22 |
import weakref |
import weakref |
23 |
from sqlite import connect |
from sqlite import connect |
24 |
|
|
25 |
|
from base import TitledObject |
26 |
|
|
27 |
import table |
import table |
28 |
|
|
29 |
sql_type_map = { |
sql_type_map = { |
87 |
self.internal_name = internal_name |
self.internal_name = internal_name |
88 |
|
|
89 |
|
|
90 |
class TransientTableBase: |
class TransientTableBase(table.OldTableInterfaceMixin): |
91 |
|
|
92 |
"""Base class for tables in the transient database""" |
"""Base class for tables in the transient database""" |
93 |
|
|
106 |
self.orig_names = [] |
self.orig_names = [] |
107 |
self.internal_to_orig = {} |
self.internal_to_orig = {} |
108 |
self.orig_to_internal = {} |
self.orig_to_internal = {} |
109 |
|
self.column_map = {} |
110 |
|
|
111 |
# Create the column objects and fill various maps and lists |
# Create the column objects and fill various maps and lists |
112 |
for col in self.columns: |
for index in range(len(self.columns)): |
113 |
|
col = self.columns[index] |
114 |
self.name_to_column[col.name] = col |
self.name_to_column[col.name] = col |
115 |
self.orig_names.append(col.name) |
self.orig_names.append(col.name) |
116 |
self.internal_to_orig[col.internal_name] = col.name |
self.internal_to_orig[col.internal_name] = col.name |
117 |
self.orig_to_internal[col.name] = col.internal_name |
self.orig_to_internal[col.name] = col.internal_name |
118 |
|
self.column_map[col.name] = col |
119 |
|
self.column_map[index] = col |
120 |
|
|
121 |
# Build the CREATE TABLE statement and create the table in the |
# Build the CREATE TABLE statement and create the table in the |
122 |
# database |
# database |
123 |
table_types = [] |
table_types = ["id INTEGER PRIMARY KEY"] |
124 |
for col in self.columns: |
for col in self.columns: |
125 |
table_types.append("%s %s" % (col.internal_name, |
table_types.append("%s %s" % (col.internal_name, |
126 |
sql_type_map[col.type])) |
sql_type_map[col.type])) |
144 |
self.db.execute(stmt) |
self.db.execute(stmt) |
145 |
self.indexed_columns[column] = 1 |
self.indexed_columns[column] = 1 |
146 |
|
|
147 |
def field_count(self): |
def NumColumns(self): |
148 |
return len(self.columns) |
return len(self.columns) |
149 |
|
|
150 |
def field_info(self, i): |
def NumRows(self): |
|
col = self.columns[i] |
|
|
return col.type, col.name, 0, 0 |
|
|
|
|
|
def field_info_by_name(self, name): |
|
|
for col in self.columns: |
|
|
if col.name == name: |
|
|
return col.type, col.name, 0, 0 |
|
|
else: |
|
|
return None |
|
|
|
|
|
def record_count(self): |
|
151 |
result = self.db.execute("SELECT count(*) FROM %s;" % self.tablename) |
result = self.db.execute("SELECT count(*) FROM %s;" % self.tablename) |
152 |
return int(result[0]) |
return int(result[0]) |
153 |
|
|
154 |
def read_record(self, index): |
def Columns(self): |
155 |
|
return self.columns |
156 |
|
|
157 |
|
def Column(self, col): |
158 |
|
return self.column_map[col] |
159 |
|
|
160 |
|
def HasColumn(self, col): |
161 |
|
"""Return whether the table has a column with the given name or index |
162 |
|
""" |
163 |
|
return self.column_map.has_key(col) |
164 |
|
|
165 |
|
def ReadRowAsDict(self, index): |
166 |
|
# Implementation Strategy: Executing a completely new select |
167 |
|
# statement every time this method is called is too slow. The |
168 |
|
# most important usage is to read the records more or less |
169 |
|
# sequentially. This happens e.g. when drawing a layer with a |
170 |
|
# classification where the shapes are drawn in order of the |
171 |
|
# shape ids. Another pattern is that the same row is requested |
172 |
|
# several times in a row. This happens in the table view, for |
173 |
|
# instance. |
174 |
|
|
175 |
|
