Page MenuHomec4science
Files F65176534

dbquery.py
No OneTemporary
Actions

File Metadata

Created
Sat, Jun 1, 13:11

dbquery.py
View Options

## This file is part of Invenio.
## Copyright (C) 2008, 2009, 2010, 2011, 2012 CERN.
##
## Invenio 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; either version 2 of the
## License, or (at your option) any later version.
##
## Invenio 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 Invenio; if not, write to the Free Software Foundation, Inc.,
## 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA.
"""
Invenio utilities to run SQL queries.
The main API functions are:
- run_sql()
- run_sql_many()
- run_sql_with_limit()
but see the others as well.
"""
__revision__ = "$Id$"
# dbquery clients can import these from here:
# pylint: disable=W0611
from MySQLdb import Warning, Error, InterfaceError, DataError, \
DatabaseError, OperationalError, IntegrityError, \
InternalError, NotSupportedError, \
ProgrammingError
import gc
import os
import string
import time
import marshal
import re
import atexit
from zlib import compress, decompress
from thread import get_ident
from invenio.config import CFG_ACCESS_CONTROL_LEVEL_SITE, \
CFG_MISCUTIL_SQL_USE_SQLALCHEMY, \
CFG_MISCUTIL_SQL_RUN_SQL_MANY_LIMIT
if CFG_MISCUTIL_SQL_USE_SQLALCHEMY:
try:
import sqlalchemy.pool as pool
import MySQLdb as mysqldb
mysqldb = pool.manage(mysqldb, use_threadlocal=True)
connect = mysqldb.connect
except ImportError:
CFG_MISCUTIL_SQL_USE_SQLALCHEMY = False
from MySQLdb import connect
else:
from MySQLdb import connect
## DB config variables. These variables are to be set in
## invenio-local.conf by admins and then replaced in situ in this file
## by calling "inveniocfg --update-dbexec".
## Note that they are defined here and not in config.py in order to
## prevent them from being exported accidentally elsewhere, as no-one
## should know DB credentials but this file.
## FIXME: this is more of a blast-from-the-past that should be fixed
## both here and in inveniocfg when the time permits.
CFG_DATABASE_HOST = 'localhost'
CFG_DATABASE_PORT = '3306'
CFG_DATABASE_NAME = 'invenio'
CFG_DATABASE_USER = 'invenio'
CFG_DATABASE_PASS = 'my123p$ss'
CFG_DATABASE_SLAVE = ''
_DB_CONN = {}
_DB_CONN[CFG_DATABASE_HOST] = {}
_DB_CONN[CFG_DATABASE_SLAVE] = {}
def unlock_all():
for dbhost in _DB_CONN.keys():
for db in _DB_CONN[dbhost].values():
try:
cur = db.cur()
cur.execute("UNLOCK TABLES")
except:
pass
atexit.register(unlock_all)
class InvenioDbQueryWildcardLimitError(Exception):
"""Exception raised when query limit reached."""
def __init__(self, res):
"""Initialization."""
self.res = res
def _db_login(dbhost=CFG_DATABASE_HOST, relogin=0):
"""Login to the database."""
## Note: we are using "use_unicode=False", because we want to
## receive strings from MySQL as Python UTF-8 binary string
## objects, not as Python Unicode string objects, as of yet.
## Note: "charset='utf8'" is needed for recent MySQLdb versions
## (such as 1.2.1_p2 and above). For older MySQLdb versions such
## as 1.2.0, an explicit "init_command='SET NAMES utf8'" parameter
## would constitute an equivalent. But we are not bothering with
## older MySQLdb versions here, since we are recommending to
## upgrade to more recent versions anyway.
if CFG_MISCUTIL_SQL_USE_SQLALCHEMY:
return connect(host=dbhost, port=int(CFG_DATABASE_PORT),
db=CFG_DATABASE_NAME, user=CFG_DATABASE_USER,
passwd=CFG_DATABASE_PASS,
use_unicode=False, charset='utf8')
else:
thread_ident = (os.getpid(), get_ident())
if relogin:
connection = _DB_CONN[dbhost][thread_ident] = connect(host=dbhost,
