18 |
# $Source$ |
# $Source$ |
19 |
# $Id$ |
# $Id$ |
20 |
|
|
21 |
import os |
# Pysqlite version 1. and 2. behaves quiet differently |
22 |
import weakref |
# Pysqlite uses a different paramstyle. The older version |
23 |
from sqlite import connect |
# support format and pyformat while pysqlite2 supports only qmark |
24 |
|
# and named. |
25 |
|
# The sqlite2 boolean variable is used to manage specific part of the code |
26 |
|
try: |
27 |
|
# Using SQLITE 2.x |
28 |
|
sqlite2 = True |
29 |
|
from pysqlite2 import dbapi2 as sqlite |
30 |
|
except ImportError: |
31 |
|
# Using SQLITE 1.x |
32 |
|
sqlite2 = False |
33 |
|
import sqlite |
34 |
|
|
35 |
|
|
36 |
|
from base import TitledObject |
37 |
|
|
38 |
import table |
import table |
39 |
|
|
53 |
|
|
54 |
def __init__(self, filename): |
def __init__(self, filename): |
55 |
self.filename = filename |
self.filename = filename |
56 |
self.conn = connect(filename) |
self.conn = sqlite.connect(filename) |
57 |
# Counters to produce unique table and column names |
# Counters to produce unique table and column names |
58 |
self.num_tables = 0 |
self.num_tables = 0 |
59 |
self.num_cols = 0 |
self.num_cols = 0 |
98 |
self.internal_name = internal_name |
self.internal_name = internal_name |
99 |
|
|
100 |
|
|
101 |
class TransientTableBase(table.OldTableInterfaceMixin): |
class TransientTableBase: |
102 |
|
|
103 |
"""Base class for tables in the transient database""" |
"""Base class for tables in the transient database""" |
104 |
|
|
173 |
""" |
""" |
174 |
return self.column_map.has_key(col) |
return self.column_map.has_key(col) |
175 |
|
|
176 |
def ReadRowAsDict(self, index): |
def RowIdToOrdinal(self, gid): |
177 |
|
"""Return the row ordinal given its id |
178 |
|
|
179 |
|
At the moment the transient tables are only used for tables that |
180 |
|
don't distinguish between row number and row id, so the value is |
181 |
|
returned unchanged. |
182 |
|
""" |
183 |
|
return gid |
184 |
|
|
185 |
|
def RowOrdinalToId(self, num): |
186 |
|
"""Return the rowid for given its ordinal |
187 |
|
|
188 |
|
At the moment the transient tables are only used for tables that |
189 |
|
don't distinguish between row number and row id, so the value is |
190 |
|
returned unchanged. |
191 |
|
""" |
192 |
|
return num |
193 |
|
|
194 |
|
def ReadRowAsDict(self, index, row_is_ordinal = 0): |
195 |
|
"""Return the entire row as a dictionary with column names as keys |
196 |
|
|
197 |
|
The row_is_ordinal is ignored because at the moment the |
198 |
|
transient tables are only used for DBF files where it doesn't |
199 |
|
matter. |
200 |
|
""" |
201 |
# Implementation Strategy: Executing a completely new select |
# Implementation Strategy: Executing a completely new select |
202 |
# statement every time this method is called is too slow. The |
# statement every time this method is called is too slow. The |
203 |
# most important usage is to read the records more or less |
# most important usage is to read the records more or less |
245 |
self.read_record_last_row = index |
self.read_record_last_row = index |
246 |
return dict(zip(self.orig_names, result)) |
return dict(zip(self.orig_names, result)) |
247 |
|
|
248 |
def ReadValue(self, row, col): |
def ReadValue(self, row, col, row_is_ordinal = 0): |
249 |
"""Return the value of the specified row and column |
"""Return the value of the specified row and column |
250 |
|
|
251 |
The col parameter may be the index of the column or its name. |
The col parameter may be the index of the column or its name. |
252 |
|
|
253 |
|
The row_is_ordinal is ignored because at the moment the |
254 |
|
transient tables are only used for DBF files where it doesn't |
255 |
|
matter. |
256 |
""" |
""" |
257 |
# Depending on the actual access patterns of the table data, it |
# Depending on the actual access patterns of the table data, it |
258 |