# We can exploit this to make access faster by having one cursor |
176 |
|
# open all the time and keeping the last row read around in case |
177 |
|
# the same row is accessed again the next time and if the row |
178 |
|
# index is larger than the row we have read last we simply fetch |
179 |
|
# rows from the cursor until we've reached the requested row. If |
180 |
|
# the requested row index is smaller then we start a new cursor. |
181 |
|
|
182 |
|
# FIXME: So far this scheme seems to work well enough. Obvious |
183 |
|
# improvements would be to start the cursor at exactly the |
184 |
|
# requested row (should be efficient and easy to do now that the |
185 |
|
# id is the primary key) and to perhaps to also start a new |
186 |
|
# cursor if the requested index is much larger than the last row |
187 |
|
# so that we don't read and discard lots of the rows. |
188 |
|
|
189 |
|
# Check whether we have to start a new cursor |
190 |
if self.read_record_cursor is None or index <self.read_record_last_row: |
if self.read_record_cursor is None or index <self.read_record_last_row: |
191 |
stmt = ("SELECT %s FROM %s;" |
stmt = ("SELECT %s FROM %s;" |
192 |
% (", ".join([c.internal_name for c in self.columns]), |
% (", ".join([c.internal_name for c in self.columns]), |
202 |
assert index >= self.read_record_last_row |
assert index >= self.read_record_last_row |
203 |
|
|
204 |
if index == self.read_record_last_row: |
if index == self.read_record_last_row: |
205 |
result = self.read_record_last_result |
result = self.read_record_last_result |
206 |
else: |
else: |
207 |
for i in range(index - self.read_record_last_row): |
for i in range(index - self.read_record_last_row): |
208 |
result = self.read_record_cursor.fetchone() |
result = self.read_record_cursor.fetchone() |
210 |
self.read_record_last_row = index |
self.read_record_last_row = index |
211 |
return dict(zip(self.orig_names, result)) |
return dict(zip(self.orig_names, result)) |
212 |
|
|
213 |
def field_range(self, colname): |
def ReadValue(self, row, col): |
214 |
col = self.name_to_column[colname] |
"""Return the value of the specified row and column |
215 |
|
|
216 |
|
The col parameter may be the index of the column or its name. |
217 |
|
""" |
218 |
|
# Depending on the actual access patterns of the table data, it |
219 |
|
# might be a bit faster in some circumstances to not implement |
220 |
|
# this via ReadRowAsDict, but this simple implementation should |
221 |
|
# be fast enough for most purposes. |
222 |
|
return self.ReadRowAsDict(row)[self.column_map[col].name] |
223 |
|
|
224 |
|
def ValueRange(self, col): |
225 |
|
col = self.column_map[col] |
226 |
iname = col.internal_name |
iname = col.internal_name |
227 |
min, max = self.db.execute("SELECT min(%s), max(%s) FROM %s;" |
min, max = self.db.execute("SELECT min(%s), max(%s) FROM %s;" |
228 |
% (iname, iname, self.tablename)) |
% (iname, iname, self.tablename)) |
229 |
converter = type_converter_map[col.type] |
converter = type_converter_map[col.type] |
230 |
return ((converter(min), None), (converter(max), None)) |
return (converter(min), converter(max)) |
231 |
|
|
232 |
def GetUniqueValues(self, colname): |
def UniqueValues(self, col): |
233 |
iname = self.orig_to_internal[colname] |
iname = self.column_map[col].internal_name |
234 |
cursor = self.db.cursor() |
cursor = self.db.cursor() |
235 |
cursor.execute("SELECT %s FROM %s GROUP BY %s;" |
cursor.execute("SELECT %s FROM %s GROUP BY %s;" |
236 |
% (iname, self.tablename, iname)) |
% (iname, self.tablename, iname)) |
242 |
result.append(row[0]) |
result.append(row[0]) |
243 |
return result |
return result |
244 |
|
|
245 |
|
def Width(self, col): |
246 |
|
"""Return the maximum width of values in the column |
247 |
|
|
248 |
|
