Speed tweaks for pokedex load in SQLite
[zzz-pokedex.git] / pokedex / db / load.py
1 """CSV to database or vice versa."""
2 import csv
3 import fnmatch
4 import os.path
5 import sys
6
7 from sqlalchemy.orm.attributes import instrumentation_registry
8 import sqlalchemy.sql.util
9 import sqlalchemy.types
10
11 from pokedex.db import metadata
12 import pokedex.db.tables as tables
13 from pokedex.defaults import get_default_csv_dir
14
15
16 def _get_table_names(metadata, patterns):
17 """Returns a list of table names from the given metadata. If `patterns`
18 exists, only tables matching one of the patterns will be returned.
19 """
20 if patterns:
21 table_names = set()
22 for pattern in patterns:
23 if '.' in pattern or '/' in pattern:
24 # If it looks like a filename, pull out just the table name
25 _, filename = os.path.split(pattern)
26 table_name, _ = os.path.splitext(filename)
27 pattern = table_name
28
29 table_names.update(fnmatch.filter(metadata.tables.keys(), pattern))
30 else:
31 table_names = metadata.tables.keys()
32
33 return list(table_names)
34
35 def _get_verbose_prints(verbose):
36 """If `verbose` is true, returns three functions: one for printing a
37 starting message, one for printing an interim status update, and one for
38 printing a success or failure message when finished.
39
40 If `verbose` is false, returns no-op functions.
41 """
42
43 if not verbose:
44 # Return dummies
45 def dummy(*args, **kwargs):
46 pass
47
48 return dummy, dummy, dummy
49
50 ### Okay, verbose == True; print stuff
51
52 def print_start(thing):
53 # Truncate to 66 characters, leaving 10 characters for a success
54 # or failure message
55 truncated_thing = thing[0:66]
56
57 # Also, space-pad to keep the cursor in a known column
58 num_spaces = 66 - len(truncated_thing)
59
60 print "%s...%s" % (truncated_thing, ' ' * num_spaces),
61 sys.stdout.flush()
62
63 if sys.stdout.isatty():
64 # stdout is a terminal; stupid backspace tricks are OK.
65 # Don't use print, because it always adds magical spaces, which
66 # makes backspace accounting harder
67
68 backspaces = [0]
69 def print_status(msg):
70 # Overwrite any status text with spaces before printing
71 sys.stdout.write('\b' * backspaces[0])
72 sys.stdout.write(' ' * backspaces[0])
73 sys.stdout.write('\b' * backspaces[0])
74 sys.stdout.write(msg)
75 sys.stdout.flush()
76 backspaces[0] = len(msg)
77
78 def print_done(msg='ok'):
79 # Overwrite any status text with spaces before printing
80 sys.stdout.write('\b' * backspaces[0])
81 sys.stdout.write(' ' * backspaces[0])
82 sys.stdout.write('\b' * backspaces[0])
83 sys.stdout.write(msg + "\n")
84 sys.stdout.flush()
85 backspaces[0] = 0
86
87 else:
88 # stdout is a file (or something); don't bother with status at all
89 def print_status(msg):
90 pass
91
92 def print_done(msg='ok'):
93 print msg
94
95 return print_start, print_status, print_done
96
97
98 def load(session, tables=[], directory=None, drop_tables=False, verbose=False, safe=True):
99 """Load data from CSV files into the given database session.
100
101 Tables are created automatically.
102
103 `session`
104 SQLAlchemy session to use.
105
106 `tables`
107 List of tables to load. If omitted, all tables are loaded.
108
109 `directory`
110 Directory the CSV files reside in. Defaults to the `pokedex` data
111 directory.
112
113 `drop_tables`
114 If set to True, existing `pokedex`-related tables will be dropped.
115
116 `verbose`
117 If set to True, status messages will be printed to stdout.
118
119 `safe`
120 If set to False, load can be faster, but can corrupt the database if
121 it crashes or is interrupted.
122 """
123
124 # First take care of verbosity
125 print_start, print_status, print_done = _get_verbose_prints(verbose)
126
127
128 if directory is None:
129 directory = get_default_csv_dir()
130
131 table_names = _get_table_names(metadata, tables)
132 table_objs = [metadata.tables[name] for name in table_names]
133 table_objs = sqlalchemy.sql.util.sort_tables(table_objs)
134
135 # SQLite speed tweaks
136 if not safe and session.connection().dialect.name == 'sqlite':
137 session.connection().execute("PRAGMA synchronous=OFF")
138 session.connection().execute("PRAGMA journal_mode=OFF")
139
140 # Drop all tables if requested
141 if drop_tables:
142 print_start('Dropping tables')
143 for table in reversed(table_objs):
144 table.drop(checkfirst=True)
145 print_done()
146
147 for table in table_objs:
148 table.create()
149 connection = session.connection()
150
151 # Okay, run through the tables and actually load the data now
152 for table_obj in table_objs:
153 table_name = table_obj.name
154 insert_stmt = table_obj.insert()
155
156 print_start(table_name)
157
158 try:
159 csvpath = "%s/%s.csv" % (directory, table_name)
160 csvfile = open(csvpath, 'rb')
161 except IOError:
162 # File doesn't exist; don't load anything!
