = 1
    category = 'Formatting values'
    __doc__ = doc = _('human_readable(v) -- return a string '
                      'representing the number v in KB, MB, GB, etc.'
            )

    def evaluate(self, formatter, kwargs, mi, locals, val):
        try:
            return human_readable(round(float(val)))
        except:
            return ''


class BuiltinFormatNumber(BuiltinFormatterFunction):
    name = 'format_number'
    arg_count = 2
    category = 'Formatting values'
    __doc__ = doc = _('format_number(v, template) -- format the number v using '
                  'a Python formatting template such as "{0:5.2f}" or '
                  '"{0:,d}" or "${0:5,.2f}". The field_name part of the '
                  'template must be a 0 (zero) (the "{0:" in the above examples). '
                  'See the template language and Python documentation for more '
                  'examples. You can leave off the leading "{0:" and trailing '
                  '"}" if the template contains only a format. Returns the empty '
                  'string if formatting fails.'
            )

    def evaluate(self, formatter, kwargs, mi, locals, val, template):
        if val == '' or val == 'None':
            return ''
        if '{' not in template:
            template = '{0:' + template + '}'
        try:
            v1 = float(val)
        except:
            return ''
        try:  # Try formatting the value as a float
            return template.format(v1)
        except:
            pass
        try:  # Try formatting the value as an int
            v2 = trunc(v1)
            if v2 == v1:
                return template.format(v2)
        except:
            pass
        return ''


class BuiltinSublist(BuiltinFormatterFunction):
    name = 'sublist'
    arg_count = 4
    category = 'List manipulation'
    __doc__ = doc = _('sublist(val, start_index, end_index, separator) -- interpret the '
            'value as a list of items separated by `separator`, returning a '
            'new list made from the `start_index` to the `end_index` item. '
            'The first item is number zero. If an index is negative, then it '
            'counts from the end of the list. As a special case, an end_index '
            'of zero is assumed to be the length of the list. Examples using '
            'basic template mode and assuming that the tags column (which is '
            'comma-separated) contains "A, B, C": '
            '{tags:sublist(0,1,\\\\,)} returns "A". '
            '{tags:sublist(-1,0,\\\\,)} returns "C". '
            '{tags:sublist(0,-1,\\\\,)} returns "A, B".'
            )

    def evaluate(self, formatter, kwargs, mi, locals, val, start_index, end_index, sep):
        if not val:
            return ''
        si = int(start_index)
        ei = int(end_index)
        # allow empty list items so counts are what the user expects
        val = [v.strip() for v in val.split(sep)]

        if sep == ',':
            sep = ', '
        try:
            if ei == 0:
                return sep.join(val[si:])
            else:
                return sep.join(val[si:ei])
        except:
            return ''


class BuiltinSubitems(BuiltinFormatterFunction):
    name = 'subitems'
    arg_count = 3
    category = 'List manipulation'
    __doc__ = doc = _('subitems(val, start_index, end_index) -- This function is used to '
            'break apart lists of items such as genres. It interprets the value '
            'as a comma-separated list of items, where each item is a period-'
            'separated list. Returns a new list made by first finding all the '
            'period-separated items, then for each such item extracting the '
            '`start_index` to the `end_index` components, then combining '
            'the results back together. The first component in a period-'
            'separated list has an index of zero. If an index is negative, '
            'then it counts from the end of the list. As a special case, an '
            'end_index of zero is assumed to be the length of the list. '
            'Example using basic template mode and assuming a #genre value of '
            '"A.B.C": {#genre:subitems(0,1)} returns "A". {#genre:subitems(0,2)} '
            'returns "A.B". {#genre:subitems(1,0)} returns "B.C". Assuming a #genre '
            'value of "A.B.C, D.E.F", {#genre:subitems(0,1)} returns "A, D". '
            '{#genre:subitems(0,2)} returns "A.B, D.E"')

    period_pattern = re.compile(r'(?<=[^\.\s])\.(?=[^\.\s])', re.U)

    def evaluate(self, formatter, kwargs, mi, locals, val, start_index, end_index):
        if not val:
            return ''
        si = int(start_index)
        ei = int(end_index)
        has_periods = '.' in val
        items = [v.strip() for v in val.split(',')]
        rv = set()
        for item in items:
            if has_periods and '.' in item:
                components = self.period_pattern.split(item)
            else:
                components = [item]
            try:
                if ei == 0:
                    rv.add('.'.join(components[si:]))
                else:
                    rv.add('.'.join(components[si:ei]))
            except:
                pass
        return ', '.join(sorted(rv, key=sort_key))