The return value is the the maximum length of string representation |
249 |
|
of the values in the column (represented by index or name).""" |
250 |
|
max = 0 |
251 |
|
|
252 |
|
type = self.column_map[col].type |
253 |
|
iname = self.column_map[col].internal_name |
254 |
|
cursor = self.db.cursor() |
255 |
|
cursor.execute("SELECT %s FROM %s;" % (iname, self.tablename)) |
256 |
|
values = [ i[0] for i in cursor.fetchall()] |
257 |
|
if not values: |
258 |
|
return None |
259 |
|
|
260 |
|
if type == sql_type_map[table.FIELDTYPE_DOUBLE]: |
261 |
|
format = "%.12f" |
262 |
|
elif type == sql_type_map[table.FIELDTYPE_INT]: |
263 |
|
format = "%d" |
264 |
|
else: |
265 |
|
format = "%s" |
266 |
|
for value in values: |
267 |
|
if value is None: continue |
268 |
|
l = len(format % value) |
269 |
|
if l > max: |
270 |
|
max = l |
271 |
|
|
272 |
|
return max |
273 |
|
|
274 |
|
def SimpleQuery(self, left, comparison, right): |
275 |
|
"""Return the indices of all rows that matching a condition. |
276 |
|
|
277 |
|
Parameters: |
278 |
|
left -- The column object for the left side of the comparison |
279 |
|
|
280 |
|
comparison -- The comparison operator as a string. It must be |
281 |
|
one of '==', '!=', '<', '<=', '>=', '>' |
282 |
|
|
283 |
|
right -- The right hand side of the comparison. It must be |
284 |
|
either a column object or a value, i.e. a string, |
285 |
|
int or float. |
286 |
|
|
287 |
|
The return value is a sorted list of the indices of the rows |
288 |
|
where the condition is true. |
289 |
|
""" |
290 |
|
if comparison not in ("==", "!=", "<", "<=", ">=", ">"): |
291 |
|
raise ValueError("Comparison operator %r not allowed" % comparison) |
292 |
|
|
293 |
|
if hasattr(right, "internal_name"): |
294 |
|
right_template = right.internal_name |
295 |
|
params = () |
296 |
|
else: |
297 |
|
right_template = "%s" |
298 |
|
params = (right,) |
299 |
|
|
300 |
class TransientTable(TransientTableBase): |
query = "SELECT id FROM %s WHERE %s %s %s ORDER BY id;" \ |
301 |
|
% (self.tablename, left.internal_name, comparison, |
302 |
|
right_template) |
303 |
|
|
304 |
|
cursor = self.db.cursor() |
305 |
|
cursor.execute(query, params) |
306 |
|
result = [] |
307 |
|
while 1: |
308 |
|
row = cursor.fetchone() |
309 |
|
if row is None: |
310 |
|
break |
311 |
|
result.append(row[0]) |
312 |
|
return result |
313 |
|
|
314 |
|
def Dependencies(self): |
315 |
|
"""Placeholder for a method in a derived class. |
316 |
|
|
317 |
|
Return a sequence with the tables and other data objects that |
318 |
|
self depends on. |
319 |
|
""" |
320 |
|
raise NotImplementedError |
321 |
|
|
322 |
|
|
323 |
|
class TransientTable(TitledObject, TransientTableBase): |
324 |
|
|
325 |
"""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.""" |
326 |
|
|
331 |
interface. |
interface. |
332 |
""" |
""" |
333 |
TransientTableBase.__init__(self, transient_db) |
TransientTableBase.__init__(self, transient_db) |
334 |
|
TitledObject.__init__(self, table.Title()) |
335 |
self.create(table) |
self.create(table) |
336 |
|
|
337 |
def create(self, table): |
def create(self, table): |
338 |
columns = [] |
columns = [] |
339 |
for i in range(table.field_count()): |
for col in table.Columns(): |
340 |
type, name = table.field_info(i)[:2] |
columns.append(ColumnReference(col.name, col.type, |
|
columns.append(ColumnReference(name, type, |
|
341 |
self.db.new_column_name())) |
self.db.new_column_name())) |
342 |
TransientTableBase.create(self, columns) |
TransientTableBase.create(self, columns) |
343 |
|
|
344 |
# copy the input table to the transient db |
# copy the input table to the transient db |
345 |
insert_template = "INSERT INTO %s (%s) VALUES (%s);" \ |
|
346 |
|
# A key to insert to use for the formatting of the insert |
347 |