163 print_done('missing?')
164 continue
165
166 csvsize = os.stat(csvpath).st_size
167
168 reader = csv.reader(csvfile, lineterminator='\n')
169 column_names = [unicode(column) for column in reader.next()]
170
171 # Self-referential tables may contain rows with foreign keys of other
172 # rows in the same table that do not yet exist. Pull these out and add
173 # them to the session last
174 # ASSUMPTION: Self-referential tables have a single PK called "id"
175 deferred_rows = [] # ( row referring to id, [foreign ids we need] )
176 seen_ids = {} # primary key we've seen => 1
177
178 # Fetch foreign key columns that point at this table, if any
179 self_ref_columns = []
180 for column in table_obj.c:
181 if any(_.references(table_obj) for _ in column.foreign_keys):
182 self_ref_columns.append(column)
183
184 new_rows = []
185 def insert_and_commit():
186 session.connection().execute(insert_stmt, new_rows)
187 session.commit()
188 new_rows[:] = []
189
190 progress = "%d%%" % (100 * csvfile.tell() // csvsize)
191 print_status(progress)
192
193 for csvs in reader:
194 row_data = {}
195
196 for column_name, value in zip(column_names, csvs):
197 column = table_obj.c[column_name]
198 if column.nullable and value == '':
199 # Empty string in a nullable column really means NULL
200 value = None
201 elif isinstance(column.type, sqlalchemy.types.Boolean):
202 # Boolean values are stored as string values 0/1, but both
203 # of those evaluate as true; SQLA wants True/False
204 if value == '0':
205 value = False
206 else:
207 value = True
208 else:
209 # Otherwise, unflatten from bytes
210 value = value.decode('utf-8')
211
212 # nb: Dictionaries flattened with ** have to have string keys
213 row_data[ str(column_name) ] = value
214
215 # May need to stash this row and add it later if it refers to a
216 # later row in this table
217 if self_ref_columns:
218 foreign_ids = [row_data[_.name] for _ in self_ref_columns]
219 foreign_ids = [_ for _ in foreign_ids if _] # remove NULL ids
220
221 if not foreign_ids:
222 # NULL key. Remember this row and add as usual.
223 seen_ids[row_data['id']] = 1
224
225 elif all(_ in seen_ids for _ in foreign_ids):
226 # Non-NULL key we've already seen. Remember it and commit
227 # so we know the old row exists when we add the new one
228 insert_and_commit()
229 seen_ids[row_data['id']] = 1
230
231 else:
232 # Non-NULL future id. Save this and insert it later!
233 deferred_rows.append((row_data, foreign_ids))
234 continue
235
236 # Insert row!
237 new_rows.append(row_data)
238
239 # Remembering some zillion rows in the session consumes a lot of
240 # RAM. Let's not do that. Commit every 1000 rows
241 if len(new_rows) >= 1000:
242 insert_and_commit()
243
244 insert_and_commit()
245
246 # Attempt to add any spare rows we've collected
247 for row_data, foreign_ids in deferred_rows:
248 if not all(_ in seen_ids for _ in foreign_ids):
249 # Could happen if row A refers to B which refers to C.
250 # This is ridiculous and doesn't happen in my data so far
251 raise ValueError("Too many levels of self-reference! "
252 "Row was: " + str(row))
253
254 session.connection().execute(
255 insert_stmt.values(**row_data)
256 )
257 seen_ids[row_data['id']] = 1
258 session.commit()
259
260 print_done()
261
262 # SQLite check
263 if session.connection().dialect.name == 'sqlite':
264 session.connection().execute("PRAGMA integrity_check")
265
266
267
268 def dump(session, tables=[], directory=None, verbose=False):
269 """Dumps the contents of a database to a set of CSV files. Probably not
270 useful to anyone besides a developer.
271
272 `session`
273 SQLAlchemy session to use.
274
275 `tables`
276 List of tables to dump. If omitted, all tables are dumped.
277
278 `directory`
279 Directory the CSV files should be put in. Defaults to the `pokedex`
280 data directory.
281
282 `verbose`
283 If set to True, status messages will be printed to stdout.
284 """
285
286 # First take care of verbosity
287 print_start, print_status, print_done = _get_verbose_prints(verbose)
288
289
290 if not directory:
291 directory = get_default_csv_dir()
292
293 table_names = _get_table_names(metadata, tables)
294 table_names.sort()
295
296
297 for table_name in table_names:
298 print_start(table_name)
299 table = metadata.tables[table_name]
300
301 writer = csv.writer(open("%s/%s.csv" % (directory, table_name), 'wb'),
302 lineterminator='\n')
303 columns = [col.name for col in table.columns]
304 writer.writerow(columns)
305
306 primary_key = table.primary_key
307 for row in session.query(table).order_by(*primary_key).all():
308 csvs = []
309 for col in columns:
310 # Convert Pythony values to something more universal
311 val = getattr(row, col)
312 if val == None:
313 val = ''
314 elif val == True:
315 val = '1'
316 elif val == False:
317 val = '0'
318 else:
319 val = unicode(val).encode('utf-8')
320
321 csvs.append(val)
322
323 writer.writerow(csvs)
324
325 print_done()