class BuiltinFormatDate(BuiltinFormatterFunction):
    name = 'format_date'
    arg_count = 2
    category = 'Formatting values'
    __doc__ = doc = _('format_date(val, format_string) -- format the value, '
            'which must be a date, using the format_string, returning a string. '
            'The formatting codes are: '
            'd    : the day as number without a leading zero (1 to 31) '
            'dd   : the day as number with a leading zero (01 to 31) '
            'ddd  : the abbreviated localized day name (e.g. "Mon" to "Sun"). '
            'dddd : the long localized day name (e.g. "Monday" to "Sunday"). '
            'M    : the month as number without a leading zero (1 to 12). '
            'MM   : the month as number with a leading zero (01 to 12) '
            'MMM  : the abbreviated localized month name (e.g. "Jan" to "Dec"). '
            'MMMM : the long localized month name (e.g. "January" to "December"). '
            'yy   : the year as two digit number (00 to 99). '
            'yyyy : the year as four digit number. '
            'h    : the hours without a leading 0 (0 to 11 or 0 to 23, depending on am/pm) '
            'hh   : the hours with a leading 0 (00 to 11 or 00 to 23, depending on am/pm) '
            'm    : the minutes without a leading 0 (0 to 59) '
            'mm   : the minutes with a leading 0 (00 to 59) '
            's    : the seconds without a leading 0 (0 to 59) '
            'ss   : the seconds with a leading 0 (00 to 59) '
            'ap   : use a 12-hour clock instead of a 24-hour clock, with "ap" replaced by the localized string for am or pm '
            'AP   : use a 12-hour clock instead of a 24-hour clock, with "AP" replaced by the localized string for AM or PM '
            'iso  : the date with time and timezone. Must be the only format present')

    def evaluate(self, formatter, kwargs, mi, locals, val, format_string):
        if not val or val == 'None':
            return ''
        try:
            dt = parse_date(val)
            s = format_date(dt, format_string)
        except:
            s = 'BAD DATE'
        return s


class BuiltinUppercase(BuiltinFormatterFunction):
    name = 'uppercase'
    arg_count = 1
    category = 'String case changes'
    __doc__ = doc = _('uppercase(val) -- return val in upper case')

    def evaluate(self, formatter, kwargs, mi, locals, val):
        return val.upper()


class BuiltinLowercase(BuiltinFormatterFunction):
    name = 'lowercase'
    arg_count = 1
    category = 'String case changes'
    __doc__ = doc = _('lowercase(val) -- return val in lower case')

    def evaluate(self, formatter, kwargs, mi, locals, val):
        return val.lower()


class BuiltinTitlecase(BuiltinFormatterFunction):
    name = 'titlecase'
    arg_count = 1
    category = 'String case changes'
    __doc__ = doc = _('titlecase(val) -- return val in title case')

    def evaluate(self, formatter, kwargs, mi, locals, val):
        return titlecase(val)


class BuiltinCapitalize(BuiltinFormatterFunction):
    name = 'capitalize'
    arg_count = 1
    category = 'String case changes'
    __doc__ = doc = _('capitalize(val) -- return val capitalized')

    def evaluate(self, formatter, kwargs, mi, locals, val):
        return capitalize(val)


class BuiltinBooksize(BuiltinFormatterFunction):
    name = 'booksize'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('booksize() -- return value of the size field. '
                'This function works only in the GUI. If you want to use this value '
                'in save-to-disk or send-to-device templates then you '
                'must make a custom "Column built from other columns", use '
                'the function in that column\'s template, and use that '
                'column\'s value in your save/send templates')

    def evaluate(self, formatter, kwargs, mi, locals):
        if hasattr(mi, '_proxy_metadata'):
            try:
                v = mi._proxy_metadata.book_size
                if v is not None:
                    return str(mi._proxy_metadata.book_size)
                return ''
            except:
                pass
            return ''
        return _('This function can be used only in the GUI')


class BuiltinOndevice(BuiltinFormatterFunction):
    name = 'ondevice'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('ondevice() -- return Yes if ondevice is set, otherwise return '
              'the empty string. This function works only in the GUI. If you want to '
              'use this value in save-to-disk or send-to-device templates then you '
              'must make a custom "Column built from other columns", use '
              'the function in that column\'s template, and use that '
              'column\'s value in your save/send templates')