|
# statement. The key must not be equal to any of the column |
348 |
|
# names so we construct one by building a string of x's that is |
349 |
|
# longer than any of the column names |
350 |
|
id_key = max([len(col.name) for col in self.columns]) * "x" |
351 |
|
|
352 |
|
insert_template = "INSERT INTO %s (id, %s) VALUES (%%(%s)s, %s);" \ |
353 |
% (self.tablename, |
% (self.tablename, |
354 |
", ".join([col.internal_name |
", ".join([col.internal_name |
355 |
for col in self.columns]), |
for col in self.columns]), |
356 |
|
id_key, |
357 |
", ".join(["%%(%s)s" % col.name |
", ".join(["%%(%s)s" % col.name |
358 |
for col in self.columns])) |
for col in self.columns])) |
359 |
cursor = self.db.cursor() |
cursor = self.db.cursor() |
360 |
for i in range(table.record_count()): |
for i in range(table.NumRows()): |
361 |
cursor.execute(insert_template, table.read_record(i)) |
row = table.ReadRowAsDict(i) |
362 |
|
row[id_key] = i |
363 |
|
cursor.execute(insert_template, row) |
364 |
self.db.conn.commit() |
self.db.conn.commit() |
365 |
|
|
366 |
|
|
367 |
|
|
368 |
class TransientJoinedTable(TransientTableBase): |
class TransientJoinedTable(TitledObject, TransientTableBase): |
369 |
|
|
370 |
"""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""" |
371 |
|
|
372 |
def __init__(self, transient_db, left_table, left_field, |
def __init__(self, transient_db, left_table, left_field, |
373 |
right_table, right_field = None): |
right_table, right_field = None, outer_join = False): |
374 |
"""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. |
375 |
|
|
376 |
Both input tables, left_table and right_table must have a |
Both input tables, left_table and right_table must have a |
377 |
transient_table method that returns a table object for a table |
transient_table method that returns a table object for a table |
378 |
in the trnsient database. The join is performed on the condition |
in the transient database. The join is performed on the condition |
379 |
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 |
380 |
equal to the value of the right_field in the right_table. |
equal to the value of the right_field in the right_table. |
381 |
|
|
382 |
The joined table contains all columns of the input tables with |
The joined table contains all columns of the input tables, |
383 |
one exception: Any column in the right_table with the same name |
however, the column names of the right table may be changed |
384 |
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 |
385 |
somewhat of an implementation detail, but is done so that the |
currently done by appending a sufficient number of underscores |
386 |
column names of the joined table can be the same as the column |
('_'). |
|
names of the input tables without having to create prefixes. |
|
387 |
""" |
""" |
388 |
TransientTableBase.__init__(self, transient_db) |
TransientTableBase.__init__(self, transient_db) |
389 |
|
self.dependencies = (left_table, right_table) |
390 |
self.left_table = left_table.transient_table() |
self.left_table = left_table.transient_table() |
391 |
self.left_field = left_field |
self.left_field = left_field |
392 |
self.right_table = right_table.transient_table() |
self.right_table = right_table.transient_table() |
394 |
self.right_field = right_field |
self.right_field = right_field |
395 |
else: |
else: |
396 |
self.right_field = self.left_field |
self.right_field = self.left_field |
397 |
|
self.outer_join = outer_join |
398 |
|
|
399 |
|
title = "Join of %(left)s and %(right)s" \ |
400 |
|
% {"left": self.left_table.Title(), |
401 |
|
"right": self.right_table.Title()} |
402 |
|
TitledObject.__init__(self, title) |
403 |
|
|
404 |
self.create() |
self.create() |
405 |
|
|
406 |