# might be a bit faster in some circumstances to not implement |
# might be a bit faster in some circumstances to not implement |
261 |
return self.ReadRowAsDict(row)[self.column_map[col].name] |
return self.ReadRowAsDict(row)[self.column_map[col].name] |
262 |
|
|
263 |
def ValueRange(self, col): |
def ValueRange(self, col): |
264 |
|
# Performance notes: |
265 |
|
# |
266 |
|
# In sqlite 2.8.6 the min and max aggregate functions can use an |
267 |
|
# index but only when used as the only expression in the select |
268 |
|
# statement (i.e. 'select min(col), max(col) from tbl;' will not |
269 |
|
# use the index but 'select min(col) from tbl;' will) so we |
270 |
|
# query the minimum and maximum separately. |
271 |
|
# |
272 |
|
# With the separate statements we can take advantage of an index |
273 |
|
# if it exists. If the index doesn't exist, creating it first |
274 |
|
# and then using it in the query is slower than the queries |
275 |
|
# without an index. Creating the index is only an advantage if |
276 |
|
# the queries are performed multiple times. With the current use |
277 |
|
# patterns where ValueRange is only used occasionally by the |
278 |
|
# classification generation dialog creating the index only for |
279 |
|
# this usage is not really worth it, so we don't. |
280 |
col = self.column_map[col] |
col = self.column_map[col] |
281 |
iname = col.internal_name |
iname = col.internal_name |
282 |
min, max = self.db.execute("SELECT min(%s), max(%s) FROM %s;" |
min = self.db.execute("SELECT min(%s) FROM %s;" |
283 |
% (iname, iname, self.tablename)) |
% (iname, self.tablename))[0] |
284 |
|
max = self.db.execute("SELECT max(%s) FROM %s;" |
285 |
|
% (iname, self.tablename))[0] |
286 |
converter = type_converter_map[col.type] |
converter = type_converter_map[col.type] |
287 |
return (converter(min), converter(max)) |
return (converter(min), converter(max)) |
288 |
|
|
289 |
def UniqueValues(self, col): |
def UniqueValues(self, col): |
290 |
|
# Performance notes: |
291 |
|
# |
292 |
|
# In sqlite 2.8.6 there doesn't seem to be a way to query the |
293 |
|
# unique items that uses an index. I've tried |
294 |
|
# |
295 |
|
# SELECT col FROM tbl GROUP BY col; |
296 |
|
# |
297 |
|
# and |
298 |
|
# |
299 |
|
# SELECT DISTINCT col FROM tbl; |
300 |
|
# |
301 |
|
# and in both cases the index is not used. If the index isn't |
302 |
|
# used it doesn't make sense to call self.ensure_index. |
303 |
iname = self.column_map[col].internal_name |
iname = self.column_map[col].internal_name |
304 |
cursor = self.db.cursor() |
cursor = self.db.cursor() |
305 |
cursor.execute("SELECT %s FROM %s GROUP BY %s;" |
cursor.execute("SELECT %s FROM %s GROUP BY %s;" |
312 |
result.append(row[0]) |
result.append(row[0]) |
313 |
return result |
return result |
314 |
|
|
315 |
|
def Width(self, col): |
316 |
|
"""Return the maximum width of values in the column |
317 |
|
|
318 |
|
The return value is the the maximum length of string |
319 |
|
representation of the values in the column (represented by index |
320 |
|
or name). |
321 |
|
""" |
322 |
|
max = 0 |
323 |
|
|
324 |
|
type = self.column_map[col].type |
325 |
|
iname = self.column_map[col].internal_name |
326 |
|
cursor = self.db.cursor() |
327 |
|
cursor.execute("SELECT %s FROM %s;" % (iname, self.tablename)) |
328 |
|
values = [ i[0] for i in cursor.fetchall()] |
329 |
|
if not values: |
330 |
|
return None |
331 |
|
|
332 |
|
if type == table.FIELDTYPE_DOUBLE: |
333 |
|
format = "%.12f" |
334 |
|
elif type == table.FIELDTYPE_INT: |
335 |
|
format = "%d" |
336 |
|
else: |
337 |
|
format = "%s" |
338 |
|
for value in values: |
339 |
|
if value is None: continue |
340 |
|
l = len(format % value) |
341 |
|
if l > max: |
342 |
|
max = l |
343 |
|
|
344 |
|
return max |
345 |
|
|
346 |
def SimpleQuery(self, left, comparison, right): |
