#-----------------------------------------------------------------------

#

# Copyright (C) 2000, 2001 by Autonomous Zone Industries

# Copyright (C) 2002 Gregory P. Smith

#

# License:      This is free software.  You may use this software for any

#               purpose including modification/redistribution, so long as

#               this header remains intact and that you do not claim any

#               rights of ownership or authorship of this software.  This

#               software has been tested, but no warranty is expressed or

#               implied.

#

#   --  Gregory P. Smith <greg@krypto.org>

# This provides a simple database table interface built on top of

# the Python Berkeley DB 3 interface.

#

_cvsid = '$Id: dbtables.py 66088 2008-08-31 14:00:51Z jesus.cea $'

import re

import sys

import copy

import random

import struct

import cPickle as pickle

try:

    # For Pythons w/distutils pybsddb

    from bsddb3 import db

except ImportError:

    # For Python 2.3

    from bsddb import db

# XXX(nnorwitz): is this correct? DBIncompleteError is conditional in _bsddb.c

if not hasattr(db,"DBIncompleteError") :

    class DBIncompleteError(Exception):

        pass

    db.DBIncompleteError = DBIncompleteError

class TableDBError(StandardError):

    pass

class TableAlreadyExists(TableDBError):

    pass

class Cond:

    """This condition matches everything"""

    def __call__(self, s):

        return 1

class ExactCond(Cond):

    """Acts as an exact match condition function"""

    def __init__(self, strtomatch):

        self.strtomatch = strtomatch

    def __call__(self, s):

        return s == self.strtomatch

class PrefixCond(Cond):

    """Acts as a condition function for matching a string prefix"""

    def __init__(self, prefix):

        self.prefix = prefix

    def __call__(self, s):

        return s[:len(self.prefix)] == self.prefix

class PostfixCond(Cond):

    """Acts as a condition function for matching a string postfix"""

    def __init__(self, postfix):

        self.postfix = postfix

    def __call__(self, s):

        return s[-len(self.postfix):] == self.postfix

class LikeCond(Cond):

    """

    Acts as a function that will match using an SQL 'LIKE' style

    string.  Case insensitive and % signs are wild cards.

    This isn't perfect but it should work for the simple common cases.

    """

    def __init__(self, likestr, re_flags=re.IGNORECASE):

        # escape python re characters

        chars_to_escape = '.*+()[]?'

        for char in chars_to_escape :

            likestr = likestr.replace(char, '\\'+char)

        # convert %s to wildcards

        self.likestr = likestr.replace('%', '.*')

        self.re = re.compile('^'+self.likestr+'$', re_flags)

    def __call__(self, s):

        return self.re.match(s)

#

# keys used to store database metadata

#

_table_names_key = '__TABLE_NAMES__'  # list of the tables in this db

_columns = '._COLUMNS__'  # table_name+this key contains a list of columns

def _columns_key(table):

    return table + _columns

#

# these keys are found within table sub databases

#

_data =  '._DATA_.'  # this+column+this+rowid key contains table data

_rowid = '._ROWID_.' # this+rowid+this key contains a unique entry for each

                     # row in the table.  (no data is stored)

_rowid_str_len = 8   # length in bytes of the unique rowid strings

def _data_key(table, col, rowid):

    return table + _data + col + _data + rowid

def _search_col_data_key(table, col):

    return table + _data + col + _data

def _search_all_data_key(table):

    return table + _data

def _rowid_key(table, rowid):

    return table + _rowid + rowid + _rowid

def _search_rowid_key(table):

    return table + _rowid

def contains_metastrings(s) :

    """Verify that the given string does not contain any

    metadata strings that might interfere with dbtables database operation.

    """

    if (s.find(_table_names_key) >= 0 or

        s.find(_columns) >= 0 or

        s.find(_data) >= 0 or

        s.find(_rowid) >= 0):

        # Then

        return 1

    else:

        return 0

class bsdTableDB :

    def __init__(self, filename, dbhome, create=0, truncate=0, mode=0600,

                 recover=0, dbflags=0):

        """bsdTableDB(filename, dbhome, create=0, truncate=0, mode=0600)

        Open database name in the dbhome Berkeley DB directory.