def create(self): |
def create(self): |
409 |
|
|
410 |
self.right_table.ensure_index(self.right_field) |
self.right_table.ensure_index(self.right_field) |
411 |
|
|
412 |
|
# determine the internal column names to join on before |
413 |
|
# coalescing the column information because if the external |
414 |
|
# column names are the same they will be mapped to the same |
415 |
|
# internal name afterwards. |
416 |
|
internal_left_col = self.left_table.orig_to_internal[self.left_field] |
417 |
|
internal_right_col =self.right_table.orig_to_internal[self.right_field] |
418 |
|
|
419 |
# Coalesce the column information |
# Coalesce the column information |
420 |
visited = {} |
visited = {} |
421 |
columns = [] |
columns = [] |
422 |
for col in self.left_table.columns + self.right_table.columns: |
newcolumns = [] |
423 |
if col.name in visited: |
for table in (self.left_table, self.right_table): |
424 |
continue |
for col in table.Columns(): |
425 |
columns.append(col) |
colname = col.name |
426 |
TransientTableBase.create(self, columns) |
# We can't allow multiple columns with the same |
427 |
|
# original name, so append '_' to this one until |
428 |
|
# it is unique. |
429 |
|
# FIXME: There should be a better solution. |
430 |
|
while colname in visited: |
431 |
|
colname = colname + '_' |
432 |
|
columns.append((table.tablename, col)) |
433 |
|
newcol = ColumnReference(colname, col.type, |
434 |
|
"Col%03d" % (len(newcolumns)+1)) |
435 |
|
newcolumns.append(newcol) |
436 |
|
visited[colname] = 1 |
437 |
|
TransientTableBase.create(self, newcolumns) |
438 |
|
|
439 |
# Copy the joined data to the table. |
# Copy the joined data to the table. |
440 |
internal_names = [col.internal_name for col in self.columns] |
newinternal_names = [col.internal_name for col in self.columns] |
441 |
stmt = "INSERT INTO %s (%s) SELECT %s FROM %s JOIN %s ON %s = %s;" \ |
internal_references = ["%s.%s" % (table, col.internal_name) |
442 |
% (self.tablename, |
for table, col in columns] |
443 |
", ".join(internal_names), |
if self.outer_join: |
444 |
", ".join(internal_names), |
join_operator = 'LEFT OUTER JOIN' |
445 |
self.left_table.tablename, |
else: |
446 |
self.right_table.tablename, |
join_operator = 'JOIN' |
447 |
self.orig_to_internal[self.left_field], |
stmt = ("INSERT INTO %s (id, %s) SELECT %s.id, %s FROM %s" |
448 |
self.orig_to_internal[self.right_field]) |
" %s %s ON %s.%s = %s.%s;" |
449 |
|
% (self.tablename, |
450 |
|
", ".join(newinternal_names), |
451 |
|
self.left_table.tablename, |
452 |
|
", ".join(internal_references), |
453 |
|
self.left_table.tablename, |
454 |
|
join_operator, |
455 |
|
self.right_table.tablename, |
456 |
|
self.left_table.tablename, |
457 |
|
internal_left_col, |
458 |
|
self.right_table.tablename, |
459 |
|
internal_right_col)) |
460 |
self.db.execute(stmt) |
self.db.execute(stmt) |
461 |
|
|
462 |
|
def Dependencies(self): |
463 |
|
"""Return a tuple with the two tables the join depends on.""" |
464 |
|
return self.dependencies |
465 |
|
|
466 |
class AutoTransientTable: |
|
467 |
|
class AutoTransientTable(TitledObject, table.OldTableInterfaceMixin): |
468 |
|
|
469 |
"""Table that copies data to a transient table on demand. |
"""Table that copies data to a transient table on demand. |
470 |
|
|
473 |
""" |
""" |
474 |
|
|
475 |
def __init__(self, transient_db, table): |
def __init__(self, transient_db, table): |
476 |
|
TitledObject.__init__(self, table.Title()) |
477 |
self.transient_db = transient_db |
self.transient_db = transient_db |
478 |
self.table = table |
self.table = table |
479 |
self.t_table = None |
self.t_table = None |
480 |
|
|
481 |
def record_count(self): |
def Columns(self): |
482 |
"""Return the number of records""" |