port=int(CFG_DATABASE_PORT),
db=CFG_DATABASE_NAME,
user=CFG_DATABASE_USER,
passwd=CFG_DATABASE_PASS,
use_unicode=False, charset='utf8')
connection.autocommit(True)
return connection
else:
if _DB_CONN[dbhost].has_key(thread_ident):
return _DB_CONN[dbhost][thread_ident]
else:
connection = _DB_CONN[dbhost][thread_ident] = connect(host=dbhost,
port=int(CFG_DATABASE_PORT),
db=CFG_DATABASE_NAME,
user=CFG_DATABASE_USER,
passwd=CFG_DATABASE_PASS,
use_unicode=False, charset='utf8')
connection.autocommit(True)
return connection
def _db_logout(dbhost=CFG_DATABASE_HOST):
"""Close a connection."""
try:
del _DB_CONN[dbhost][(os.getpid(), get_ident())]
except KeyError:
pass
def close_connection(dbhost=CFG_DATABASE_HOST):
"""
Enforce the closing of a connection
Highly relevant in multi-processing and multi-threaded modules
"""
try:
_DB_CONN[dbhost][(os.getpid(), get_ident())].close()
del(_DB_CONN[dbhost][(os.getpid(), get_ident())])
except KeyError:
pass
def run_sql(sql, param=None, n=0, with_desc=False, with_dict=False, run_on_slave=False):
"""Run SQL on the server with PARAM and return result.
@param param: tuple of string params to insert in the query (see
notes below)
@param n: number of tuples in result (0 for unbounded)
@param with_desc: if True, will return a DB API 7-tuple describing
columns in query.
@param with_dict: if True, will return a list of dictionaries
composed of column-value pairs
@return: If SELECT, SHOW, DESCRIBE statements, return tuples of data,
followed by description if parameter with_desc is
provided.
If SELECT and with_dict=True, return a list of dictionaries
composed of column-value pairs, followed by description
if parameter with_desc is provided.
If INSERT, return last row id.
Otherwise return SQL result as provided by database.
@note: When the site is closed for maintenance (as governed by the
config variable CFG_ACCESS_CONTROL_LEVEL_SITE), do not attempt
to run any SQL queries but return empty list immediately.
Useful to be able to have the website up while MySQL database
is down for maintenance, hot copies, table repairs, etc.
@note: In case of problems, exceptions are returned according to
the Python DB API 2.0. The client code can import them from
this file and catch them.
"""
if CFG_ACCESS_CONTROL_LEVEL_SITE == 3:
# do not connect to the database as the site is closed for maintenance:
return []
if param:
param = tuple(param)
dbhost = CFG_DATABASE_HOST
if run_on_slave and CFG_DATABASE_SLAVE:
dbhost = CFG_DATABASE_SLAVE
### log_sql_query(dbhost, sql, param) ### UNCOMMENT ONLY IF you REALLY want to log all queries
try:
db = _db_login(dbhost)
cur = db.cursor()
gc.disable()
rc = cur.execute(sql, param)
gc.enable()
except (OperationalError, InterfaceError): # unexpected disconnect, bad malloc error, etc
# FIXME: now reconnect is always forced, we may perhaps want to ping() first?
try:
db = _db_login(dbhost, relogin=1)
cur = db.cursor()
gc.disable()
rc = cur.execute(sql, param)
gc.enable()
except (OperationalError, InterfaceError): # unexpected disconnect, bad malloc error, etc
raise
if string.upper(string.split(sql)[0]) in ("SELECT", "SHOW", "DESC", "DESCRIBE"):
if n:
recset = cur.fetchmany(n)
else:
recset = cur.fetchall()
if with_dict: # return list of dictionaries
# let's extract column names
keys = [row[0] for row in cur.description]
# let's construct a list of dictionaries
list_dict_results = [dict(zip(*[keys, values])) for values in recset]
if with_desc:
return list_dict_results, cur.description
else:
return list_dict_results
else:
if with_desc:
return recset, cur.description