def SimpleQuery(self, left, comparison, right): |
347 |
"""Return the indices of all rows that matching a condition. |
"""Return the indices of all rows that matching a condition. |
348 |
|
|
366 |
right_template = right.internal_name |
right_template = right.internal_name |
367 |
params = () |
params = () |
368 |
else: |
else: |
369 |
right_template = "%s" |
if sqlite2: |
370 |
|
right_template = "?" |
371 |
|
else: right_template = "%s" |
372 |
params = (right,) |
params = (right,) |
373 |
|
|
374 |
query = "SELECT id FROM %s WHERE %s %s %s ORDER BY id;" \ |
query = "SELECT id FROM %s WHERE %s %s %s ORDER BY id;" \ |
385 |
result.append(row[0]) |
result.append(row[0]) |
386 |
return result |
return result |
387 |
|
|
388 |
|
def Dependencies(self): |
389 |
|
"""Placeholder for a method in a derived class. |
390 |
|
|
391 |
|
Return a sequence with the tables and other data objects that |
392 |
|
self depends on. |
393 |
|
""" |
394 |
|
raise NotImplementedError |
395 |
|
|
396 |
|
|
397 |
class TransientTable(TransientTableBase): |
class TransientTable(TitledObject, TransientTableBase): |
398 |
|
|
399 |
"""A Table in a transient DB that starts as the copy of a Thuban Table.""" |
"""A Table in a transient DB that starts as the copy of a Thuban Table.""" |
400 |
|
|
405 |
interface. |
interface. |
406 |
""" |
""" |
407 |
TransientTableBase.__init__(self, transient_db) |
TransientTableBase.__init__(self, transient_db) |
408 |
|
TitledObject.__init__(self, table.Title()) |
409 |
self.create(table) |
self.create(table) |
410 |
|
|
411 |
def create(self, table): |
def create(self, table): |
423 |
# longer than any of the column names |
# longer than any of the column names |
424 |
id_key = max([len(col.name) for col in self.columns]) * "x" |
id_key = max([len(col.name) for col in self.columns]) * "x" |
425 |
|
|
426 |
insert_template = "INSERT INTO %s (id, %s) VALUES (%%(%s)s, %s);" \ |
if sqlite2: |
427 |
|
insert_template = "INSERT INTO %s (id, %s) VALUES (%s, %s);" \ |
428 |
|
% (self.tablename, |
429 |
|
", ".join([col.internal_name |
430 |
|
for col in self.columns]), |
431 |
|
'?', |
432 |
|
", ".join(["?" for col in self.columns])) |
433 |
|
|
434 |
|
else: |
435 |
|
insert_template = "INSERT INTO %s (id, %s) VALUES (%%(%s)s, %s);" \ |
436 |
% (self.tablename, |
% (self.tablename, |
437 |
", ".join([col.internal_name |
", ".join([col.internal_name |
438 |
for col in self.columns]), |
for col in self.columns]), |
443 |
for i in range(table.NumRows()): |
for i in range(table.NumRows()): |
444 |
row = table.ReadRowAsDict(i) |
row = table.ReadRowAsDict(i) |
445 |
row[id_key] = i |
row[id_key] = i |
446 |
cursor.execute(insert_template, row) |
if sqlite2: |
447 |
|
params = [i] |
448 |
|
for col in self.columns: |
449 |
|
params.append(row[col.name]) |
450 |
|
cursor.execute(insert_template, params) |
451 |
|
else: |
452 |
|
cursor.execute(insert_template, row) |
453 |
self.db.conn.commit() |
self.db.conn.commit() |
454 |
|
|
455 |
|
|
456 |
|
|
457 |
class TransientJoinedTable(TransientTableBase): |
class TransientJoinedTable(TitledObject, TransientTableBase): |
458 |
|
|
459 |
"""A Table in the transient DB that contains a join of two tables""" |
"""A Table in the transient DB that contains a join of two tables""" |
460 |
|
|
461 |
def __init__(self, transient_db, left_table, left_field, |
def __init__(self, transient_db, left_table, left_field, |
462 |
right_table, right_field = None): |
right_table, right_field = None, outer_join = False): |
463 |
"""Create a new table in the transient DB as a join of two tables. |
"""Create a new table in the transient DB as a join of two tables. |
464 |
|
|
465 |
Both input tables, left_table and right_table must have a |
Both input tables, left_table and right_table must have a |
466 |
transient_table method that returns a table object for a table |
