/[thuban]/branches/WIP-pyshapelib-bramz/Thuban/Model/transientdb.py

Diff of /branches/WIP-pyshapelib-bramz/Thuban/Model/transientdb.py

Parent Directory | Revision Log | View Patch Patch

-revision 777 by bh,
Tue Apr 29 14:54:12 2003 UTC
+revision 1923 by bh,
Fri Nov  7 12:07:01 2003 UTC
 Line 18 
 __version__ = "$Revision$"
  # $Source$
  # $Id$
- import os
- import weakref
  from sqlite import connect
+ from base import TitledObject
  import table
  sql_type_map = {
 Line 85 
 class ColumnReference:
          self.internal_name = internal_name
- class TransientTableBase:
+ class TransientTableBase(table.OldTableInterfaceMixin):
      """Base class for tables in the transient database"""
 Line 96 
 class TransientTableBase:
          self.indexed_columns = {}
          self.read_record_cursor = None
          self.read_record_last_row = None
+         self.read_record_last_result = None
      def create(self, columns):
          self.columns = columns
-Line 103 
 class TransientTableBase:
+Line 104 
 class TransientTableBase:
          self.orig_names = []
          self.internal_to_orig = {}
          self.orig_to_internal = {}
+         self.column_map = {}
          # Create the column objects and fill various maps and lists
-         for col in self.columns:
+         for index in range(len(self.columns)):
+             col = self.columns[index]
              self.name_to_column[col.name] = col
              self.orig_names.append(col.name)
              self.internal_to_orig[col.internal_name] = col.name
              self.orig_to_internal[col.name] = col.internal_name
+             self.column_map[col.name] = col
+             self.column_map[index] = col
          # Build the CREATE TABLE statement and create the table in the
          # database
-         table_types = []
+         table_types = ["id INTEGER PRIMARY KEY"]
          for col in self.columns:
              table_types.append("%s %s" % (col.internal_name,
                                            sql_type_map[col.type]))
-Line 137 
 class TransientTableBase:
+Line 142 
 class TransientTableBase:
              self.db.execute(stmt)
              self.indexed_columns[column] = 1
-     def field_count(self):
+     def NumColumns(self):
          return len(self.columns)
-     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):
          result = self.db.execute("SELECT count(*) FROM %s;" % self.tablename)
          return int(result[0])
-     def read_record(self, index):
+     def Columns(self):
+         return self.columns
+     def Column(self, col):
+         return self.column_map[col]
+     def HasColumn(self, col):
+         """Return whether the table has a column with the given name or index
+         """
+         return self.column_map.has_key(col)
+     def RowIdToOrdinal(self, gid):
+         """Return the row ordinal given its id
+         At the moment the transient tables are only used for tables that
+         don't distinguish between row number and row id, so the value is
+         returned unchanged.
+         """
+         return gid
+     def RowOrdinalToId(self, num):
+         """Return the rowid for given its ordinal
+         At the moment the transient tables are only used for tables that
+         don't distinguish between row number and row id, so the value is
+         returned unchanged.
+         """
+         return num
+     def ReadRowAsDict(self, index, row_is_ordinal = 0):
+         """Return the entire row as a dictionary with column names as keys
+         The row_is_ordinal is ignored because at the moment the
+         transient tables are only used for DBF files where it doesn't
+         matter.
+         """
+         # Implementation Strategy: Executing a completely new select
+         # statement every time this method is called is too slow. The
+         # most important usage is to read the records more or less
+         # sequentially. This happens e.g. when drawing a layer with a
+         # classification where the shapes are drawn in order of the
+         # shape ids. Another pattern is that the same row is requested
+         # several times in a row. This happens in the table view, for
+         # instance.
+         # We can exploit this to make access faster by having one cursor
+         # open all the time and keeping the last row read around in case
+         # the same row is accessed again the next time and if the row
+         # index is larger than the row we have read last we simply fetch
+         # rows from the cursor until we've reached the requested row. If
+         # the requested row index is smaller then we start a new cursor.
+         # FIXME: So far this scheme seems to work well enough. Obvious
+         # improvements would be to start the cursor at exactly the
+         # requested row (should be efficient and easy to do now that the
+         # id is the primary key) and to perhaps to also start a new
+         # cursor if the requested index is much larger than the last row
+         # so that we don't read and discard lots of the rows.
+         # Check whether we have to start a new cursor
          if self.read_record_cursor is None or index <self.read_record_last_row:
              stmt = ("SELECT %s FROM %s;"
                      % (", ".join([c.internal_name for c in self.columns]),
-Line 163 
 class TransientTableBase:
