| Server IP : 180.180.241.3 / Your IP : 216.73.216.35 Web Server : Microsoft-IIS/7.5 System : Windows NT NETWORK-NHRC 6.1 build 7601 (Windows Server 2008 R2 Standard Edition Service Pack 1) i586 User : IUSR ( 0) PHP Version : 5.3.28 Disable Function : NONE MySQL : ON | cURL : ON | WGET : OFF | Perl : OFF | Python : OFF | Sudo : OFF | Pkexec : OFF Directory : C:/Program Files/MySQL/MySQL Workbench 6.3 CE/modules/ |
Upload File : |
# Copyright (c) 2012, 2014, Oracle and/or its affiliates. All rights reserved.
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License as
# published by the Free Software Foundation; version 2 of the
# License.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
# 02110-1301 USA
import grt
from grt.modules import Workbench
import mforms
from wb import wbinputs
from wb_utils_grt import ModuleInfo
@ModuleInfo.plugin('wb.util.copySQLToClipboard', caption='Copy SQL to Clipboard', input= [wbinputs.objectOfClass('db.DatabaseObject')], groups= ['Catalog/Utilities', 'Menu/Objects'])
@ModuleInfo.export(grt.INT, grt.classes.GrtNamedObject)
def copySQLToClipboard(obj):
script = []
# workaround until diff sql generator handles routine groups
if isinstance(obj, grt.classes.db_RoutineGroup):
for routine in obj.routines:
script.append(grt.modules.DbMySQL.makeCreateScriptForObject(routine))
script.append(';\n\n')
else:
script.append(grt.modules.DbMySQL.makeCreateScriptForObject(obj))
Workbench.copyToClipboard(''.join(script))
return 0
@ModuleInfo.plugin('wb.util.copyColumnNamesToClipboard', caption='Copy Column Names to Clipboard', input= [wbinputs.objectOfClass('db.Table')], groups= ['Catalog/Utilities', 'Menu/Objects'])
@ModuleInfo.export(grt.INT, grt.classes.db_Table)
def copyColumnNamesToClipboard(table):
data = ', '.join([column.name for column in table.columns])
Workbench.copyToClipboard(data)
return 0
@ModuleInfo.plugin('wb.util.copyTableListToClipboard', caption='Copy Table List to Clipboard', input= [wbinputs.currentCatalog()], groups= ['Catalog/Utilities', 'Menu/Catalog'])
@ModuleInfo.export(grt.INT, grt.classes.db_Catalog)
def copyTableListToClipboard(cat):
#insert = ['`'+schema.name+'`.`'+tbl.name+'`' for tbl in schema.tables for schema in cat.schemata ]
insert = ''
for schema in cat.schemata:
insert = insert + ', '.join(['`'+schema.name+'`.`'+tbl.name+'`' for tbl in schema.tables])
Workbench.copyToClipboard(insert)
return 0
def generateName(name_prefix, names_map):
name_suffix = 0
byte_a = ord('a')
while True:
for i in range(0, 25):
next_name = ''
if name_prefix == '':
next_name = chr(byte_a + i) + str(name_suffix)
else:
next_name = name_prefix + str(name_suffix)
if not next_name in names_map:
names_map[next_name] = 1
return next_name
name_suffix = name_suffix + 1
@ModuleInfo.plugin('wb.util.obfuscateCatalog', caption='Obfuscate Object Names in Catalog', input= [wbinputs.currentModel()], groups= ['Catalog/Utilities', 'Menu/Utilities'])
@ModuleInfo.export(grt.INT, grt.classes.workbench_physical_Model)
def obfuscateCatalog(model):
result = mforms.Utilities.show_warning('Warning', '''This operation will change names of all schemata and tables in the model.
Would you like to also clear SQL code for views and stored procedures?
This action cannot be undone!''', 'Clear SQL', 'Cancel', 'Leave SQL')
if result == mforms.ResultCancel:
return 1
clear_sql_code = (result == mforms.ResultOk)
schemata_map = {}
objects_map = {}
def createFiguresMap(model):
figures_map = {}
for diagram in model.diagrams:
for figure in diagram.figures:
if hasattr(figure, 'table'):
figures_map[figure.table.__id__] = figure
elif hasattr(figure, 'view'):
figures_map[figure.view.__id__] = figure
elif hasattr(figure, 'routine'):
figures_map[figure.routine.__id__] = figure
elif hasattr(figure, 'routineGroup'):
figures_map[figure.routineGroup.__id__] = figure
return figures_map
figures_map = createFiguresMap(model)
def renameFigure(obj, figures_map):
if obj.__id__ in figures_map:
figures_map[obj.__id__].name = obj.name
def renameDbObject(obj, obj_map=objects_map, clear_sql_code=clear_sql_code, figures_map=figures_map):
obj.name = generateName('', obj_map)
if clear_sql_code and hasattr(obj, 'sqlBody'):
obj.sqlBody = ''
if clear_sql_code and hasattr(obj, 'sqlDefinition'):
obj.sqlDefinition = ''
renameFigure(obj, figures_map)
for schema in model.catalog.schemata:
schema.name = generateName('', schemata_map)
for table in schema.tables:
renameDbObject(table)
for view in schema.views:
renameDbObject(view)
for routine in schema.routines:
renameDbObject(routine)
for routineGroup in schema.routineGroups:
renameDbObject(routineGroup)
for routine in routineGroup.routines:
renameDbObject(routine)
return 0
@ModuleInfo.plugin('wb.util.prefixTables', caption='Give a Prefix to All Tables in Catalog', input= [wbinputs.currentCatalog()], groups= ['Catalog/Utilities', 'Menu/Catalog'])
@ModuleInfo.export(grt.INT, grt.classes.db_Catalog)
def prefixTables(cat):
ret, prefix = mforms.Utilities.request_input("Give a Prefix to All Tables in Catalog", "Please specify the prefix:copy table ", "")
if not ret:
return 1
for schema in cat.schemata:
for tbl in schema.tables:
tbl.name = prefix + tbl.name
return 0
@ModuleInfo.plugin('wb.util.changeStorageEngines', caption='Change the Storage Engine of All Tables', input= [wbinputs.currentCatalog()], groups= ['Catalog/Utilities', 'Menu/Catalog'])
@ModuleInfo.export(grt.INT, grt.classes.db_Catalog)
def changeStorageEngines(cat):
ret, new_engine = mforms.Utilities.request_input("Change the Storage Engine of All Tables", "Type the new storage engine name:", "")
if not ret:
return 1
def getTableEngines():
result = grt.root.wb.options.options['@db.mysql.Table:tableEngine/Items']
items = [item.strip(' \t') for item in result.split(',')]
return items
# validate the engine name and fix its case
engines = getTableEngines()
engine_found = False
for engine_name in engines:
if engine_name.find(':') != -1:
engine_name = engine_name[engine_name.find(':') + 1:]
if new_engine.lower() == engine_name.lower():
engine_found = True
new_engine = engine_name
if not engine_found:
Workbench.confirm('Change Storage Engines', 'Invalid storage engine name: ' + new_engine)
return 2
for schema in cat.schemata:
for tbl in schema.tables:
tbl.tableEngine = new_engine
return 0