18 |
# $Source$ |
# $Source$ |
19 |
# $Id$ |
# $Id$ |
20 |
|
|
|
import os |
|
|
import weakref |
|
21 |
from sqlite import connect |
from sqlite import connect |
22 |
|
|
23 |
|
from base import TitledObject |
24 |
|
|
25 |
import table |
import table |
26 |
|
|
27 |
sql_type_map = { |
sql_type_map = { |
160 |
""" |
""" |
161 |
return self.column_map.has_key(col) |
return self.column_map.has_key(col) |
162 |
|
|
163 |
def ReadRowAsDict(self, index): |
def RowIdToOrdinal(self, gid): |
164 |
|
"""Return the row ordinal given its id |
165 |
|
|
166 |
|
At the moment the transient tables are only used for tables that |
167 |
|
don't distinguish between row number and row id, so the value is |
168 |
|
returned unchanged. |
169 |
|
""" |
170 |
|
return gid |
171 |
|
|
172 |
|
def RowOrdinalToId(self, num): |
173 |
|
"""Return the rowid for given its ordinal |
174 |
|
|
175 |
|
At the moment the transient tables are only used for tables that |
176 |
|
don't distinguish between row number and row id, so the value is |
177 |
|
returned unchanged. |
178 |
|
""" |
179 |
|
return num |
180 |
|
|
181 |
|
def ReadRowAsDict(self, index, row_is_ordinal = 0): |
182 |
|
"""Return the entire row as a dictionary with column names as keys |
183 |
|
|
184 |
|
The row_is_ordinal is ignored because at the moment the |
185 |
|
transient tables are only used for DBF files where it doesn't |
186 |
|
matter. |
187 |
|
""" |
188 |
# Implementation Strategy: Executing a completely new select |
# Implementation Strategy: Executing a completely new select |
189 |
# statement every time this method is called is too slow. The |
# statement every time this method is called is too slow. The |
190 |
# most important usage is to read the records more or less |
# most important usage is to read the records more or less |
232 |
self.read_record_last_row = index |
self.read_record_last_row = index |
233 |
return dict(zip(self.orig_names, result)) |
return dict(zip(self.orig_names, result)) |
234 |
|
|
235 |
def ReadValue(self, row, col): |
def ReadValue(self, row, col, row_is_ordinal = 0): |
236 |
"""Return the value of the specified row and column |
"""Return the value of the specified row and column |
237 |
|
|
238 |
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. |
239 |
|
|
240 |
|
The row_is_ordinal is ignored because at the moment the |
241 |
|
transient tables are only used for DBF files where it doesn't |
242 |
|
matter. |
243 |
""" |
""" |
244 |
# Depending on the actual access patterns of the table data, it |
# Depending on the actual access patterns of the table data, it |
245 |
# might be a bit faster in some circumstances to not implement |
# might be a bit faster in some circumstances to not implement |
248 |
return self.ReadRowAsDict(row)[self.column_map[col].name] |
return self.ReadRowAsDict(row)[self.column_map[col].name] |
249 |
|
|
250 |
def ValueRange(self, col): |
def ValueRange(self, col): |
251 |
|
# Performance notes: |
252 |
|
# |
253 |
|
# In sqlite 2.8.6 the min and max aggregate functions can use an |
254 |
|
# index but only when used as the only expression in the select |
255 |
|
# statement (i.e. 'select min(col), max(col) from tbl;' will not |
256 |
|
# use the index but 'select min(col) from tbl;' will) so we |
257 |
|
# query the minimum and maximum separately. |
258 |
|
# |
259 |
|
# With the separate statements we can take advantage of an index |
260 |
|
# if it exists. If the index doesn't exist, creating it first |
261 |
|
# and then using it in the query is slower than the queries |
262 |
|
# without an index. Creating the index is only an advantage if |
263 |
|