+Line 216 
 class TransientTableBase:
              self.read_record_cursor = self.db.cursor()
              self.read_record_cursor.execute(stmt)
              self.read_record_last_row = -1
-         for i in range(index - self.read_record_last_row):
+             self.read_record_last_result = None
-             result = self.read_record_cursor.fetchone()
+         # Now we should have a cursor at a position less than or equal
+         # to the index so the following if statement will always set
+         # result to a suitable value
+         assert index >= self.read_record_last_row
+         if index == self.read_record_last_row:
+             result = self.read_record_last_result
+         else:
+             for i in range(index - self.read_record_last_row):
+                 result = self.read_record_cursor.fetchone()
+                 self.read_record_last_result = result
          self.read_record_last_row = index
-         result = dict(zip(self.orig_names, result))
+         return dict(zip(self.orig_names, result))
-         return result
+     def ReadValue(self, row, col, row_is_ordinal = 0):
+         """Return the value of the specified row and column
-     def field_range(self, colname):
+         The col parameter may be the index of the column or its name.
-         col = self.name_to_column[colname]
+         The row_is_ordinal is ignored because at the moment the
+         transient tables are only used for DBF files where it doesn't
+         matter.
+         """
+         # Depending on the actual access patterns of the table data, it
+         # might be a bit faster in some circumstances to not implement
+         # this via ReadRowAsDict, but this simple implementation should
+         # be fast enough for most purposes.
+         return self.ReadRowAsDict(row)[self.column_map[col].name]
+     def ValueRange(self, col):
+         # Performance notes:
+         #
+         # In sqlite 2.8.6 the min and max aggregate functions can use an
+         # index but only when used as the only expression in the select
+         # statement (i.e. 'select min(col), max(col) from tbl;' will not
+         # use the index but 'select min(col) from tbl;' will) so we
+         # query the minimum and maximum separately.
+         #
+         # With the separate statements we can take advantage of an index
+         # if it exists. If the index doesn't exist, creating it first
+         # and then using it in the query is slower than the queries
+         # without an index. Creating the index is only an advantage if
+         # the queries are performed multiple times. With the current use
+         # patterns where ValueRange is only used occasionally by the
+         # classification generation dialog creating the index only for
+         # this usage is not really worth it, so we don't.
+         col = self.column_map[col]
          iname = col.internal_name
-         min, max = self.db.execute("SELECT min(%s), max(%s) FROM %s;"
+         min = self.db.execute("SELECT min(%s) FROM %s;"
-                                    % (iname, iname, self.tablename))
+                               % (iname, self.tablename))[0]
+         max = self.db.execute("SELECT max(%s) FROM %s;"
+                               % (iname, self.tablename))[0]
          converter = type_converter_map[col.type]
-         return ((converter(min), None), (converter(max), None))
+         return (converter(min), converter(max))
-     def GetUniqueValues(self, colname):
+     def UniqueValues(self, col):
-         iname = self.orig_to_internal[colname]
+         # Performance notes:
+         #
+         # In sqlite 2.8.6 there doesn't seem to be a way to query the
+         # unique items that uses an index. I've tried
+         #
+         #   SELECT col FROM tbl GROUP BY col;
+         #
+         # and
+         #
+         #   SELECT DISTINCT col FROM tbl;
+         #
+         # and in both cases the index is not used. If the index isn't
+         # used it doesn't make sense to call self.ensure_index.
+         iname = self.column_map[col].internal_name
          cursor = self.db.cursor()
          cursor.execute("SELECT %s FROM %s GROUP BY %s;"
                         % (iname, self.tablename, iname))
-Line 190 
 class TransientTableBase:
+Line 299 
 class TransientTableBase:
              result.append(row[0])
          return result
+     def Width(self, col):
+         """Return the maximum width of values in the column
+         The return value is the the maximum length of string
+         representation of the values in the column (represented by index
+         or name).
+         """
+         max = 0
+         type  = self.column_map[col].type
+         iname = self.column_map[col].internal_name
+         cursor = self.db.cursor()
+         cursor.execute("SELECT %s FROM %s;" % (iname, self.tablename))
+         values = [ i[0] for i in cursor.fetchall()]
+         if not values:
+             return None
+         if type == table.FIELDTYPE_DOUBLE:
+             format = "%.12f"
+         elif type == table.FIELDTYPE_INT:
+             format = "%d"
+         else:
+             format = "%s"
+         for value in values:
+             if value is None: continue
+             l = len(format % value)
+             if l > max:
+                 max = l
+         return max