else:
return recset
else:
if string.upper(string.split(sql)[0]) == "INSERT":
rc = cur.lastrowid
return rc
def run_sql_many(query, params, limit=CFG_MISCUTIL_SQL_RUN_SQL_MANY_LIMIT, run_on_slave=False):
"""Run SQL on the server with PARAM.
This method does executemany and is therefore more efficient than execute
but it has sense only with queries that affect state of a database
(INSERT, UPDATE). That is why the results just count number of affected rows
@param params: tuple of tuple of string params to insert in the query
@param limit: query will be executed in parts when number of
parameters is greater than limit (each iteration runs at most
`limit' parameters)
@return: SQL result as provided by database
"""
dbhost = CFG_DATABASE_HOST
if run_on_slave and CFG_DATABASE_SLAVE:
dbhost = CFG_DATABASE_SLAVE
i = 0
r = None
while i < len(params):
## make partial query safely (mimicking procedure from run_sql())
try:
db = _db_login(dbhost)
cur = db.cursor()
gc.disable()
rc = cur.executemany(query, params[i:i + limit])
gc.enable()
except (OperationalError, InterfaceError):
try:
db = _db_login(dbhost, relogin=1)
cur = db.cursor()
gc.disable()
rc = cur.executemany(query, params[i:i + limit])
gc.enable()
except (OperationalError, InterfaceError):
raise
## collect its result:
if r is None:
r = rc
else:
r += rc
i += limit
return r
def run_sql_with_limit(query, param=None, n=0, with_desc=False, wildcard_limit=0, run_on_slave=False):
"""This function should be used in some cases, instead of run_sql function, in order
to protect the db from queries that might take a log time to respond
Ex: search queries like [a-z]+ ; cern*; a->z;
The parameters are exactly the ones for run_sql function.
In case the query limit is reached, an InvenioDbQueryWildcardLimitError will be raised.
"""
try:
dummy = int(wildcard_limit)
except ValueError:
raise
if wildcard_limit < 1:#no limit on the wildcard queries
return run_sql(query, param, n, with_desc, run_on_slave=run_on_slave)
safe_query = query + " limit %s" %wildcard_limit
res = run_sql(safe_query, param, n, with_desc, run_on_slave=run_on_slave)
if len(res) == wildcard_limit:
raise InvenioDbQueryWildcardLimitError(res)
return res
def blob_to_string(ablob):
"""Return string representation of ABLOB. Useful to treat MySQL
BLOBs in the same way for both recent and old MySQLdb versions.
"""
if ablob:
if type(ablob) is str:
# BLOB is already a string in MySQLdb 0.9.2
return ablob
else:
# BLOB is array.array in MySQLdb 1.0.0 and later
return ablob.tostring()
else:
return ablob
def log_sql_query(dbhost, sql, param=None):
"""Log SQL query into prefix/var/log/dbquery.log log file. In order
to enable logging of all SQL queries, please uncomment one line
in run_sql() above. Useful for fine-level debugging only!
"""
from invenio.config import CFG_LOGDIR
from invenio.dateutils import convert_datestruct_to_datetext
from invenio.textutils import indent_text
log_path = CFG_LOGDIR + '/dbquery.log'
date_of_log = convert_datestruct_to_datetext(time.localtime())
message = date_of_log + '-->\n'
message += indent_text('Host:\n' + indent_text(str(dbhost), 2, wrap=True), 2)
message += indent_text('Query:\n' + indent_text(str(sql), 2, wrap=True), 2)
message += indent_text('Params:\n' + indent_text(str(param), 2, wrap=True), 2)
message += '-----------------------------\n\n'
try:
log_file = open(log_path, 'a+')
log_file.writelines(message)
log_file.close()
except:
pass
def get_table_update_time(tablename, run_on_slave=False):
"""Return update time of TABLENAME. TABLENAME can contain