transient_table method that returns a table object for a table |
467 |
in the trnsient database. The join is performed on the condition |
in the transient database. The join is performed on the condition |
468 |
that the value of the left_field column the the left table is |
that the value of the left_field column the the left table is |
469 |
equal to the value of the right_field in the right_table. |
equal to the value of the right_field in the right_table. |
470 |
|
|
471 |
The joined table contains all columns of the input tables with |
The joined table contains all columns of the input tables, |
472 |
one exception: Any column in the right_table with the same name |
however, the column names of the right table may be changed |
473 |
as one of the columns in the left_table will be omitted. This is |
slightly to make them unique in the joined table. This is |
474 |
somewhat of an implementation detail, but is done so that the |
currently done by appending a sufficient number of underscores |
475 |
column names of the joined table can be the same as the column |
('_'). |
|
names of the input tables without having to create prefixes. |
|
476 |
""" |
""" |
477 |
TransientTableBase.__init__(self, transient_db) |
TransientTableBase.__init__(self, transient_db) |
478 |
|
self.dependencies = (left_table, right_table) |
479 |
self.left_table = left_table.transient_table() |
self.left_table = left_table.transient_table() |
480 |
self.left_field = left_field |
self.left_field = left_field |
481 |
self.right_table = right_table.transient_table() |
self.right_table = right_table.transient_table() |
483 |
self.right_field = right_field |
self.right_field = right_field |
484 |
else: |
else: |
485 |
self.right_field = self.left_field |
self.right_field = self.left_field |
486 |
|
self.outer_join = outer_join |
487 |
|
|
488 |
|
title = "Join of %(left)s and %(right)s" \ |
489 |
|
% {"left": self.left_table.Title(), |
490 |
|
"right": self.right_table.Title()} |
491 |
|
TitledObject.__init__(self, title) |
492 |
|
|
493 |
self.create() |
self.create() |
494 |
|
|
495 |
def create(self): |
def create(self): |
498 |
|
|
499 |
self.right_table.ensure_index(self.right_field) |
self.right_table.ensure_index(self.right_field) |
500 |
|
|
501 |
|
# determine the internal column names to join on before |
502 |
|
# coalescing the column information because if the external |
503 |
|
# column names are the same they will be mapped to the same |
504 |
|
# internal name afterwards. |
505 |
|
internal_left_col = self.left_table.orig_to_internal[self.left_field] |
506 |
|
internal_right_col =self.right_table.orig_to_internal[self.right_field] |
507 |
|
|
508 |
# Coalesce the column information |
# Coalesce the column information |
509 |
visited = {} |
visited = {} |
510 |
columns = [] |
columns = [] |
511 |
for col in self.left_table.columns + self.right_table.columns: |
newcolumns = [] |
512 |
if col.name in visited: |
for table in (self.left_table, self.right_table): |
513 |
# We can't allow multiple columns with the same original |
for col in table.Columns(): |
514 |
# name, so omit this one. FIXME: There should be a |
colname = col.name |
515 |
# better solution. |
# We can't allow multiple columns with the same |
516 |
continue |
# original name, so append '_' to this one until |
517 |
columns.append(col) |
# it is unique. |
518 |
TransientTableBase.create(self, columns) |
# FIXME: There should be a better solution. |
519 |
|
while colname in visited: |
520 |
|
colname = colname + '_' |
521 |
|
columns.append((table.tablename, col)) |
522 |
|
newcol = ColumnReference(colname, col.type, |
523 |
|
"Col%03d" % (len(newcolumns)+1)) |
524 |
|
newcolumns.append(newcol) |
525 |
|
visited[colname] = 1 |
526 |
|
TransientTableBase.create(self, newcolumns) |
527 |
|
|
528 |
