populate.py


"""
populate.py is run to initially create and then keep updated a local sqlite database from the .json files
maintained at mtgjson.com. It can do this for any .json files, I think, parsing list and dict
data-types as well as normal strings and ints.

The scraping only updates what the json file says has changed after the initial run, so should be pretty low overhead

uses: http://mtgjson.com/json/changelog.json to update card db entries that have changed using the SET updater

after running populate.py, run picfinder.py to check magiccards.info for downloadable pictures of newly added cards
"""
import os
import path
import json
import sqlite3
import requests
import grequests
from collections import Counter
import time
import sys
reload(sys).setdefaultencoding("utf8")


DEBUG = True
__author__ = 'suber1'
__sqlext__ = '.sqlite'
__sqlsets__ = os.getcwd() + os.sep + 'mtg_sets' + __sqlext__
__sqlcards__ = os.getcwd() + os.sep + 'mtg_cards' + __sqlext__
__sqlfiles__ = [__sqlsets__, __sqlcards__]
__mtgpics__ = os.getcwd() + os.sep + 'pics'
__needs__ = os.getcwd() + os.sep + 'needs_pic_links.json'
__jsonsets__ = 'http://mtgjson.com/json/SetCodes.json'
__jsonupdate__ = 'http://mtgjson.com/json/changelog.json'
__one_set__ = 'http://mtgjson.com/json/{}.json'
__req_limit__ = 21          # can the web server handle getting pounded by this many?
__test_quant__ = 0          # set to zero for normal full run
__max_errors__ = 0          # set positive to explore new import data
__last_update__ = os.getcwd() + os.sep + 'last_update.json'
__set_hdr_excluded__ = [u'cards', u'booster']
__cards_hdr_excluded__ = [u'booster', u'foreignNames']
__newness__ = [u"newSetFiles", u"updatedSetFiles"]
__sets_key__ = u'code'
__sets_t__ = 'set_infos'
__cards_key__ = u'id'
__cards_t__ = 'cards'
__cards_dk__ = u'cards'
__types__ = {"<type 'list'>": 'json',
             "<type 'unicode'>": 'TEXT',
             "<type 'str'>": 'TEXT',
             "<type 'float'>": 'REAL',
             "<type 'int'>": 'INTEGER',
             "<type 'dict'>": 'json',
             "<type 'bool'>": 'TEXT',
             "<type 'buffer'>": 'BLOB'}


class DBMagic (object):
    """
    DBcolumns = {db_tablename: '''CREATE TABLE db_tablename (column_name1 data_type PRIMARY KEY?,
                                column_name2 data_type, )''', ...}
    user: get the columns from the json entry for a card, or make up your own
    """
    def __init__(self, DBfn=None, DBcolumns=None, DB_DEBUG=False):
        self.DB_DEBUG = DB_DEBUG
        self.DBfn = DBfn
        self.DBcolumns = DBcolumns
        if self.DBfn is None:
            self.DBfn = os.path.join(os.path.expanduser('~'), 'Desktop', "MagicDB", __sqlext__)
            print("WARNING, creating/using a default database: {}".format(self.DBfn))
        if not os.path.isdir(os.path.dirname(self.DBfn)):
            os.makedirs(os.path.dirname(self.DBfn))
        sqlite3.register_converter("json", json.loads)
        sqlite3.register_adapter(list, json.dumps)
        sqlite3.register_adapter(dict, json.dumps)
        self.con = sqlite3.connect(self.DBfn, detect_types=sqlite3.PARSE_DECLTYPES|sqlite3.PARSE_COLNAMES)
        self.con.row_factory = sqlite3.Row
        self.con.text_factory = sqlite3.OptimizedUnicode
        self.cur = self.con.cursor()
        self.newDB = False
        # check that tables exist. if not, make them
        for t, v in self.DBcolumns.viewitems():
            if not self.cur.execute('''PRAGMA table_info ('{}')'''.format(t)).fetchall():
                self.cur.execute(v)
                self.con.commit()
                print("Created new table: {}".format(t))
                self.newDB = True
            else:
                print("using existing table: {} ".format(t))
            print("in file: {}".format(self.DBfn))
        self.tables = [a[0] for a in
                       self.cur.execute('''SELECT name FROM sqlite_master WHERE type='table' ''').fetchall()]