return self.table.Columns() |
483 |
return self.table.record_count() |
|
484 |
|
def Column(self, col): |
485 |
def field_count(self): |
return self.table.Column(col) |
|
"""Return the number of fields in a record""" |
|
|
return self.table.field_count() |
|
|
|
|
|
def field_info(self, field): |
|
|
"""Return a tuple (type, name, width, prec) for the field no. field |
|
|
|
|
|
type is the data type of the field, name the name, width the |
|
|
field width in characters and prec the decimal precision. |
|
|
""" |
|
|
info = self.table.field_info(field) |
|
|
if info: |
|
|
info = info[:2] + (0, 0) |
|
|
return info |
|
|
|
|
|
def field_info_by_name(self, fieldName): |
|
|
info = self.table.field_info_by_name(fieldName) |
|
|
if info: |
|
|
info = info[:2] + (0, 0) |
|
|
return info |
|
486 |
|
|
487 |
def read_record(self, record): |
def HasColumn(self, col): |
488 |
|
"""Return whether the table has a column with the given name or index |
489 |
|
""" |
490 |
|
return self.table.HasColumn(col) |
491 |
|
|
492 |
|
def NumRows(self): |
493 |
|
return self.table.NumRows() |
494 |
|
|
495 |
|
def NumColumns(self): |
496 |
|
return self.table.NumColumns() |
497 |
|
|
498 |
|
def ReadRowAsDict(self, record): |
499 |
"""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 |
500 |
""" |
""" |
501 |
if self.t_table is not None: |
if self.t_table is not None: |
502 |
return self.t_table.read_record(record) |
return self.t_table.ReadRowAsDict(record) |
503 |
else: |
else: |
504 |
return self.table.read_record(record) |
return self.table.ReadRowAsDict(record) |
505 |
|
|
506 |
def write_record(self, record, values): |
def ReadValue(self, row, col): |
507 |
raise NotImplementedError |
"""Return the value of the specified row and column |
508 |
|
|
509 |
|
The col parameter may be the index of the column or its name. |
510 |
|
""" |
511 |
|
if self.t_table is not None: |
512 |
|
return self.t_table.ReadValue(row, col) |
513 |
|
else: |
514 |
|
return self.table.ReadValue(row, col) |
515 |
|
|
516 |
def copy_to_transient(self): |
def copy_to_transient(self): |
517 |
"""Internal: Create a transient table and copy the data into it""" |
"""Internal: Create a transient table and copy the data into it""" |
525 |
self.copy_to_transient() |
self.copy_to_transient() |
526 |
return self.t_table |
return self.t_table |
527 |
|
|
528 |
def field_range(self, colname): |
def ValueRange(self, col): |
529 |
|
if self.t_table is None: |
530 |
|
self.copy_to_transient() |
531 |
|
return self.t_table.ValueRange(col) |
532 |
|
|
533 |
|
def UniqueValues(self, col): |
534 |
if self.t_table is None: |
if self.t_table is None: |
535 |
self.copy_to_transient() |
self.copy_to_transient() |
536 |
return self.t_table.field_range(colname) |
return self.t_table.UniqueValues(col) |
537 |
|
|
538 |
def GetUniqueValues(self, colname): |
def SimpleQuery(self, left, comparison, right): |
539 |
if self.t_table is None: |
if self.t_table is None: |
540 |
self.copy_to_transient() |
self.copy_to_transient() |
541 |
return self.t_table.GetUniqueValues(colname) |
# Make sure to use the column object of the transient table. The |
542 |
|
# left argument is always a column object so we can just ask the |
543 |
|
# t_table for the right object. |
544 |
|
if hasattr(right, "name"): |
545 |
|
return self.t_table.SimpleQuery(self.t_table.Column(left.name), |
546 |
|
comparison, |
547 |
|
self.t_table.Column(right.name)) |
548 |
|
else: |
549 |
|
return self.t_table.SimpleQuery(self.t_table.Column(left.name), |
550 |
|
comparison, right) |
551 |
|
|
552 |
|
def Dependencies(self): |
553 |
|
"""Return a tuple containing the original table""" |
554 |
|
return (self.table,) |
555 |
|
|
556 |
|
def Width(self, col): |
557 |
|
return self.table.Width(col) |