# the queries are performed multiple times. With the current use |
264 |
|
# patterns where ValueRange is only used occasionally by the |
265 |
|
# classification generation dialog creating the index only for |
266 |
|
# this usage is not really worth it, so we don't. |
267 |
col = self.column_map[col] |
col = self.column_map[col] |
268 |
iname = col.internal_name |
iname = col.internal_name |
269 |
min, max = self.db.execute("SELECT min(%s), max(%s) FROM %s;" |
min = self.db.execute("SELECT min(%s) FROM %s;" |
270 |
% (iname, iname, self.tablename)) |
% (iname, self.tablename))[0] |
271 |
|
max = self.db.execute("SELECT max(%s) FROM %s;" |
272 |
|
% (iname, self.tablename))[0] |
273 |
converter = type_converter_map[col.type] |
converter = type_converter_map[col.type] |
274 |
return (converter(min), converter(max)) |
return (converter(min), converter(max)) |
275 |
|
|
276 |
def UniqueValues(self, col): |
def UniqueValues(self, col): |
277 |
|
# Performance notes: |
278 |
|
# |
279 |
|
# In sqlite 2.8.6 there doesn't seem to be a way to query the |
280 |
|
# unique items that uses an index. I've tried |
281 |
|
# |
282 |
|
# SELECT col FROM tbl GROUP BY col; |
283 |
|
# |
284 |
|
# and |
285 |
|
# |
286 |
|
# SELECT DISTINCT col FROM tbl; |
287 |
|
# |
288 |
|
# and in both cases the index is not used. If the index isn't |
289 |
|
# used it doesn't make sense to call self.ensure_index. |
290 |
iname = self.column_map[col].internal_name |
iname = self.column_map[col].internal_name |
291 |
cursor = self.db.cursor() |
cursor = self.db.cursor() |
292 |
cursor.execute("SELECT %s FROM %s GROUP BY %s;" |
cursor.execute("SELECT %s FROM %s GROUP BY %s;" |
299 |
result.append(row[0]) |
result.append(row[0]) |
300 |
return result |
return result |
301 |
|
|
302 |
|
def Width(self, col): |
303 |
|
"""Return the maximum width of values in the column |
304 |
|
|
305 |
|
The return value is the the maximum length of string |
306 |
|
representation of the values in the column (represented by index |
307 |
|
or name). |
308 |
|
""" |
309 |
|
max = 0 |
310 |
|
|
311 |
|
type = self.column_map[col].type |
312 |
|
iname = self.column_map[col].internal_name |
313 |
|
cursor = self.db.cursor() |
314 |
|
cursor.execute("SELECT %s FROM %s;" % (iname, self.tablename)) |
315 |
|
values = [ i[0] for i in cursor.fetchall()] |
316 |
|
if not values: |
317 |
|
return None |
318 |
|
|
319 |
|
if type == table.FIELDTYPE_DOUBLE: |
320 |
|
format = "%.12f" |
321 |
|
elif type == table.FIELDTYPE_INT: |
322 |
|
format = "%d" |
323 |
|
else: |
324 |
|
format = "%s" |
325 |
|
for value in values: |
326 |
|
if value is None: continue |
327 |
|
l = len(format % value) |
328 |
|
if l > max: |
329 |
|
max = l |
330 |
|
|
331 |
|
return max |
332 |
|
|
333 |
def SimpleQuery(self, left, comparison, right): |
def SimpleQuery(self, left, comparison, right): |
334 |
"""Return the indices of all rows that matching a condition. |
"""Return the indices of all rows that matching a condition. |
335 |
|
|
379 |
raise NotImplementedError |
raise NotImplementedError |
380 |
|
|
381 |
|
|
382 |
class TransientTable(TransientTableBase): |
class TransientTable(TitledObject, TransientTableBase): |
383 |
|
|
384 |
"""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.""" |
385 |
|
|
390 |
interface. |
interface. |
391 |
""" |
""" |
392 |
TransientTableBase.__init__(self, transient_db) |
TransientTableBase.__init__(self, transient_db) |
393 |
|
TitledObject.__init__(self, table.Title()) |
394 |
self.create(table) |
self.create(table) |
395 |
|
|
396 |
def create(self, table): |
def create(self, table): |
424 |
|
|
425 |
|
|
