pokedex/__init__.py

   1 # encoding: utf8
   2 import sys
   3
   4 from sqlalchemy.exc import IntegrityError
   5 import sqlalchemy.types
   6
   7 from .db import connect, metadata, tables as tables_module
   8 from pokedex.lookup import lookup as pokedex_lookup
   9
  10 def main():
  11     if len(sys.argv) <= 1:
  12         help()
  13
  14     command = sys.argv[1]
  15     args = sys.argv[2:]
  16
  17     # Find the command as a function in this file
  18     func = globals().get("command_%s" % command, None)
  19     if func:
  20         func(*args)
  21     else:
  22         command_help()
  23
  24
  25 def command_csvimport(engine_uri, directory='.'):
  26     import csv
  27
  28     from sqlalchemy.orm.attributes import instrumentation_registry
  29
  30     session = connect(engine_uri)
  31
  32     metadata.create_all()
  33
  34     # SQLAlchemy is retarded and there is no way for me to get a list of ORM
  35     # classes besides to inspect the module they all happen to live in for
  36     # things that look right.
  37     table_base = tables_module.TableBase
  38     orm_classes = {}  # table object => table class
  39
  40     for name in dir(tables_module):
  41         # dir() returns strings!  How /convenient/.
  42         thingy = getattr(tables_module, name)
  43
  44         if not isinstance(thingy, type):
  45             # Not a class; bail
  46             continue
  47         elif not issubclass(thingy, table_base):
  48             # Not a declarative table; bail
  49             continue
  50         elif thingy == table_base:
  51             # Declarative table base, so not a real table; bail
  52             continue
  53
  54         # thingy is definitely a table class!  Hallelujah.
  55         orm_classes[thingy.__table__] = thingy
  56
  57     # Okay, run through the tables and actually load the data now
  58     for table_obj in metadata.sorted_tables:
  59         table_class = orm_classes[table_obj]
  60         table_name = table_obj.name
  61
  62         # Print the table name but leave the cursor in a fixed column
  63         print table_name + '...', ' ' * (40 - len(table_name)),
  64         sys.stdout.flush()
  65
  66         try:
  67             csvfile = open("%s/%s.csv" % (directory, table_name), 'rb')
  68         except IOError:
  69             # File doesn't exist; don't load anything!
  70             print 'no data!'
  71             continue
  72
  73         reader = csv.reader(csvfile, lineterminator='\n')
  74         column_names = [unicode(column) for column in reader.next()]
  75
  76         # Self-referential tables may contain rows with foreign keys of other
  77         # rows in the same table that do not yet exist.  Pull these out and add
  78         # them to the session last
  79         # ASSUMPTION: Self-referential tables have a single PK called "id"
  80         deferred_rows = []  # ( row referring to id, [foreign ids we need] )
  81         seen_ids = {}       # primary key we've seen => 1
  82
  83         # Fetch foreign key columns that point at this table, if any
  84         self_ref_columns = []
  85         for column in table_obj.c:
  86             if any(_.references(table_obj) for _ in column.foreign_keys):
  87                 self_ref_columns.append(column)
  88
  89         for csvs in reader:
  90             row = table_class()
  91
  92             for column_name, value in zip(column_names, csvs):
  93                 column = table_obj.c[column_name]
  94                 if column.nullable and value == '':
  95                     # Empty string in a nullable column really means NULL
  96                     value = None
  97                 elif isinstance(column.type, sqlalchemy.types.Boolean):
  98                     # Boolean values are stored as string values 0/1, but both
  99                     # of those evaluate as true; SQLA wants True/False
 100                     if value == '0':
 101                         value = False
 102                     else:
 103                         value = True
 104                 else:
 105                     # Otherwise, unflatten from bytes
 106                     value = value.decode('utf-8')
 107
 108                 setattr(row, column_name, value)
 109
 110             # May need to stash this row and add it later if it refers to a
 111             # later row in this table
 112             if self_ref_columns:
 113                 foreign_ids = [getattr(row, _.name) for _ in self_ref_columns]
 114                 foreign_ids = [_ for _ in foreign_ids if _]  # remove NULL ids
 115
 116                 if not foreign_ids:
 117                     # NULL key.  Remember this row and add as usual.
 118                     seen_ids[row.id] = 1
 119
 120                 elif all(_ in seen_ids for _ in foreign_ids):
 121                     # Non-NULL key we've already seen.  Remember it and commit
 122                     # so we know the old row exists when we add the new one
 123                     session.commit()
 124                     seen_ids[row.id] = 1
 125
 126                 else:
 127                     # Non-NULL future id.  Save this and insert it later!
 128                     deferred_rows.append((row, foreign_ids))
 129                     continue
 130
 131             session.add(row)
 132
 133         session.commit()
 134
 135         # Attempt to add any spare rows we've collected
 136         for row, foreign_ids in deferred_rows:
 137             if not all(_ in seen_ids for _ in foreign_ids):
 138                 # Could happen if row A refers to B which refers to C.
 139                 # This is ridiculous and doesn't happen in my data so far
 140                 raise ValueError("Too many levels of self-reference!  "
 141                                  "Row was: " + str(row.__dict__))
 142
 143             session.add(row)
 144             seen_ids[row.id] = 1
 145             session.commit()
 146
 147         print 'loaded'
 148
 149 def command_csvexport(engine_uri, directory='.'):
 150     import csv
 151     session = connect(engine_uri)
 152
 153     for table_name in sorted(metadata.tables.keys()):
 154         print table_name
 155         table = metadata.tables[table_name]
 156
 157         writer = csv.writer(open("%s/%s.csv" % (directory, table_name), 'wb'),
 158                             lineterminator='\n')
 159         columns = [col.name for col in table.columns]
 160         writer.writerow(columns)
 161
 162         primary_key = table.primary_key
 163         for row in session.query(table).order_by(*primary_key).all():
 164             csvs = []
 165             for col in columns:
 166                 # Convert Pythony values to something more universal
 167                 val = getattr(row, col)
 168                 if val == None:
 169                     val = ''
 170                 elif val == True:
 171                     val = '1'
 172                 elif val == False:
 173                     val = '0'
 174                 else:
 175                     val = unicode(val).encode('utf-8')
 176
 177                 csvs.append(val)
 178
 179             writer.writerow(csvs)
 180
 181 def command_lookup(engine_uri, name):
 182     # XXX don't require uri!  somehow
 183     session = connect(engine_uri)
 184
 185     results, exact = pokedex_lookup(session, name)
 186     if exact:
 187         print "Matched:"
 188     else:
 189         print "Fuzzy-matched:"
 190
 191     for object in results:
 192         print object.__tablename__, object.name
 193
 194
 195 def command_help():
 196     print u"""pokedex -- a command-line Pokédex interface
 197
 198     help                        Displays this message.
 199     lookup {uri} [name]         Look up something in the Pokédex.
 200
 201   These commands are only useful for developers:
 202     csvimport {uri} [dir]       Import data from a set of CSVs to the database
 203                                   given by the URI.
 204     csvexport {uri} [dir]       Export data from the database given by the URI
 205                                   to a set of CSVs.
 206                                 Directory defaults to cwd.
 207 """.encode(sys.getdefaultencoding(), 'replace')
 208
 209     sys.exit(0)