wildcard `%' in which case we return the maximum update time
value.
"""
# Note: in order to work with all of MySQL 4.0, 4.1, 5.0, this
# function uses SHOW TABLE STATUS technique with a dirty column
# position lookup to return the correct value. (Making use of
# Index_Length column that is either of type long (when there are
# some indexes defined) or of type None (when there are no indexes
# defined, e.g. table is empty). When we shall use solely
# MySQL-5.0, we can employ a much cleaner technique of using
# SELECT UPDATE_TIME FROM INFORMATION_SCHEMA.TABLES WHERE
# table_name='collection'.
res = run_sql("SHOW TABLE STATUS LIKE %s", (tablename,),
run_on_slave=run_on_slave)
update_times = [] # store all update times
for row in res:
if type(row[10]) is long or \
row[10] is None:
# MySQL-4.1 and 5.0 have creation_time in 11th position,
# so return next column:
update_times.append(str(row[12]))
else:
# MySQL-4.0 has creation_time in 10th position, which is
# of type datetime.datetime or str (depending on the
# version of MySQLdb), so return next column:
update_times.append(str(row[11]))
return max(update_times)
def get_table_status_info(tablename, run_on_slave=False):
"""Return table status information on TABLENAME. Returned is a
dict with keys like Name, Rows, Data_length, Max_data_length,
etc. If TABLENAME does not exist, return empty dict.
"""
# Note: again a hack so that it works on all MySQL 4.0, 4.1, 5.0
res = run_sql("SHOW TABLE STATUS LIKE %s", (tablename,),
run_on_slave=run_on_slave)
table_status_info = {} # store all update times
for row in res:
if type(row[10]) is long or \
row[10] is None:
# MySQL-4.1 and 5.0 have creation time in 11th position:
table_status_info['Name'] = row[0]
table_status_info['Rows'] = row[4]
table_status_info['Data_length'] = row[6]
table_status_info['Max_data_length'] = row[8]
table_status_info['Create_time'] = row[11]
table_status_info['Update_time'] = row[12]
else:
# MySQL-4.0 has creation_time in 10th position, which is
# of type datetime.datetime or str (depending on the
# version of MySQLdb):
table_status_info['Name'] = row[0]
table_status_info['Rows'] = row[3]
table_status_info['Data_length'] = row[5]
table_status_info['Max_data_length'] = row[7]
table_status_info['Create_time'] = row[10]
table_status_info['Update_time'] = row[11]
return table_status_info
def serialize_via_marshal(obj):
"""Serialize Python object via marshal into a compressed string."""
return compress(marshal.dumps(obj))
def deserialize_via_marshal(astring):
"""Decompress and deserialize string into a Python object via marshal."""
return marshal.loads(decompress(astring))
def wash_table_column_name(colname):
"""
Evaluate table-column name to see if it is clean.
This function accepts only names containing [a-zA-Z0-9_].
@param colname: The string to be checked
@type colname: str
@return: colname if test passed
@rtype: str
@raise Exception: Raises an exception if colname is invalid.
"""
if re.search('[^\w]', colname):
raise Exception('The table column %s is not valid.' % repr(colname))
return colname
def real_escape_string(unescaped_string, run_on_slave=False):
"""
Escapes special characters in the unescaped string for use in a DB query.
@param unescaped_string: The string to be escaped
@type unescaped_string: str
@return: Returns the escaped string
@rtype: str
"""
dbhost = CFG_DATABASE_HOST
if run_on_slave and CFG_DATABASE_SLAVE:
dbhost = CFG_DATABASE_SLAVE
connection_object = _db_login(dbhost)
escaped_string = connection_object.escape_string(unescaped_string)
return escaped_string

Event Timeline

c4science · Help