+     def SimpleQuery(self, left, comparison, right):
+         """Return the indices of all rows that matching a condition.
+         Parameters:
+            left -- The column object for the left side of the comparison
+            comparison -- The comparison operator as a string. It must be
+                          one of '==', '!=', '<', '<=', '>=', '>'
+            right -- The right hand side of the comparison. It must be
+                     either a column object or a value, i.e. a string,
+                     int or float.
+         The return value is a sorted list of the indices of the rows
+         where the condition is true.
+         """
+         if comparison not in ("==", "!=", "<", "<=", ">=", ">"):
+             raise ValueError("Comparison operator %r not allowed" % comparison)
+         if hasattr(right, "internal_name"):
+             right_template = right.internal_name
+             params = ()
+         else:
+             right_template = "%s"
+             params = (right,)
+         query = "SELECT id FROM %s WHERE %s %s %s ORDER BY id;" \
+                 % (self.tablename, left.internal_name, comparison,
+                    right_template)
+         cursor = self.db.cursor()
+         cursor.execute(query, params)
+         result = []
+         while 1:
+             row = cursor.fetchone()
+             if row is None:
+                 break
+             result.append(row[0])
+         return result
+     def Dependencies(self):
+         """Placeholder for a method in a derived class.
- class TransientTable(TransientTableBase):
+         Return a sequence with the tables and other data objects that
+         self depends on.
+         """
+         raise NotImplementedError
+ class TransientTable(TitledObject, TransientTableBase):
      """A Table in a transient DB that starts as the copy of a Thuban Table."""
-Line 202 
 class TransientTable(TransientTableBase)
+Line 390 
 class TransientTable(TransientTableBase)
          interface.
          """
          TransientTableBase.__init__(self, transient_db)
+         TitledObject.__init__(self, table.Title())
          self.create(table)
      def create(self, table):
          columns = []
-         for i in range(table.field_count()):
+         for col in table.Columns():
-             type, name = table.field_info(i)[:2]
+             columns.append(ColumnReference(col.name, col.type,
-             columns.append(ColumnReference(name, type,
                                             self.db.new_column_name()))
          TransientTableBase.create(self, columns)
          # copy the input table to the transient db
-         insert_template = "INSERT INTO %s (%s) VALUES (%s);" \
+         # A key to insert to use for the formatting of the insert
+         # statement. The key must not be equal to any of the column
+         # names so we construct one by building a string of x's that is
+         # longer than any of the column names
+         id_key = max([len(col.name) for col in self.columns]) * "x"
+         insert_template = "INSERT INTO %s (id, %s) VALUES (%%(%s)s, %s);" \
                                 % (self.tablename,
                                    ", ".join([col.internal_name
                                               for col in self.columns]),
+                                   id_key,
                                    ", ".join(["%%(%s)s" % col.name
                                               for col in self.columns]))
          cursor = self.db.cursor()
-         for i in range(table.record_count()):
+         for i in range(table.NumRows()):
-             cursor.execute(insert_template, table.read_record(i))
+             row = table.ReadRowAsDict(i)
+             row[id_key] = i
+             cursor.execute(insert_template, row)
          self.db.conn.commit()
- class TransientJoinedTable(TransientTableBase):
+ class TransientJoinedTable(TitledObject, TransientTableBase):
      """A Table in the transient DB that contains a join of two tables"""
      def __init__(self, transient_db, left_table, left_field,
-                  right_table, right_field = None):
+                  right_table, right_field = None, outer_join = False):
          """Create a new table in the transient DB as a join of two tables.
          Both input tables, left_table and right_table must have a
          transient_table method that returns a table object for a table
-         in the trnsient database. The join is performed on the condition
+         in the transient database. The join is performed on the condition
          that the value of the left_field column the the left table is
          equal to the value of the right_field in the right_table.
-         The joined table contains all columns of the input tables with
+         The joined table contains all columns of the input tables,
-         one exception: Any column in the right_table with the same name
+         however, the column names of the right table may be changed
-         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
-         somewhat of an implementation detail, but is done so that the
+         currently done by appending a sufficient number of underscores
-         column names of the joined table can be the same as the column
+         ('_').
-         names of the input tables without having to create prefixes.