    def show_columns(self, tablename):
        return [t[1] for t in self.cur.execute('''PRAGMA table_info ('{}')'''.format(tablename)).fetchall()]

    def add_columns(self, tableup, column_map):
        """
        Parameters
        ----------
        tableup: the table to which we are adding columns
        column_map: {'column_foo': sqlite database-type, as string}
                          eg  {'column-name': 'INTEGER', ...}
        """
        present_columns = self.show_columns(tableup)
        for newcol, sql_dtype in column_map.viewitems():
            if newcol not in present_columns:
                self.cur.execute('''ALTER TABLE {} ADD {} {}'''.format(tableup, newcol, sql_dtype))
                if self.DB_DEBUG:
                    print("added column: '{}' of type: '{}' to table: {}".format(newcol, sql_dtype, tableup))
        self.con.commit()

    def add_data(self, data, tbl, key_column=None):
        """
        populate database with list-of-dict values for dict.keys() that are in db-columns
        Parameters
        ----------
        data - list of dict cum json objects whose wanted keys have been made DB column names
        tbl - db_table to which we add this data
        key - the sqlite PRIMARY KEY that the table index is being done on

        Returns
        -------
        side effect = database entry. Made an UPSERT from strings that accepts any number of key: vals
        as long as keys are column names. sqlite3 has no native UPSERT command.
        """
        n, error_count = -1, 0
        if key_column is None:
            hdrs = self.cur.execute('''PRAGMA table_info ({})'''.format(tbl)).fetchall()
            for h in hdrs:
                if bool(h[5]):
                    key_column = h[1]
                    if self.DB_DEBUG:
                        print("guessing primary key is: {}".format(key_column))

        approved_columns = self.show_columns(tbl)

        def hunk(c):
            return c + "=(?)"

        for n, line in enumerate(data):
            line_item = {k: v for k, v in line.viewitems() if k in approved_columns}
            SQL1 = '''INSERT OR IGNORE INTO {}({}) VALUES({})'''.format(str(tbl), ', '.join(line_item.keys()),
                                                                        ':' + ', :'.join(line_item.keys()))
            SQL2 = '''UPDATE {} SET {} WHERE changes()=0 and {}=("{}")'''\
                .format(str(tbl), ", ".join(map(hunk, line_item.keys())), key_column, line_item[key_column])
            try:
                self.cur.execute(SQL1, line_item)
                self.cur.execute(SQL2, line_item.values())
            except (sqlite3.OperationalError, sqlite3.InterfaceError, sqlite3.ProgrammingError) as e:
                error_count += 1
                print("error: {} *****table: {} ***  row#{}   #items={}  >>> {}  ***"
                      .format(error_count, tbl, n, len(line_item), e))
                print(SQL1, SQL2)
                print(line_item)
                if error_count > __max_errors__:
                    print("** data entry has too many problems to ignore. exiting, harshly **")
                    exit(1)
                continue
        self.con.commit()
        return n, error_count


createstr = '''CREATE TABLE {} ({} TEXT PRIMARY KEY)'''
set_db = DBMagic(DBfn=__sqlsets__,
                 DBcolumns={__sets_t__: createstr.format(__sets_t__, __sets_key__)},
                 DB_DEBUG=DEBUG)
card_db = DBMagic(DBfn=__sqlcards__,
                  DBcolumns={__cards_t__: createstr.format(__cards_t__, __cards_key__)},
                  DB_DEBUG=DEBUG)


def asynch_getter(unsent, groupsize_limit=None, DBG=DEBUG):
    """
    Parameters
    ----------
    unsent: generator-prepared requests objects that are not yet responses
    groupsize_limit: can keep us from spamming web server. 12-at-a-time works