426 |
|
|
427 |
class TransientJoinedTable(TransientTableBase): |
class TransientJoinedTable(TitledObject, TransientTableBase): |
428 |
|
|
429 |
"""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""" |
430 |
|
|
431 |
def __init__(self, transient_db, left_table, left_field, |
def __init__(self, transient_db, left_table, left_field, |
432 |
right_table, right_field = None): |
right_table, right_field = None, outer_join = False): |
433 |
"""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. |
434 |
|
|
435 |
Both input tables, left_table and right_table must have a |
Both input tables, left_table and right_table must have a |
436 |
transient_table method that returns a table object for a table |
transient_table method that returns a table object for a table |
437 |
in the trnsient database. The join is performed on the condition |
in the transient database. The join is performed on the condition |
438 |
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 |
439 |
equal to the value of the right_field in the right_table. |
equal to the value of the right_field in the right_table. |
440 |
|
|
441 |
The joined table contains all columns of the input tables with |
The joined table contains all columns of the input tables, |
442 |
one exception: Any column in the right_table with the same name |
however, the column names of the right table may be changed |
443 |
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 |
444 |
somewhat of an implementation detail, but is done so that the |
currently done by appending a sufficient number of underscores |
445 |
column names of the joined table can be the same as the column |
('_'). |
|
names of the input tables without having to create prefixes. |
|
446 |
""" |
""" |
447 |
TransientTableBase.__init__(self, transient_db) |
TransientTableBase.__init__(self, transient_db) |
448 |
self.dependencies = (left_table, right_table) |
self.dependencies = (left_table, right_table) |
453 |
self.right_field = right_field |
self.right_field = right_field |
454 |
else: |
else: |
455 |
self.right_field = self.left_field |
self.right_field = self.left_field |
456 |
|
self.outer_join = outer_join |
457 |
|
|
458 |
|
title = "Join of %(left)s and %(right)s" \ |
459 |
|
% {"left": self.left_table.Title(), |
460 |
|
"right": self.right_table.Title()} |
461 |
|
TitledObject.__init__(self, title) |
462 |
|
|
463 |
self.create() |
self.create() |
464 |
|
|
465 |
def create(self): |
def create(self): |
468 |
|
|
469 |
self.right_table.ensure_index(self.right_field) |
self.right_table.ensure_index(self.right_field) |
470 |
|
|
471 |
|
# determine the internal column names to join on before |
472 |
|
# coalescing the column information because if the external |
473 |
|
# column names are the same they will be mapped to the same |
474 |
|
# internal name afterwards. |
475 |
|
internal_left_col = self.left_table.orig_to_internal[self.left_field] |
476 |
|
internal_right_col =self.right_table.orig_to_internal[self.right_field] |
477 |
|
|
478 |
# Coalesce the column information |
# Coalesce the column information |
479 |
visited = {} |
visited = {} |
480 |
columns = [] |
columns = [] |
481 |
for col in self.left_table.columns + self.right_table.columns: |
newcolumns = [] |
482 |
if col.name in visited: |
for table in (self.left_table, self.right_table): |
483 |
# We can't allow multiple columns with the same original |
for col in table.Columns(): |
484 |
# name, so omit this one. FIXME: There should be a |
colname = col.name |
485 |
# better solution. |
# We can't allow multiple columns with the same |
486 |
continue |
# original name, so append '_' to this one until |
487 |