    def evaluate(self, formatter, kwargs, mi, locals):
        if hasattr(mi, '_proxy_metadata'):
            if mi._proxy_metadata.ondevice_col:
                return _('Yes')
            return ''
        return _('This function can be used only in the GUI')


class BuiltinSeriesSort(BuiltinFormatterFunction):
    name = 'series_sort'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('series_sort() -- return the series sort value')

    def evaluate(self, formatter, kwargs, mi, locals):
        if mi.series:
            return title_sort(mi.series)
        return ''


class BuiltinHasCover(BuiltinFormatterFunction):
    name = 'has_cover'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('has_cover() -- return Yes if the book has a cover, '
                      'otherwise return the empty string')

    def evaluate(self, formatter, kwargs, mi, locals):
        if mi.has_cover:
            return _('Yes')
        return ''


class BuiltinFirstNonEmpty(BuiltinFormatterFunction):
    name = 'first_non_empty'
    arg_count = -1
    category = 'Iterating over values'
    __doc__ = doc = _('first_non_empty(value, value, ...) -- '
            'returns the first value that is not empty. If all values are '
            'empty, then the empty value is returned. '
            'You can have as many values as you want.')

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        i = 0
        while i < len(args):
            if args[i]:
                return args[i]
            i += 1
        return ''


class BuiltinAnd(BuiltinFormatterFunction):
    name = 'and'
    arg_count = -1
    category = 'Boolean'
    __doc__ = doc = _('and(value, value, ...) -- '
            'returns the string "1" if all values are not empty, otherwise '
            'returns the empty string. This function works well with test or '
            'first_non_empty. You can have as many values as you want. ')

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        i = 0
        while i < len(args):
            if not args[i]:
                return ''
            i += 1
        return '1'


class BuiltinOr(BuiltinFormatterFunction):
    name = 'or'
    arg_count = -1
    category = 'Boolean'
    __doc__ = doc = _('or(value, value, ...) -- '
            'returns the string "1" if any value is not empty, otherwise '
            'returns the empty string. This function works well with test or '
            'first_non_empty. You can have as many values as you want.')

    def evaluate(self, formatter, kwargs, mi, locals, *args):
        i = 0
        while i < len(args):
            if args[i]:
                return '1'
            i += 1
        return ''


class BuiltinNot(BuiltinFormatterFunction):
    name = 'not'
    arg_count = 1
    category = 'Boolean'
    __doc__ = doc = _('not(value) -- '
            'returns the string "1" if the value is empty, otherwise '
            'returns the empty string. This function works well with test or '
            'first_non_empty.')

    def evaluate(self, formatter, kwargs, mi, locals, val):
        return '' if val else '1'


class BuiltinListUnion(BuiltinFormatterFunction):
    name = 'list_union'
    arg_count = 3
    category = 'List manipulation'
    __doc__ = doc = _('list_union(list1, list2, separator) -- '
            'return a list made by merging the items in list1 and list2, '
            'removing duplicate items using a case-insensitive comparison. If '
            'items differ in case, the one in list1 is used. '
            'The items in list1 and list2 are separated by separator, as are '
            'the items in the returned list.')
    aliases = ['merge_lists']

    def evaluate(self, formatter, kwargs, mi, locals, list1, list2, separator):
        res = [l.strip() for l in list1.split(separator) if l.strip()]
        l2 = [l.strip() for l in list2.split(separator) if l.strip()]
        lcl1 = {icu_lower(l) for l in res}

        for i in l2:
            if icu_lower(i) not in lcl1 and i not in res:
                res.append(i)
        if separator == ',':
            return ', '.join(res)
        return separator.join(res)


class BuiltinListDifference(BuiltinFormatterFunction):
    name = 'list_difference'
    arg_count = 3
    category = 'List manipulation'
    __doc__ = doc = _('list_difference(list1, list2, separator) -- '
            'return a list made by removing from list1 any item found in list2, '
            'using a case-insensitive comparison. The items in list1 and list2 '
            'are separated by separator, as are the items in the returned list.')