    Returns
    -------
    flat list of 'response' objects from each URL contained in unsent
    """
    if groupsize_limit is None:
        groupsize_limit = len(unsent)
    d = []
    while unsent:
        bunch = []
        groupsize = min(len(unsent), groupsize_limit)
        for x in xrange(groupsize):
            bunch.append(unsent.pop())
        if DBG:
            print("requesting from {} urls".format(groupsize))
        d.extend(grequests.map(bunch))
    return d


def column_parser(datas, exclusions=None, DEBUG=DEBUG):
    """
    called by column_type_parser() now. It filters out unwanted columns so they don't
    get into the database. Just for peace-of-mind in avoiding bloat.
    Parameters
    ----------
    datas: a list of dicts, possibly derived from downloaded json card data
    exclusions: the keys from datas we don't want in the database

    Returns
    -------
    keys_from_json: Counter object with undesirable keys set to 0 (BUT NOT DELETED!)
    """
    if exclusions is None:
        exclusions = []
        print("using all parsed column names; no exclusions")
    keys_from_json = Counter()
    for i, d in enumerate(datas):
        kks = d.keys()
        for k in kks:
            keys_from_json[k] += 1
    for e in exclusions:
        if e in keys_from_json.keys():
            keys_from_json[e] = 0
    return keys_from_json


def column_type_parser(datas, exclusions=None, types_map=None, DEBUG=DEBUG):
    """
    this supplies columns & types that need to be added to db from import data
    .json is a custom sqlite type here so it can 'natively' handle nested dict & list!
    datas: your list of dict containing the data that may not have column-headers (keys) in db
    types_map: user created {'python type() output as string': 'sqlite-data-type', ...}
    - returns -
    type_defs: {'column-name': 'appropriate sqlite-data-type', ...}
    """
    if types_map is None:
        types_map = __types__
    if exclusions is None:
        exclusions = []

    type_defs = {}
    usable_columns = [c for c, quant in column_parser(datas, exclusions=exclusions).viewitems() if quant > 0]
    for k in usable_columns:
        for e in datas:
            if k in e.keys():
                try:
                    type_defs[k] = types_map[str(type(e[k]))]
                except KeyError as err:
                    print("{} - is the key, so add to the python-to-sqlite type definitions".format(err))
                    print("PROBLEM: {}".format(e[k]))
                    exit()
                break
    return type_defs


def check_for_updates(update_url, oldfn, need_all=False, DBG=DEBUG):
    """
    Parameters
    ----------
    update_url = url for json-encoded changes to the mtgjson.com data
    oldfn = local file-name for .json file showing the most recent changes incorporated locally

    Returns
    -------
    list of unique set-codes that need to be updated since the last successful check
    """
    newness = __newness__
    check_codes = []
    most_recent = u"0.0.0"
    # see if there have been previous updates recorded in local files
    try:
        with open(oldfn, mode='rb') as fob:
            old_stuff = json.load(fob)
            most_recent = old_stuff[0][u"version"]
    except Exception as e:
        print("{}: couldn't use changes log file ({})".format(e, oldfn))
        need_all = True
    if need_all:
        most_recent = u"0.0.0"
    old_version = [int(a) for a in most_recent.split(u".")]
    # get update file from web
    try:
        req_new = requests.get(update_url).json()
    except Exception as e:
        print("{}: couldn't obtain recent updates from: {}".format(e, update_url))
        return check_codes
    print("latest updates version processed locally: {} \n   *   *   *  newest version available: {}"
          .format(most_recent, req_new[0][u'version']))
    i = 0
    for i in xrange(len(req_new)):
        new_version = [int(a) > b for a, b in zip(req_new[i][u'version'].split(u"."), old_version)]
        if any(new_version):
            for tag in newness:
                if tag in req_new[i].keys():
                    check_codes.extend(req_new[i][tag])
        else:  # versions are indexed newest to oldest
            break
    # save the update file locally for comparison next time
    try:
        with open(oldfn, mode='wb') as wob:
            json.dump(req_new, wob)
    except Exception as e:
        print("{}: couldn't save the changes log as {}".format(e, oldfn))

    check_codes = list(set(check_codes))
    if DBG:
        print("there are {} unprocessed updates from {} ".format(i, update_url))
    return check_codes


