# -*- coding: utf-8 -*-

# copyright 2003-2011 LOGILAB S.A. (Paris, FRANCE), all rights reserved.

# contact http://www.logilab.fr/ -- mailto:contact@logilab.fr

#

# This file is part of CubicWeb.

#

# CubicWeb is free software: you can redistribute it and/or modify it under the

# terms of the GNU Lesser General Public License as published by the Free

# Software Foundation, either version 2.1 of the License, or (at your option)

# any later version.

#

# CubicWeb is distributed in the hope that it will be useful, but WITHOUT

# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS

# FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public License for more

# details.

#

# You should have received a copy of the GNU Lesser General Public License along

# with CubicWeb.  If not, see <http://www.gnu.org/licenses/>.

"""This module provides tools to import tabular data.

Example of use (run this with `cubicweb-ctl shell instance import-script.py`):

.. sourcecode:: python

  from cubicweb.dataimport import *

  # define data generators

  GENERATORS = []

  USERS = [('Prenom', 'firstname', ()),

           ('Nom', 'surname', ()),

           ('Identifiant', 'login', ()),

           ]

  def gen_users(ctl):

      for row in ctl.iter_and_commit('utilisateurs'):

          entity = mk_entity(row, USERS)

          entity['upassword'] = 'motdepasse'

          ctl.check('login', entity['login'], None)

          entity = ctl.store.create_entity('CWUser', **entity)

          email = ctl.store.create_entity('EmailAddress', address=row['email'])

          ctl.store.relate(entity.eid, 'use_email', email.eid)

          ctl.store.rql('SET U in_group G WHERE G name "users", U eid %(x)s', {'x':entity['eid']})

  CHK = [('login', check_doubles, 'Utilisateurs Login',

          'Deux utilisateurs ne devraient pas avoir le même login.'),

         ]

  GENERATORS.append( (gen_users, CHK) )

  # create controller

  if 'cnx' in globals():

      ctl = CWImportController(RQLObjectStore(cnx))

  else:

      print 'debug mode (not connected)'

      print 'run through cubicweb-ctl shell to access an instance'

      ctl = CWImportController(ObjectStore())

  ctl.askerror = 1

  ctl.generators = GENERATORS

  ctl.data['utilisateurs'] = lazytable(ucsvreader(open('users.csv')))

  # run

  ctl.run()

.. BUG file with one column are not parsable

.. TODO rollback() invocation is not possible yet

"""

__docformat__ = "restructuredtext en"

import sys

import csv

import traceback

import os.path as osp

from StringIO import StringIO

from copy import copy

from datetime import datetime

from logilab.common import shellutils, attrdict

from logilab.common.date import strptime

from logilab.common.decorators import cached

from logilab.common.deprecation import deprecated

from cubicweb.schema import META_RTYPES, VIRTUAL_RTYPES

from cubicweb.server.utils import eschema_eid

from cubicweb.server.edition import EditedEntity

def count_lines(stream_or_filename):

    if isinstance(stream_or_filename, basestring):

        f = open(stream_or_filename)

    else:

        f = stream_or_filename

        f.seek(0)

    for i, line in enumerate(f):

        pass

    f.seek(0)

    return i+1

def ucsvreader_pb(stream_or_path, encoding='utf-8', separator=',', quote='"',

                  skipfirst=False, withpb=True):

    """same as ucsvreader but a progress bar is displayed as we iter on rows"""

    if isinstance(stream_or_path, basestring):

        if not osp.exists(stream_or_path):

            raise Exception("file doesn't exists: %s" % stream_or_path)

        stream = open(stream_or_path)

    else:

        stream = stream_or_path

    rowcount = count_lines(stream)

    if skipfirst:

        rowcount -= 1

    if withpb:

        pb = shellutils.ProgressBar(rowcount, 50)

    for urow in ucsvreader(stream, encoding, separator, quote, skipfirst):

        yield urow

        if withpb:

            pb.update()

    print ' %s rows imported' % rowcount

def ucsvreader(stream, encoding='utf-8', separator=',', quote='"',

               skipfirst=False):

    """A csv reader that accepts files with any encoding and outputs unicode

    strings

    """

    it = iter(csv.reader(stream, delimiter=separator, quotechar=quote))

    if skipfirst:

        it.next()

    for row in it:

        yield [item.decode(encoding) for item in row]

def callfunc_every(func, number, iterable):

    """yield items of `iterable` one by one and call function `func`

    every `number` iterations. Always call function `func` at the end.