columns.append(col) |
# it is unique. |
488 |
TransientTableBase.create(self, columns) |
# FIXME: There should be a better solution. |
489 |
|
while colname in visited: |
490 |
|
colname = colname + '_' |
491 |
|
columns.append((table.tablename, col)) |
492 |
|
newcol = ColumnReference(colname, col.type, |
493 |
|
"Col%03d" % (len(newcolumns)+1)) |
494 |
|
newcolumns.append(newcol) |
495 |
|
visited[colname] = 1 |
496 |
|
TransientTableBase.create(self, newcolumns) |
497 |
|
|
498 |
# Copy the joined data to the table. |
# Copy the joined data to the table. |
499 |
internal_names = [col.internal_name for col in self.columns] |
newinternal_names = [col.internal_name for col in self.columns] |
500 |
|
internal_references = ["%s.%s" % (table, col.internal_name) |
501 |
|
for table, col in columns] |
502 |
|
if self.outer_join: |
503 |
|
join_operator = 'LEFT OUTER JOIN' |
504 |
|
else: |
505 |
|
join_operator = 'JOIN' |
506 |
stmt = ("INSERT INTO %s (id, %s) SELECT %s.id, %s FROM %s" |
stmt = ("INSERT INTO %s (id, %s) SELECT %s.id, %s FROM %s" |
507 |
" JOIN %s ON %s = %s;" |
" %s %s ON %s.%s = %s.%s;" |
508 |
% (self.tablename, |
% (self.tablename, |
509 |
", ".join(internal_names), |
", ".join(newinternal_names), |
510 |
|
self.left_table.tablename, |
511 |
|
", ".join(internal_references), |
512 |
self.left_table.tablename, |
self.left_table.tablename, |
513 |
", ".join(internal_names), |
join_operator, |
514 |
|
self.right_table.tablename, |
515 |
self.left_table.tablename, |
self.left_table.tablename, |
516 |
|
internal_left_col, |
517 |
self.right_table.tablename, |
self.right_table.tablename, |
518 |
self.orig_to_internal[self.left_field], |
internal_right_col)) |
|
self.orig_to_internal[self.right_field])) |
|
519 |
self.db.execute(stmt) |
self.db.execute(stmt) |
520 |
|
|
521 |
def Dependencies(self): |
def Dependencies(self): |
522 |
"""Return a tuple with the two tables the join depends on.""" |
"""Return a tuple with the two tables the join depends on.""" |
523 |
return self.dependencies |
return self.dependencies |
524 |
|
|
525 |
|
def JoinType(self): |
526 |
|
"""Return the type of the join (either 'INNER' or 'LEFT OUTER')""" |
527 |
|
if self.outer_join: |
528 |
|
return "LEFT OUTER" |
529 |
|
else: |
530 |
|
return "INNER" |
531 |
|
|
532 |
|
|
533 |
class AutoTransientTable(table.OldTableInterfaceMixin): |
class AutoTransientTable(TitledObject, table.OldTableInterfaceMixin): |
534 |
|
|
535 |
"""Table that copies data to a transient table on demand. |
"""Table that copies data to a transient table on demand. |
536 |
|
|
539 |
""" |
""" |
540 |
|
|
541 |
def __init__(self, transient_db, table): |
def __init__(self, transient_db, table): |
542 |
|
TitledObject.__init__(self, table.Title()) |
543 |
self.transient_db = transient_db |
self.transient_db = transient_db |
544 |
self.table = table |
self.table = table |
545 |
self.t_table = None |
self.t_table = None |
561 |
def NumColumns(self): |
def NumColumns(self): |
562 |
return self.table.NumColumns() |
return self.table.NumColumns() |
563 |
|
|
564 |
def ReadRowAsDict(self, record): |
def RowIdToOrdinal(self, gid): |
565 |
|
"""Return the row ordinal given its id""" |
566 |
|
if self.t_table is not None: |
567 |
|
return self.t_table.RowIdToOrdinal(gid) |
568 |
|
else: |
569 |
|
return self.table.RowIdToOrdinal(gid) |
570 |
|
|
571 |
|
def RowOrdinalToId(self, num): |
572 |
|
"""Return the rowid for given its ordinal""" |
573 |
|
if self.t_table is not None: |
574 |
|
return self.t_table.RowOrdinalToId(num) |
575 |
|
else: |
576 |
|
return self.table.RowOrdinalToId(num) |
577 |
|