def info_grubber(unsent_greq, datas=None, tries=10):
    """
    don't (always) take no for an answer
    """
    if datas is None:
        datas = []
    fails = []
    while tries:
        fails = []
        for n, d in enumerate(asynch_getter(unsent_greq, groupsize_limit=__req_limit__)):
            if d.status_code == 200:
                datas.append(d.json())
            else:
                fails.append(d.url)
        if fails:
            print("< {} > retrying {} urls ".format(tries, len(fails)) + "".join('*' * tries))
            time.sleep(.4)
            tries -= 1
            unsent_greq = [grequests.get(u) for u in fails]
        else:
            tries = 0
    return datas, fails


def xando(setcodes, mk='-x'):
    """
    use each setcode only once, so if & only if an 'ABC-x' version exists, remove the 'ABC' version
    """
    keepers, stripped, stumpys = [], [], []
    for x in setcodes:
        if mk in x:
            keepers.append(x)
            stripped.append(x.split(mk)[0])
        else:
            stumpys.append(x)
    for y in stumpys:
        if y not in stripped:
            keepers.append(y)
    return keepers


def deckify(dats):
    """ add a column 'code' to make set queries a bit easier
        # put all cards from all sets into one flat list"""
    deck = []
    for e in dats:
        for cardline in e[__cards_dk__]:
            cardline.setdefault(u'code', e[__sets_key__])
            deck.append(cardline)
    return deck


def starting_sets(sql_dbs):
    """
    looks at web for all required set-codes that need to be added / updated
    """
    setcodes, needs = [], False
    if not len(sql_dbs):
        print("setup: missing required database files: {}".format([f for f in __sqlfiles__ if f not in sql_dbs]))
        print("Getting all set codes from: {}".format(__jsonsets__))
        setcodes = requests.get(__jsonsets__).json()
        needs = True
    print("Checking for new or updated set codes at: {}".format(__jsonupdate__))
    return list(set(setcodes + check_for_updates(__jsonupdate__, __last_update__, need_all=needs)))


def bootup():
    """
    Returns the bases and paths of required items needed to run this thing.
    """
    homedir = os.getcwd()
    if not os.path.isdir(__mtgpics__):
        os.mkdir(__mtgpics__)
    picdir = __mtgpics__
    return homedir, picdir, \
           [unicode(q.basename()) for q in path.path(homedir).files('*' + __sqlext__)], \
           [unicode(p.basename()) for p in path.path(picdir).dirs()]


def main():
    homedir, picdir, sqldbs, picsets = bootup()

    if DEBUG:
        print("using existing sqlite db fns: {}".format(sqldbs))

    urls = [__one_set__.format(s) for s in xando(starting_sets(sqldbs))]
    unsent = [grequests.get(u) for u in urls[:(__test_quant__ or len(urls))]]

    datas, fails = info_grubber(unsent, tries=8)

    for p in fails:
        print("url didn't play nice: {}".format(p))

    # aggregate all the columns (while assigning their data-types) for each data-set
    columns_for_sets = column_type_parser(datas, exclusions=__set_hdr_excluded__)
    deck = deckify(datas)
    columns_for_cards = column_type_parser(deck, exclusions=__cards_hdr_excluded__)

    if DEBUG:
        print("'set_infos' columns to update: {}".format(columns_for_sets))
        print("'cards' columns to update: {}".format(columns_for_cards))

    # adding the approved column-headers to existing tables (if not already present)
    set_db.add_columns(set_db.tables[0], columns_for_sets)
    card_db.add_columns(card_db.tables[0], columns_for_cards)

    # should be ready to add the data to local database!
    a, b = card_db.add_data(deck, card_db.tables[0], key_column=__cards_key__)
    print("processed cards: {}, import errors: {}".format(a+1, b))
    c, d = set_db.add_data(datas, set_db.tables[0], key_column=__sets_key__)
    print("processed sets: {}, import errors: {}".format(c+1, d))
    return 1

if __name__ == "__main__":
    exit(main())