    """

    for idx, item in enumerate(iterable):

        yield item

        if idx % number:

            func()

    func()

def lazytable(reader):

    """The first row is taken to be the header of the table and

    used to output a dict for each row of data.

    >>> data = lazytable(ucsvreader(open(filename)))

    """

    header = reader.next()

    for row in reader:

        yield dict(zip(header, row))

def mk_entity(row, map):

    """Return a dict made from sanitized mapped values.

    ValueError can be raised on unexpected values found in checkers

    >>> row = {'myname': u'dupont'}

    >>> map = [('myname', u'name', (call_transform_method('title'),))]

    >>> mk_entity(row, map)

    {'name': u'Dupont'}

    >>> row = {'myname': u'dupont', 'optname': u''}

    >>> map = [('myname', u'name', (call_transform_method('title'),)),

    ...        ('optname', u'MARKER', (optional,))]

    >>> mk_entity(row, map)

    {'name': u'Dupont', 'optname': None}

    """

    res = {}

    assert isinstance(row, dict)

    assert isinstance(map, list)

    for src, dest, funcs in map:

        res[dest] = row[src]

        try:

            for func in funcs:

                res[dest] = func(res[dest])

                if res[dest] is None:

                    break

        except ValueError, err:

            raise ValueError('error with %r field: %s' % (src, err))

    return res

# user interactions ############################################################

def tell(msg):

    print msg

def confirm(question):

    """A confirm function that asks for yes/no/abort and exits on abort."""

    answer = shellutils.ASK.ask(question, ('Y', 'n', 'abort'), 'Y')

    if answer == 'abort':

        sys.exit(1)

    return answer == 'Y'

class catch_error(object):

    """Helper for @contextmanager decorator."""

    def __init__(self, ctl, key='unexpected error', msg=None):

        self.ctl = ctl

        self.key = key

        self.msg = msg

    def __enter__(self):

        return self

    def __exit__(self, type, value, traceback):

        if type is not None:

            if issubclass(type, (KeyboardInterrupt, SystemExit)):

                return # re-raise

            if self.ctl.catcherrors:

                self.ctl.record_error(self.key, None, type, value, traceback)

                return True # silent

# base sanitizing/coercing functions ###########################################

def optional(value):

    """checker to filter optional field

    If value is undefined (ex: empty string), return None that will

    break the checkers validation chain

    General use is to add 'optional' check in first condition to avoid

    ValueError by further checkers

    >>> MAPPER = [(u'value', 'value', (optional, int))]

    >>> row = {'value': u'XXX'}

    >>> mk_entity(row, MAPPER)

    {'value': None}

    >>> row = {'value': u'100'}

    >>> mk_entity(row, MAPPER)

    {'value': 100}

    """

    if value:

        return value

    return None

def required(value):

    """raise ValueError if value is empty

    This check should be often found in last position in the chain.

    """

    if value:

        return value

    raise ValueError("required")

def todatetime(format='%d/%m/%Y'):

    """return a transformation function to turn string input value into a

    `datetime.datetime` instance, using given format.

    Follow it by `todate` or `totime` functions from `logilab.common.date` if

    you want a `date`/`time` instance instead of `datetime`.