    def evaluate(self, formatter, kwargs, mi, locals, list1, list2, separator):
        l1 = [l.strip() for l in list1.split(separator) if l.strip()]
        l2 = {icu_lower(l.strip()) for l in list2.split(separator) if l.strip()}

        res = []
        for i in l1:
            if icu_lower(i) not in l2 and i not in res:
                res.append(i)
        if separator == ',':
            return ', '.join(res)
        return separator.join(res)


class BuiltinListIntersection(BuiltinFormatterFunction):
    name = 'list_intersection'
    arg_count = 3
    category = 'List manipulation'
    __doc__ = doc = _('list_intersection(list1, list2, separator) -- '
            'return a list made by removing from list1 any item not found in list2, '
            'using a case-insensitive comparison. The items in list1 and list2 '
            'are separated by separator, as are the items in the returned list.')

    def evaluate(self, formatter, kwargs, mi, locals, list1, list2, separator):
        l1 = [l.strip() for l in list1.split(separator) if l.strip()]
        l2 = {icu_lower(l.strip()) for l in list2.split(separator) if l.strip()}

        res = []
        for i in l1:
            if icu_lower(i) in l2 and i not in res:
                res.append(i)
        if separator == ',':
            return ', '.join(res)
        return separator.join(res)


class BuiltinListSort(BuiltinFormatterFunction):
    name = 'list_sort'
    arg_count = 3
    category = 'List manipulation'
    __doc__ = doc = _('list_sort(list, direction, separator) -- '
            'return list sorted using a case-insensitive sort. If direction is '
            'zero, the list is sorted ascending, otherwise descending. The list items '
            'are separated by separator, as are the items in the returned list.')

    def evaluate(self, formatter, kwargs, mi, locals, list1, direction, separator):
        res = [l.strip() for l in list1.split(separator) if l.strip()]
        if separator == ',':
            return ', '.join(sorted(res, key=sort_key, reverse=direction != "0"))
        return separator.join(sorted(res, key=sort_key, reverse=direction != "0"))


class BuiltinListEquals(BuiltinFormatterFunction):
    name = 'list_equals'
    arg_count = 6
    category = 'List manipulation'
    __doc__ = doc = _('list_equals(list1, sep1, list2, sep2, yes_val, no_val) -- '
            'return yes_val if list1 and list2 contain the same items, '
            'otherwise return no_val. The items are determined by splitting '
            'each list using the appropriate separator character (sep1 or '
            'sep2). The order of items in the lists is not relevant. '
            'The comparison is case insensitive.')

    def evaluate(self, formatter, kwargs, mi, locals, list1, sep1, list2, sep2, yes_val, no_val):
        s1 = {icu_lower(l.strip()) for l in list1.split(sep1) if l.strip()}
        s2 = {icu_lower(l.strip()) for l in list2.split(sep2) if l.strip()}
        if s1 == s2:
            return yes_val
        return no_val


class BuiltinListRe(BuiltinFormatterFunction):
    name = 'list_re'
    arg_count = 4
    category = 'List manipulation'
    __doc__ = doc = _('list_re(src_list, separator, include_re, opt_replace) -- '
            'Construct a list by first separating src_list into items using '
            'the separator character. For each item in the list, check if it '
            'matches include_re. If it does, then add it to the list to be '
            'returned. If opt_replace is not the empty string, then apply the '
            'replacement before adding the item to the returned list.')

    def evaluate(self, formatter, kwargs, mi, locals, src_list, separator, include_re, opt_replace):
        l = [l.strip() for l in src_list.split(separator) if l.strip()]
        res = []
        for item in l:
            if re.search(include_re, item, flags=re.I) is not None:
                if opt_replace:
                    item = re.sub(include_re, opt_replace, item)
                for i in [t.strip() for t in item.split(separator) if t.strip()]:
                    if i not in res:
                        res.append(i)
        if separator == ',':
            return ', '.join(res)
        return separator.join(res)


class BuiltinListReGroup(BuiltinFormatterFunction):
    name = 'list_re_group'
    arg_count = -1
    category = 'List manipulation'
    __doc__ = doc = _('list_re_group(src_list, separator, include_re, search_re, group_1_template, ...) -- '
                      'Like list_re except replacements are not optional. It '
                      'uses re_group(list_item, search_re, group_1_template, ...) when '
                      'doing the replacements on the resulting list.')