          """
          TransientTableBase.__init__(self, transient_db)
+         self.dependencies = (left_table, right_table)
          self.left_table = left_table.transient_table()
          self.left_field = left_field
          self.right_table = right_table.transient_table()
-Line 255 
 class TransientJoinedTable(TransientTabl
+Line 453 
 class TransientJoinedTable(TransientTabl
              self.right_field = right_field
          else:
              self.right_field = self.left_field
+         self.outer_join = outer_join
+         title = "Join of %(left)s and %(right)s" \
+                 % {"left": self.left_table.Title(),
+                    "right": self.right_table.Title()}
+         TitledObject.__init__(self, title)
          self.create()
      def create(self):
-Line 263 
 class TransientJoinedTable(TransientTabl
+Line 468 
 class TransientJoinedTable(TransientTabl
          self.right_table.ensure_index(self.right_field)
+         # determine the internal column names to join on before
+         # coalescing the column information because if the external
+         # column names are the same they will be mapped to the same
+         # internal name afterwards.
+         internal_left_col = self.left_table.orig_to_internal[self.left_field]
+         internal_right_col =self.right_table.orig_to_internal[self.right_field]
          # Coalesce the column information
          visited = {}
          columns = []
-         for col in self.left_table.columns + self.right_table.columns:
+         newcolumns = []
-             if col.name in visited:
+         for table in (self.left_table, self.right_table):
-                 continue
+             for col in table.Columns():
-             columns.append(col)
+                 colname = col.name
-         TransientTableBase.create(self, columns)
+                 # We can't allow multiple columns with the same
+                 # original name, so append '_' to this one until
+                 # it is unique.
+                 # FIXME: There should be a better solution.
+                 while colname in visited:
+                     colname = colname + '_'
+                 columns.append((table.tablename, col))
+                 newcol = ColumnReference(colname, col.type,
+                                             "Col%03d" % (len(newcolumns)+1))
+                 newcolumns.append(newcol)
+                 visited[colname] = 1
+         TransientTableBase.create(self, newcolumns)
          # Copy the joined data to the table.
-         internal_names = [col.internal_name for col in self.columns]
+         newinternal_names = [col.internal_name for col in self.columns]
-         stmt = "INSERT INTO %s (%s) SELECT %s FROM %s JOIN %s ON %s = %s;" \
+         internal_references = ["%s.%s" % (table, col.internal_name)
-                % (self.tablename,
+                                                     for table, col in columns]
-                   ", ".join(internal_names),
+         if self.outer_join:
-                   ", ".join(internal_names),
+             join_operator = 'LEFT OUTER JOIN'
-                   self.left_table.tablename,
+         else:
-                   self.right_table.tablename,
+             join_operator = 'JOIN'
-                   self.orig_to_internal[self.left_field],
+         stmt = ("INSERT INTO %s (id, %s) SELECT %s.id, %s FROM %s"
-                   self.orig_to_internal[self.right_field])
+                 " %s %s ON %s.%s = %s.%s;"
+                 % (self.tablename,
+                    ", ".join(newinternal_names),
+                    self.left_table.tablename,
+                    ", ".join(internal_references),
+                    self.left_table.tablename,
+                    join_operator,
+                    self.right_table.tablename,
+                    self.left_table.tablename,
+                    internal_left_col,
+                    self.right_table.tablename,
+                    internal_right_col))
          self.db.execute(stmt)
+     def Dependencies(self):
+         """Return a tuple with the two tables the join depends on."""
+         return self.dependencies
+     def JoinType(self):
+         """Return the type of the join (either 'INNER' or 'LEFT OUTER')"""
+         if self.outer_join:
+             return "LEFT OUTER"
+         else:
+             return "INNER"
- class AutoTransientTable:
+ class AutoTransientTable(TitledObject, table.OldTableInterfaceMixin):
      """Table that copies data to a transient table on demand.
-Line 294 
 class AutoTransientTable:
+Line 539 
 class AutoTransientTable:
      """
      def __init__(self, transient_db, table):
+         TitledObject.__init__(self, table.Title())
          self.transient_db = transient_db
          self.table = table
          self.t_table = None
-     def record_count(self):
+     def Columns(self):
-         """Return the number of records"""
+         return self.table.Columns()
-         return self.table.record_count()
+     def Column(self, col):
-     def field_count(self):
+         return self.table.Column(col)
-         """Return the number of fields in a record"""
-         return self.table.field_count()
+     def HasColumn(self, col):
+         """Return whether the table has a column with the given name or index