    """

    def coerce(value):

        return strptime(value, format)

    return coerce

def call_transform_method(methodname, *args, **kwargs):

    """return value returned by calling the given method on input"""

    def coerce(value):

        return getattr(value, methodname)(*args, **kwargs)

    return coerce

def call_check_method(methodname, *args, **kwargs):

    """check value returned by calling the given method on input is true,

    else raise ValueError

    """

    def check(value):

        if getattr(value, methodname)(*args, **kwargs):

            return value

        raise ValueError('%s not verified on %r' % (methodname, value))

    return check

# base integrity checking functions ############################################

def check_doubles(buckets):

    """Extract the keys that have more than one item in their bucket."""

    return [(k, len(v)) for k, v in buckets.items() if len(v) > 1]

def check_doubles_not_none(buckets):

    """Extract the keys that have more than one item in their bucket."""

    return [(k, len(v)) for k, v in buckets.items()

            if k is not None and len(v) > 1]

# object stores #################################################################

class ObjectStore(object):

    """Store objects in memory for *faster* validation (development mode)

    But it will not enforce the constraints of the schema and hence will miss some problems

    >>> store = ObjectStore()

    >>> user = store.create_entity('CWUser', login=u'johndoe')

    >>> group = store.create_entity('CWUser', name=u'unknown')

    >>> store.relate(user.eid, 'in_group', group.eid)

    """

    def __init__(self):

        self.items = []

        self.eids = {}

        self.types = {}

        self.relations = set()

        self.indexes = {}

        self._rql = None

        self._commit = None

    def _put(self, type, item):

        self.items.append(item)

        return len(self.items) - 1

    def create_entity(self, etype, **data):

        data = attrdict(data)

        data['eid'] = eid = self._put(etype, data)

        self.eids[eid] = data

        self.types.setdefault(etype, []).append(eid)

        return data

    @deprecated("[3.11] add is deprecated, use create_entity instead")

    def add(self, etype, item):

        assert isinstance(item, dict), 'item is not a dict but a %s' % type(item)

        data = self.create_entity(etype, **item)

        item['eid'] = data['eid']

        return item

    def relate(self, eid_from, rtype, eid_to, inlined=False):

        """Add new relation"""

        relation = eid_from, rtype, eid_to

        self.relations.add(relation)

        return relation

    def commit(self):

        """this commit method do nothing by default

        This is voluntary to use the frequent autocommit feature in CubicWeb

        when you are using hooks or another

        If you want override commit method, please set it by the

        constructor

        """

        pass

    def rql(self, *args):

        if self._rql is not None:

            return self._rql(*args)

        return []

    @property

    def nb_inserted_entities(self):

        return len(self.eids)

    @property

    def nb_inserted_types(self):

        return len(self.types)

    @property

    def nb_inserted_relations(self):

        return len(self.relations)

    @deprecated("[3.7] index support will disappear")

    def build_index(self, name, type, func=None, can_be_empty=False):

        """build internal index for further search"""

        index = {}

        if func is None or not callable(func):

            func = lambda x: x['eid']

        for eid in self.types[type]:

            index.setdefault(func(self.eids[eid]), []).append(eid)

        if not can_be_empty:

            assert index, "new index '%s' cannot be empty" % name

        self.indexes[name] = index

    @deprecated("[3.7] index support will disappear")

    def build_rqlindex(self, name, type, key, rql, rql_params=False,

                       func=None, can_be_empty=False):

        """build an index by rql query

        rql should return eid in first column

        ctl.store.build_index('index_name', 'users', 'login', 'Any U WHERE U is CWUser')

        """

        self.types[type] = []

        rset = self.rql(rql, rql_params or {})

        if not can_be_empty:

            assert rset, "new index type '%s' cannot be empty (0 record found)" % type

        for entity in rset.entities():

            getattr(entity, key) # autopopulate entity with key attribute

            self.eids[entity.eid] = dict(entity)

            if entity.eid not in self.types[type]:

                self.types[type].append(entity.eid)

        # Build index with specified key

        func = lambda x: x[key]

        self.build_index(name, type, func, can_be_empty=can_be_empty)