# Copy the joined data to the table. |
# Copy the joined data to the table. |
529 |
internal_names = [col.internal_name for col in self.columns] |
newinternal_names = [col.internal_name for col in self.columns] |
530 |
|
internal_references = ["%s.%s" % (table, col.internal_name) |
531 |
|
for table, col in columns] |
532 |
|
if self.outer_join: |
533 |
|
join_operator = 'LEFT OUTER JOIN' |
534 |
|
else: |
535 |
|
join_operator = 'JOIN' |
536 |
stmt = ("INSERT INTO %s (id, %s) SELECT %s.id, %s FROM %s" |
stmt = ("INSERT INTO %s (id, %s) SELECT %s.id, %s FROM %s" |
537 |
" JOIN %s ON %s = %s;" |
" %s %s ON %s.%s = %s.%s;" |
538 |
% (self.tablename, |
% (self.tablename, |
539 |
", ".join(internal_names), |
", ".join(newinternal_names), |
540 |
self.left_table.tablename, |
self.left_table.tablename, |
541 |
", ".join(internal_names), |
", ".join(internal_references), |
542 |
self.left_table.tablename, |
self.left_table.tablename, |
543 |
|
join_operator, |
544 |
self.right_table.tablename, |
self.right_table.tablename, |
545 |
self.orig_to_internal[self.left_field], |
self.left_table.tablename, |
546 |
self.orig_to_internal[self.right_field])) |
internal_left_col, |
547 |
|
self.right_table.tablename, |
548 |
|
internal_right_col)) |
549 |
self.db.execute(stmt) |
self.db.execute(stmt) |
550 |
|
|
551 |
|
def Dependencies(self): |
552 |
|
"""Return a tuple with the two tables the join depends on.""" |
553 |
|
return self.dependencies |
554 |
|
|
555 |
|
def JoinType(self): |
556 |
|
"""Return the type of the join (either 'INNER' or 'LEFT OUTER')""" |
557 |
|
if self.outer_join: |
558 |
|
return "LEFT OUTER" |
559 |
|
else: |
560 |
|
return "INNER" |
561 |
|
|
562 |
|
|
563 |
class AutoTransientTable(table.OldTableInterfaceMixin): |
class AutoTransientTable(TitledObject): |
564 |
|
|
565 |
"""Table that copies data to a transient table on demand. |
"""Table that copies data to a transient table on demand. |
566 |
|
|
569 |
""" |
""" |
570 |
|
|
571 |
def __init__(self, transient_db, table): |
def __init__(self, transient_db, table): |
572 |
|
TitledObject.__init__(self, table.Title()) |
573 |
self.transient_db = transient_db |
self.transient_db = transient_db |
574 |
self.table = table |
self.table = table |
575 |
self.t_table = None |
self.t_table = None |
591 |
def NumColumns(self): |
def NumColumns(self): |
592 |
return self.table.NumColumns() |
return self.table.NumColumns() |
593 |
|
|
594 |
def ReadRowAsDict(self, record): |
def RowIdToOrdinal(self, gid): |
595 |
|
"""Return the row ordinal given its id""" |
596 |
|
if self.t_table is not None: |
597 |
|
return self.t_table.RowIdToOrdinal(gid) |
598 |
|
else: |
599 |
|
return self.table.RowIdToOrdinal(gid) |
600 |
|
|
601 |
|
def RowOrdinalToId(self, num): |
602 |
|
"""Return the rowid for given its ordinal""" |
603 |
|
if self.t_table is not None: |
604 |
|
return self.t_table.RowOrdinalToId(num) |
605 |
|
else: |
606 |
|
return self.table.RowOrdinalToId(num) |
607 |
|
|
608 |
|
def ReadRowAsDict(self, record, row_is_ordinal = 0): |
609 |
"""Return the record no. record as a dict mapping field names to values |
"""Return the record no. record as a dict mapping field names to values |
610 |
""" |
""" |
611 |
if self.t_table is not None: |
if self.t_table is not None: |
612 |
return self.t_table.ReadRowAsDict(record) |
return self.t_table.ReadRowAsDict(record, |
613 |
|
row_is_ordinal = row_is_ordinal) |
614 |
else: |
else: |
615 |
return self.table.ReadRowAsDict(record) |
return self.table.ReadRowAsDict(record, |
616 |
|
row_is_ordinal = row_is_ordinal) |
617 |
|
|
618 |
def ReadValue(self, row, col): |
def ReadValue(self, row, col, row_is_ordinal = 0): |
619 |
"""Return the value of the specified row and column |
"""Return the value of the specified row and column |
620 |
|
|
621 |
The col parameter may be the index of the column or its name. |