        Use keyword arguments when calling this constructor.

        """

        self.db = None

        myflags = db.DB_THREAD

        if create:

            myflags |= db.DB_CREATE

        flagsforenv = (db.DB_INIT_MPOOL | db.DB_INIT_LOCK | db.DB_INIT_LOG |

                       db.DB_INIT_TXN | dbflags)

        # DB_AUTO_COMMIT isn't a valid flag for env.open()

        try:

            dbflags |= db.DB_AUTO_COMMIT

        except AttributeError:

            pass

        if recover:

            flagsforenv = flagsforenv | db.DB_RECOVER

        self.env = db.DBEnv()

        # enable auto deadlock avoidance

        self.env.set_lk_detect(db.DB_LOCK_DEFAULT)

        self.env.open(dbhome, myflags | flagsforenv)

        if truncate:

            myflags |= db.DB_TRUNCATE

        self.db = db.DB(self.env)

        # this code relies on DBCursor.set* methods to raise exceptions

        # rather than returning None

        self.db.set_get_returns_none(1)

        # allow duplicate entries [warning: be careful w/ metadata]

        self.db.set_flags(db.DB_DUP)

        self.db.open(filename, db.DB_BTREE, dbflags | myflags, mode)

        self.dbfilename = filename

        if sys.version_info[0] >= 3 :

            class cursor_py3k(object) :

                def __init__(self, dbcursor) :

                    self._dbcursor = dbcursor

                def close(self) :

                    return self._dbcursor.close()

                def set_range(self, search) :

                    v = self._dbcursor.set_range(bytes(search, "iso8859-1"))

                    if v != None :

                        v = (v[0].decode("iso8859-1"),

                                v[1].decode("iso8859-1"))

                    return v

                def __next__(self) :

                    v = getattr(self._dbcursor, "next")()

                    if v != None :

                        v = (v[0].decode("iso8859-1"),

                                v[1].decode("iso8859-1"))

                    return v

            class db_py3k(object) :

                def __init__(self, db) :

                    self._db = db

                def cursor(self, txn=None) :

                    return cursor_py3k(self._db.cursor(txn=txn))

                def has_key(self, key, txn=None) :

                    return getattr(self._db,"has_key")(bytes(key, "iso8859-1"),

                            txn=txn)

                def put(self, key, value, flags=0, txn=None) :

                    key = bytes(key, "iso8859-1")

                    if value != None :

                        value = bytes(value, "iso8859-1")

                    return self._db.put(key, value, flags=flags, txn=txn)

                def put_bytes(self, key, value, txn=None) :

                    key = bytes(key, "iso8859-1")

                    return self._db.put(key, value, txn=txn)

                def get(self, key, txn=None, flags=0) :

                    key = bytes(key, "iso8859-1")

                    v = self._db.get(key, txn=txn, flags=flags)

                    if v != None :

                        v = v.decode("iso8859-1")

                    return v

                def get_bytes(self, key, txn=None, flags=0) :

                    key = bytes(key, "iso8859-1")

                    return self._db.get(key, txn=txn, flags=flags)

                def delete(self, key, txn=None) :

                    key = bytes(key, "iso8859-1")

                    return self._db.delete(key, txn=txn)

                def close (self) :

                    return self._db.close()

            self.db = db_py3k(self.db)

        else :  # Python 2.x

            pass

        # Initialize the table names list if this is a new database

        txn = self.env.txn_begin()

        try:

            if not getattr(self.db, "has_key")(_table_names_key, txn):

                getattr(self.db, "put_bytes", self.db.put) \

                        (_table_names_key, pickle.dumps([], 1), txn=txn)

        # Yes, bare except

        except:

            txn.abort()

            raise

        else:

            txn.commit()