    def evaluate(self, formatter, kwargs, mi, locals, src_list, separator, include_re,
                 search_re, *args):
        from ebook_converter.utils.formatter import EvalFormatter

        l = [l.strip() for l in src_list.split(separator) if l.strip()]
        res = []
        for item in l:
            def repl(mo):
                newval = ''
                if mo and mo.lastindex:
                    for dex in range(0, mo.lastindex):
                        gv = mo.group(dex+1)
                        if gv is None:
                            continue
                        if len(args) > dex:
                            template = args[dex].replace('[[', '{').replace(']]', '}')
                            newval += EvalFormatter().safe_format(template, {'$': gv},
                                              'EVAL', None, strip_results=False)
                        else:
                            newval += gv
                return newval
            if re.search(include_re, item, flags=re.I) is not None:
                item = re.sub(search_re, repl, item, flags=re.I)
                for i in [t.strip() for t in item.split(separator) if t.strip()]:
                    if i not in res:
                        res.append(i)
        if separator == ',':
            return ', '.join(res)
        return separator.join(res)


class BuiltinToday(BuiltinFormatterFunction):
    name = 'today'
    arg_count = 0
    category = 'Date functions'
    __doc__ = doc = _('today() -- '
            'return a date string for today. This value is designed for use in '
            'format_date or days_between, but can be manipulated like any '
            'other string. The date is in ISO format.')

    def evaluate(self, formatter, kwargs, mi, locals):
        return format_date(now(), 'iso')


class BuiltinDaysBetween(BuiltinFormatterFunction):
    name = 'days_between'
    arg_count = 2
    category = 'Date functions'
    __doc__ = doc = _('days_between(date1, date2) -- '
            'return the number of days between date1 and date2. The number is '
            'positive if date1 is greater than date2, otherwise negative. If '
            'either date1 or date2 are not dates, the function returns the '
            'empty string.')

    def evaluate(self, formatter, kwargs, mi, locals, date1, date2):
        try:
            d1 = parse_date(date1)
            if d1 == UNDEFINED_DATE:
                return ''
            d2 = parse_date(date2)
            if d2 == UNDEFINED_DATE:
                return ''
        except:
            return ''
        i = d1 - d2
        return '%.1f'%(i.days + (i.seconds/(24.0*60.0*60.0)))


class BuiltinLanguageStrings(BuiltinFormatterFunction):
    name = 'language_strings'
    arg_count = 2
    category = 'Get values from metadata'
    __doc__ = doc = _('language_strings(lang_codes, localize) -- '
            'return the strings for the language codes passed in lang_codes. '
            'If localize is zero, return the strings in English. If '
            'localize is not zero, return the strings in the language of '
            'the current locale. Lang_codes is a comma-separated list.')

    def evaluate(self, formatter, kwargs, mi, locals, lang_codes, localize):
        retval = []
        for c in [c.strip() for c in lang_codes.split(',') if c.strip()]:
            try:
                n = calibre_langcode_to_name(c, localize != '0')
                if n:
                    retval.append(n)
            except:
                pass
        return ', '.join(retval)


class BuiltinLanguageCodes(BuiltinFormatterFunction):
    name = 'language_codes'
    arg_count = 1
    category = 'Get values from metadata'
    __doc__ = doc = _('language_codes(lang_strings) -- '
            'return the language codes for the strings passed in lang_strings. '
            'The strings must be in the language of the current locale. '
            'Lang_strings is a comma-separated list.')

    def evaluate(self, formatter, kwargs, mi, locals, lang_strings):
        retval = []
        for c in [c.strip() for c in lang_strings.split(',') if c.strip()]:
            try:
                cv = canonicalize_lang(c)
                if cv:
                    retval.append(canonicalize_lang(cv))
            except:
                pass
        return ', '.join(retval)


class BuiltinCurrentLibraryName(BuiltinFormatterFunction):
    name = 'current_library_name'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('current_library_name() -- '
            'return the last name on the path to the current calibre library. '
            'This function can be called in template program mode using the '
            'template "{:\'current_library_name()\'}".')

    def evaluate(self, formatter, kwargs, mi, locals):
        from ebook_converter.library import current_library_name
        return current_library_name()


class BuiltinCurrentLibraryPath(BuiltinFormatterFunction):
    name = 'current_library_path'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('current_library_path() -- '
                'return the path to the current calibre library. This function can '
                'be called in template program mode using the template '
                '"{:\'current_library_path()\'}".')