The col parameter may be the index of the column or its name. |
622 |
""" |
""" |
623 |
if self.t_table is not None: |
if self.t_table is not None: |
624 |
return self.t_table.ReadValue(row, col) |
return self.t_table.ReadValue(row, col, |
625 |
|
row_is_ordinal = row_is_ordinal) |
626 |
else: |
else: |
627 |
return self.table.ReadValue(row, col) |
return self.table.ReadValue(row, col, |
628 |
|
row_is_ordinal = row_is_ordinal) |
629 |
|
|
630 |
def copy_to_transient(self): |
def copy_to_transient(self): |
631 |
"""Internal: Create a transient table and copy the data into it""" |
"""Internal: Create a transient table and copy the data into it""" |
640 |
return self.t_table |
return self.t_table |
641 |
|
|
642 |
def ValueRange(self, col): |
def ValueRange(self, col): |
643 |
if self.t_table is None: |
# Performance of sqlite vs. DBF for this method: |
644 |
self.copy_to_transient() |
# |
645 |
return self.t_table.ValueRange(col) |
# If the table has been copied to the sqlite database it's |
646 |
|
# faster to use it even if there is no index on that column. |
647 |
|
# Otherwise it's faster to simply loop through all rows in the |
648 |
|
# DBF file. Copying the data to the sqlite database can take |
649 |
|
# very long for large amounts of data |
650 |
|
# |
651 |
|
# Of course if the table is not a DBF file the issue could be |
652 |
|
# different, although copying the data into sqlite first will |
653 |
|
# likely always be slower than simply querying the non-sqlite |
654 |
|
# table directly. Currently only DBFfiles and memory tables are |
655 |
|
# used as the underlying non-sqlite table, though. |
656 |
|
if self.t_table is not None: |
657 |
|
return self.t_table.ValueRange(col) |
658 |
|
else: |
659 |
|
return self.table.ValueRange(col) |
660 |
|
|
661 |
def UniqueValues(self, col): |
def UniqueValues(self, col): |
662 |
if self.t_table is None: |
# The performance trade-offs for this method are basically the |
663 |
self.copy_to_transient() |
# same as for ValueRange except that currently there doesn't |
664 |
return self.t_table.UniqueValues(col) |
# seem to be a way to take advantage of indexes in this case in |
665 |
|
# sqlite. However, but it's still faster to query the transient |
666 |
|
# table if it already exists. |
667 |
|
if self.t_table is not None: |
668 |
|
return self.t_table.UniqueValues(col) |
669 |
|
else: |
670 |
|
return self.table.UniqueValues(col) |
671 |
|
|
672 |
def SimpleQuery(self, left, comparison, right): |
def SimpleQuery(self, left, comparison, right): |
673 |
if self.t_table is None: |
if self.t_table is None: |
675 |
# Make sure to use the column object of the transient table. The |
# Make sure to use the column object of the transient table. The |
676 |
# left argument is always a column object so we can just ask the |
# left argument is always a column object so we can just ask the |
677 |
# t_table for the right object. |
# t_table for the right object. |
678 |
return self.t_table.SimpleQuery(self.t_table.Column(left.name), |
if hasattr(right, "name"): |
679 |
comparison, right) |
return self.t_table.SimpleQuery(self.t_table.Column(left.name), |
680 |
|
comparison, |
681 |
|
self.t_table.Column(right.name)) |
682 |
|
else: |
683 |
|
return self.t_table.SimpleQuery(self.t_table.Column(left.name), |
684 |
|
comparison, right) |
685 |
|
|
686 |
|
def Dependencies(self): |
687 |
|
"""Return a tuple containing the original table""" |
688 |
|
return (self.table,) |
689 |
|
|
690 |
|
def Width(self, col): |
691 |
|
return self.table.Width(col) |
692 |
|
|
693 |
|
def write_record(self, row, values): |
694 |
|
"""Write the values to the given row. |
695 |
|
|
696 |
|
This is a very experimental feature which doesn't work in all |
697 |
|
cases, so you better know what you're doing when calling this |
698 |
|
method. |
699 |
|
""" |
700 |
|
self.table.write_record(row, values) |