        # TODO verify more of the database's metadata?

        self.__tablecolumns = {}

    def __del__(self):

        self.close()

    def close(self):

        if self.db is not None:

            self.db.close()

            self.db = None

        if self.env is not None:

            self.env.close()

            self.env = None

    def checkpoint(self, mins=0):

        try:

            self.env.txn_checkpoint(mins)

        except db.DBIncompleteError:

            pass

    def sync(self):

        try:

            self.db.sync()

        except db.DBIncompleteError:

            pass

    def _db_print(self) :

        """Print the database to stdout for debugging"""

        print "******** Printing raw database for debugging ********"

        cur = self.db.cursor()

        try:

            key, data = cur.first()

            while 1:

                print repr({key: data})

                next = cur.next()

                if next:

                    key, data = next

                else:

                    cur.close()

                    return

        except db.DBNotFoundError:

            cur.close()

    def CreateTable(self, table, columns):

        """CreateTable(table, columns) - Create a new table in the database.

        raises TableDBError if it already exists or for other DB errors.

        """

        assert isinstance(columns, list)

        txn = None

        try:

            # checking sanity of the table and column names here on

            # table creation will prevent problems elsewhere.

            if contains_metastrings(table):

                raise ValueError(

                    "bad table name: contains reserved metastrings")

            for column in columns :

                if contains_metastrings(column):

                    raise ValueError(

                        "bad column name: contains reserved metastrings")

            columnlist_key = _columns_key(table)

            if getattr(self.db, "has_key")(columnlist_key):

                raise TableAlreadyExists, "table already exists"

            txn = self.env.txn_begin()

            # store the table's column info

            getattr(self.db, "put_bytes", self.db.put)(columnlist_key,

                    pickle.dumps(columns, 1), txn=txn)

            # add the table name to the tablelist

            tablelist = pickle.loads(getattr(self.db, "get_bytes",

                self.db.get) (_table_names_key, txn=txn, flags=db.DB_RMW))

            tablelist.append(table)

            # delete 1st, in case we opened with DB_DUP

            self.db.delete(_table_names_key, txn=txn)

            getattr(self.db, "put_bytes", self.db.put)(_table_names_key,

                    pickle.dumps(tablelist, 1), txn=txn)

            txn.commit()

            txn = None

        except db.DBError, dberror:

            if txn:

                txn.abort()

            if sys.version_info[0] < 3 :

                raise TableDBError, dberror[1]

            else :

                raise TableDBError, dberror.args[1]

    def ListTableColumns(self, table):

        """Return a list of columns in the given table.

        [] if the table doesn't exist.

        """

        assert isinstance(table, str)

        if contains_metastrings(table):

            raise ValueError, "bad table name: contains reserved metastrings"

        columnlist_key = _columns_key(table)

        if not getattr(self.db, "has_key")(columnlist_key):

            return []

        pickledcolumnlist = getattr(self.db, "get_bytes",

                self.db.get)(columnlist_key)

        if pickledcolumnlist:

            return pickle.loads(pickledcolumnlist)

        else:

            return []

    def ListTables(self):

        """Return a list of tables in this database."""

        pickledtablelist = self.db.get_get(_table_names_key)

        if pickledtablelist:

            return pickle.loads(pickledtablelist)

        else:

            return []

    def CreateOrExtendTable(self, table, columns):

        """CreateOrExtendTable(table, columns)

        Create a new table in the database.

        If a table of this name already exists, extend it to have any

        additional columns present in the given list as well as

        all of its current columns.

        """

        assert isinstance(columns, list)

        try:

            self.CreateTable(table, columns)

        except TableAlreadyExists:

            # the table already existed, add any new columns

            txn = None

            try:

                columnlist_key = _columns_key(table)

                txn = self.env.txn_begin()

                # load the current column list

                oldcolumnlist = pickle.loads(

                    getattr(self.db, "get_bytes",

                        self.db.get)(columnlist_key, txn=txn, flags=db.DB_RMW))

                # create a hash table for fast lookups of column names in the

                # loop below

                oldcolumnhash = {}

                for c in oldcolumnlist:

                    oldcolumnhash[c] = c

                # create a new column list containing both the old and new

                # column names

                newcolumnlist = copy.copy(oldcolumnlist)

                for c in columns:

                    if not oldcolumnhash.has_key(c):