    def evaluate(self, formatter, kwargs, mi, locals):
        from ebook_converter.library import current_library_path
        return current_library_path()


class BuiltinFinishFormatting(BuiltinFormatterFunction):
    name = 'finish_formatting'
    arg_count = 4
    category = 'Formatting values'
    __doc__ = doc = _('finish_formatting(val, fmt, prefix, suffix) -- apply the '
                      'format, prefix, and suffix to a value in the same way as '
                      'done in a template like `{series_index:05.2f| - |- }`. For '
                      'example, the following program produces the same output '
                      'as the above template: '
                      'program: finish_formatting(field("series_index"), "05.2f", " - ", " - ")')

    def evaluate(self, formatter, kwargs, mi, locals_, val, fmt, prefix, suffix):
        if not val:
            return val
        return prefix + formatter._do_format(val, fmt) + suffix


class BuiltinVirtualLibraries(BuiltinFormatterFunction):
    name = 'virtual_libraries'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('virtual_libraries() -- return a comma-separated list of '
                      'virtual libraries that contain this book. This function '
                      'works only in the GUI. If you want to use these values '
                      'in save-to-disk or send-to-device templates then you '
                      'must make a custom "Column built from other columns", use '
                      'the function in that column\'s template, and use that '
                      'column\'s value in your save/send templates')

    def evaluate(self, formatter, kwargs, mi, locals_):
        if hasattr(mi, '_proxy_metadata'):
            return mi._proxy_metadata.virtual_libraries
        return _('This function can be used only in the GUI')


class BuiltinUserCategories(BuiltinFormatterFunction):
    name = 'user_categories'
    arg_count = 0
    category = 'Get values from metadata'
    __doc__ = doc = _('user_categories() -- return a comma-separated list of '
                      'the user categories that contain this book. This function '
                      'works only in the GUI. If you want to use these values '
                      'in save-to-disk or send-to-device templates then you '
                      'must make a custom "Column built from other columns", use '
                      'the function in that column\'s template, and use that '
                      'column\'s value in your save/send templates')

    def evaluate(self, formatter, kwargs, mi, locals_):
        if hasattr(mi, '_proxy_metadata'):
            cats = set(k for k, v in iteritems(mi._proxy_metadata.user_categories) if v)
            cats = sorted(cats, key=sort_key)
            return ', '.join(cats)
        return _('This function can be used only in the GUI')


class BuiltinTransliterate(BuiltinFormatterFunction):
    name = 'transliterate'
    arg_count = 1
    category = 'String manipulation'
    __doc__ = doc = _('transliterate(a) -- Returns a string in a latin alphabet '
                      'formed by approximating the sound of the words in the '
                      'source string. For example, if the source is "{0}"'
                      ' the function returns "{1}".').format(
                          u"Фёдор Миха́йлович Достоевский", 'Fiodor Mikhailovich Dostoievskii')

    def evaluate(self, formatter, kwargs, mi, locals, source):
        from ebook_converter.utils.filenames import ascii_text
        return ascii_text(source)


class BuiltinAuthorLinks(BuiltinFormatterFunction):
    name = 'author_links'
    arg_count = 2
    category = 'Get values from metadata'
    __doc__ = doc = _('author_links(val_separator, pair_separator) -- returns '
                      'a string containing a list of authors and that author\'s '
                      'link values in the '
                      'form author1 val_separator author1link pair_separator '
                      'author2 val_separator author2link etc. An author is '
                      'separated from its link value by the val_separator string '
                      'with no added spaces. author:linkvalue pairs are separated '
                      'by the pair_separator string argument with no added spaces. '
                      'It is up to you to choose separator strings that do '
                      'not occur in author names or links. An author is '
                      'included even if the author link is empty.')

    def evaluate(self, formatter, kwargs, mi, locals, val_sep, pair_sep):
        if hasattr(mi, '_proxy_metadata'):
            link_data = mi._proxy_metadata.author_link_map
            if not link_data:
                return ''
            names = sorted(link_data.keys(), key=sort_key)
            return pair_sep.join(n + val_sep + link_data[n] for n in names)
        return _('This function can be used only in the GUI')


class BuiltinAuthorSorts(BuiltinFormatterFunction):
    name = 'author_sorts'
    arg_count = 1
    category = 'Get values from metadata'
    __doc__ = doc = _('author_sorts(val_separator) -- returns a string '
                      'containing a list of author\'s sort values for the '
                      'authors of the book. The sort is the one in the author '
                      'metadata (different from the author_sort in books). The '
                      'returned list has the form author sort 1 val_separator '
                      'author sort 2 etc. The author sort values in this list '
                      'are in the same order as the authors of the book. If '
                      'you want spaces around val_separator then include them '
                      'in the separator string')

    def evaluate(self, formatter, kwargs, mi, locals, val_sep):
        sort_data = mi.author_sort_map
        if not sort_data:
            return ''
        names = [sort_data.get(n) for n in mi.authors if n.strip()]
        return val_sep.join(n for n in names)