|
578 |
|
def ReadRowAsDict(self, record, row_is_ordinal = 0): |
579 |
"""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 |
580 |
""" |
""" |
581 |
if self.t_table is not None: |
if self.t_table is not None: |
582 |
return self.t_table.ReadRowAsDict(record) |
return self.t_table.ReadRowAsDict(record, |
583 |
|
row_is_ordinal = row_is_ordinal) |
584 |
else: |
else: |
585 |
return self.table.ReadRowAsDict(record) |
return self.table.ReadRowAsDict(record, |
586 |
|
row_is_ordinal = row_is_ordinal) |
587 |
|
|
588 |
def ReadValue(self, row, col): |
def ReadValue(self, row, col, row_is_ordinal = 0): |
589 |
"""Return the value of the specified row and column |
"""Return the value of the specified row and column |
590 |
|
|
591 |
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. |
592 |
""" |
""" |
593 |
if self.t_table is not None: |
if self.t_table is not None: |
594 |
return self.t_table.ReadValue(row, col) |
return self.t_table.ReadValue(row, col, |
595 |
|
row_is_ordinal = row_is_ordinal) |
596 |
else: |
else: |
597 |
return self.table.ReadValue(row, col) |
return self.table.ReadValue(row, col, |
598 |
|
row_is_ordinal = row_is_ordinal) |
599 |
|
|
600 |
def copy_to_transient(self): |
def copy_to_transient(self): |
601 |
"""Internal: Create a transient table and copy the data into it""" |
"""Internal: Create a transient table and copy the data into it""" |
610 |
return self.t_table |
return self.t_table |
611 |
|
|
612 |
def ValueRange(self, col): |
def ValueRange(self, col): |
613 |
if self.t_table is None: |
# Performance of sqlite vs. DBF for this method: |
614 |
self.copy_to_transient() |
# |
615 |
return self.t_table.ValueRange(col) |
# If the table has been copied to the sqlite database it's |
616 |
|
# faster to use it even if there is no index on that column. |
617 |
|
# Otherwise it's faster to simply loop through all rows in the |
618 |
|
# DBF file. Copying the data to the sqlite database can take |
619 |
|
# very long for large amounts of data |
620 |
|
# |
621 |
|
# Of course if the table is not a DBF file the issue could be |
622 |
|
# different, although copying the data into sqlite first will |
623 |
|
# likely always be slower than simply querying the non-sqlite |
624 |
|
# table directly. Currently only DBFfiles and memory tables are |
625 |
|
# used as the underlying non-sqlite table, though. |
626 |
|
if self.t_table is not None: |
627 |
|
return self.t_table.ValueRange(col) |
628 |
|
else: |
629 |
|
return self.table.ValueRange(col) |
630 |
|
|
631 |
def UniqueValues(self, col): |
def UniqueValues(self, col): |
632 |
if self.t_table is None: |
# The performance trade-offs for this method are basically the |
633 |
self.copy_to_transient() |
# same as for ValueRange except that currently there doesn't |
634 |
return self.t_table.UniqueValues(col) |
# seem to be a way to take advantage of indexes in this case in |
635 |
|
# sqlite. However, but it's still faster to query the transient |
636 |
|
# table if it already exists. |
637 |
|
if self.t_table is not None: |
638 |
|
return self.t_table.UniqueValues(col) |
639 |
|
else: |
640 |
|
return self.table.UniqueValues(col) |
641 |
|
|
642 |
def SimpleQuery(self, left, comparison, right): |
def SimpleQuery(self, left, comparison, right): |
643 |
if self.t_table is None: |
if self.t_table is None: |
656 |
def Dependencies(self): |
def Dependencies(self): |
657 |
"""Return a tuple containing the original table""" |
"""Return a tuple containing the original table""" |
658 |
return (self.table,) |
return (self.table,) |
659 |
|
|
660 |
|
def Width(self, col): |
661 |
|
return self.table.Width(col) |