-     def field_info(self, field):
+         """
-         """Return a tuple (type, name, width, prec) for the field no. field
+         return self.table.HasColumn(col)
-         type is the data type of the field, name the name, width the
+     def NumRows(self):
-         field width in characters and prec the decimal precision.
+         return self.table.NumRows()
-         """
-         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
-     def read_record(self, record):
+     def NumColumns(self):
+         return self.table.NumColumns()
+     def RowIdToOrdinal(self, gid):
+         """Return the row ordinal given its id"""
+         if self.t_table is not None:
+             return self.t_table.RowIdToOrdinal(gid)
+         else:
+             return self.table.RowIdToOrdinal(gid)
+     def RowOrdinalToId(self, num):
+         """Return the rowid for given its ordinal"""
+         if self.t_table is not None:
+             return self.t_table.RowOrdinalToId(num)
+         else:
+             return self.table.RowOrdinalToId(num)
+     def ReadRowAsDict(self, record, row_is_ordinal = 0):
          """Return the record no. record as a dict mapping field names to values
          """
          if self.t_table is not None:
-             return self.t_table.read_record(record)
+             return self.t_table.ReadRowAsDict(record,
+                                               row_is_ordinal = row_is_ordinal)
          else:
-             return self.table.read_record(record)
+             return self.table.ReadRowAsDict(record,
+                                             row_is_ordinal = row_is_ordinal)
-     def write_record(self, record, values):
+     def ReadValue(self, row, col, row_is_ordinal = 0):
-         raise NotImplementedError
+         """Return the value of the specified row and column
+         The col parameter may be the index of the column or its name.
+         """
+         if self.t_table is not None:
+             return self.t_table.ReadValue(row, col,
+                                           row_is_ordinal = row_is_ordinal)
+         else:
+             return self.table.ReadValue(row, col,
+                                         row_is_ordinal = row_is_ordinal)
      def copy_to_transient(self):
          """Internal: Create a transient table and copy the data into it"""
-Line 346 
 class AutoTransientTable:
+Line 609 
 class AutoTransientTable:
              self.copy_to_transient()
          return self.t_table
-     def field_range(self, colname):
+     def ValueRange(self, col):
-         if self.t_table is None:
+         # Performance of sqlite vs. DBF for this method:
-             self.copy_to_transient()
+         #
-         return self.t_table.field_range(colname)
+         # If the table has been copied to the sqlite database it's
+         # faster to use it even if there is no index on that column.
+         # Otherwise it's faster to simply loop through all rows in the
+         # DBF file. Copying the data to the sqlite database can take
+         # very long for large amounts of data
+         #
+         # Of course if the table is not a DBF file the issue could be
+         # different, although copying the data into sqlite first will
+         # likely always be slower than simply querying the non-sqlite
+         # table directly. Currently only DBFfiles and memory tables are
+         # used as the underlying non-sqlite table, though.
+         if self.t_table is not None:
+             return self.t_table.ValueRange(col)
+         else:
+             return self.table.ValueRange(col)
-     def GetUniqueValues(self, colname):
+     def UniqueValues(self, col):
+         # The performance trade-offs for this method are basically the
+         # same as for ValueRange except that currently there doesn't
+         # seem to be a way to take advantage of indexes in this case in
+         # sqlite. However, but it's still faster to query the transient
+         # table if it already exists.
+         if self.t_table is not None:
+             return self.t_table.UniqueValues(col)
+         else:
+             return self.table.UniqueValues(col)
+     def SimpleQuery(self, left, comparison, right):
          if self.t_table is None:
              self.copy_to_transient()
-         return self.t_table.GetUniqueValues(colname)
+         # Make sure to use the column object of the transient table. The
+         # left argument is always a column object so we can just ask the
+         # t_table for the right object.
+         if hasattr(right, "name"):
+             return self.t_table.SimpleQuery(self.t_table.Column(left.name),
+                                             comparison,
+                                             self.t_table.Column(right.name))
+         else:
+             return self.t_table.SimpleQuery(self.t_table.Column(left.name),
+                                             comparison, right)
+     def Dependencies(self):
+         """Return a tuple containing the original table"""
+         return (self.table,)
+     def Width(self, col):
+         return self.table.Width(col)

 Legend:



Removed from v.777
 


changed lines


 
Added in v.1923
 Legend:



Removed from v.777
 


changed lines


 
Added in v.1923
-Removed from v.777
+Added in v.1923

[email protected]	ViewVC Help
Powered by ViewVC 1.1.26