_formatter_builtins = [
    BuiltinAdd(), BuiltinAnd(), BuiltinApproximateFormats(), BuiltinAssign(),
    BuiltinAuthorLinks(), BuiltinAuthorSorts(), BuiltinBooksize(),
    BuiltinCapitalize(), BuiltinCmp(), BuiltinContains(), BuiltinCount(),
    BuiltinCurrentLibraryName(), BuiltinCurrentLibraryPath(),
    BuiltinDaysBetween(), BuiltinDivide(), BuiltinEval(), BuiltinFirstNonEmpty(),
    BuiltinField(), BuiltinFinishFormatting(), BuiltinFirstMatchingCmp(),
    BuiltinFormatDate(), BuiltinFormatNumber(), BuiltinFormatsModtimes(),
    BuiltinFormatsPaths(), BuiltinFormatsSizes(),
    BuiltinHasCover(), BuiltinHumanReadable(), BuiltinIdentifierInList(),
    BuiltinIfempty(), BuiltinLanguageCodes(), BuiltinLanguageStrings(),
    BuiltinInList(), BuiltinListDifference(), BuiltinListEquals(),
    BuiltinListIntersection(), BuiltinListitem(), BuiltinListRe(),
    BuiltinListReGroup(), BuiltinListSort(), BuiltinListUnion(), BuiltinLookup(),
    BuiltinLowercase(), BuiltinMultiply(), BuiltinNot(), BuiltinOndevice(),
    BuiltinOr(), BuiltinPrint(), BuiltinRawField(), BuiltinRawList(),
    BuiltinRe(), BuiltinReGroup(), BuiltinSelect(), BuiltinSeriesSort(),
    BuiltinShorten(), BuiltinStrcat(), BuiltinStrcatMax(),
    BuiltinStrcmp(), BuiltinStrInList(), BuiltinStrlen(), BuiltinSubitems(),
    BuiltinSublist(),BuiltinSubstr(), BuiltinSubtract(), BuiltinSwapAroundComma(),
    BuiltinSwitch(), BuiltinTemplate(), BuiltinTest(), BuiltinTitlecase(),
    BuiltinToday(), BuiltinTransliterate(), BuiltinUppercase(),
    BuiltinUserCategories(), BuiltinVirtualLibraries()
]


class FormatterUserFunction(FormatterFunction):

    def __init__(self, name, doc, arg_count, program_text):
        self.name = name
        self.doc = doc
        self.arg_count = arg_count
        self.program_text = program_text


tabs = re.compile(r'^\t*')


def compile_user_function(name, doc, arg_count, eval_func):
    def replace_func(mo):
        return mo.group().replace('\t', '    ')

    func = '    ' + '\n    '.join([tabs.sub(replace_func, line)
                                   for line in eval_func.splitlines()])
    prog = '''
from ebook_converter.utils.formatter_functions import FormatterUserFunction
from ebook_converter.utils.formatter_functions import formatter_functions
class UserFunction(FormatterUserFunction):
''' + func
    locals_ = {}
    if DEBUG and tweaks.get('enable_template_debug_printing', False):
        print(prog)
    exec(prog, locals_)
    cls = locals_['UserFunction'](name, doc, arg_count, eval_func)
    return cls


def compile_user_template_functions(funcs):
    compiled_funcs = {}
    for func in funcs:
        try:
            # Force a name conflict to test the logic
            # if func[0] == 'myFunc2':
            #     func[0] = 'myFunc3'

            # Compile the function so that the tab processing is done on the
            # source. This helps ensure that if the function already is defined
            # then white space differences don't cause them to compare differently

            cls = compile_user_function(*func)
            compiled_funcs[cls.name] = cls
        except:
            traceback.print_exc()
    return compiled_funcs


def load_user_template_functions(library_uuid, funcs, precompiled_user_functions=None):
    unload_user_template_functions(library_uuid)
    if precompiled_user_functions:
        compiled_funcs = precompiled_user_functions
    else:
        compiled_funcs = compile_user_template_functions(funcs)
    formatter_functions().register_functions(library_uuid, list(compiled_funcs.values()))


def unload_user_template_functions(library_uuid):
    formatter_functions().unregister_functions(library_uuid)