                        newcolumnlist.append(c)

                # store the table's new extended column list

                if newcolumnlist != oldcolumnlist :

                    # delete the old one first since we opened with DB_DUP

                    self.db.delete(columnlist_key, txn=txn)

                    getattr(self.db, "put_bytes", self.db.put)(columnlist_key,

                                pickle.dumps(newcolumnlist, 1),

                                txn=txn)

                txn.commit()

                txn = None

                self.__load_column_info(table)

            except db.DBError, dberror:

                if txn:

                    txn.abort()

                if sys.version_info[0] < 3 :

                    raise TableDBError, dberror[1]

                else :

                    raise TableDBError, dberror.args[1]

    def __load_column_info(self, table) :

        """initialize the self.__tablecolumns dict"""

        # check the column names

        try:

            tcolpickles = getattr(self.db, "get_bytes",

                    self.db.get)(_columns_key(table))

        except db.DBNotFoundError:

            raise TableDBError, "unknown table: %r" % (table,)

        if not tcolpickles:

            raise TableDBError, "unknown table: %r" % (table,)

        self.__tablecolumns[table] = pickle.loads(tcolpickles)

    def __new_rowid(self, table, txn) :

        """Create a new unique row identifier"""

        unique = 0

        while not unique:

            # Generate a random 64-bit row ID string

            # (note: might have <64 bits of true randomness

            # but it's plenty for our database id needs!)

            blist = []

            for x in xrange(_rowid_str_len):

                blist.append(random.randint(0,255))

            newid = struct.pack('B'*_rowid_str_len, *blist)

            if sys.version_info[0] >= 3 :

                newid = newid.decode("iso8859-1")  # 8 bits

            # Guarantee uniqueness by adding this key to the database

            try:

                self.db.put(_rowid_key(table, newid), None, txn=txn,

                            flags=db.DB_NOOVERWRITE)

            except db.DBKeyExistError:

                pass

            else:

                unique = 1

        return newid

    def Insert(self, table, rowdict) :

        """Insert(table, datadict) - Insert a new row into the table

        using the keys+values from rowdict as the column values.

        """

        txn = None

        try:

            if not getattr(self.db, "has_key")(_columns_key(table)):

                raise TableDBError, "unknown table"

            # check the validity of each column name

            if not self.__tablecolumns.has_key(table):

                self.__load_column_info(table)

            for column in rowdict.keys() :

                if not self.__tablecolumns[table].count(column):

                    raise TableDBError, "unknown column: %r" % (column,)

            # get a unique row identifier for this row

            txn = self.env.txn_begin()

            rowid = self.__new_rowid(table, txn=txn)

            # insert the row values into the table database

            for column, dataitem in rowdict.items():

                # store the value

                self.db.put(_data_key(table, column, rowid), dataitem, txn=txn)

            txn.commit()

            txn = None

        except db.DBError, dberror:

            # WIBNI we could just abort the txn and re-raise the exception?

            # But no, because TableDBError is not related to DBError via

            # inheritance, so it would be backwards incompatible.  Do the next

            # best thing.

            info = sys.exc_info()

            if txn:

                txn.abort()

                self.db.delete(_rowid_key(table, rowid))

            if sys.version_info[0] < 3 :

                raise TableDBError, dberror[1], info[2]

            else :

                raise TableDBError, dberror.args[1], info[2]

    def Modify(self, table, conditions={}, mappings={}):

        """Modify(table, conditions={}, mappings={}) - Modify items in rows matching 'conditions' using mapping functions in 'mappings'

        * table - the table name

        * conditions - a dictionary keyed on column names containing

          a condition callable expecting the data string as an

          argument and returning a boolean.

        * mappings - a dictionary keyed on column names containing a

          condition callable expecting the data string as an argument and

          returning the new string for that column.

        """

        try:

            matching_rowids = self.__Select(table, [], conditions)

            # modify only requested columns

            columns = mappings.keys()

            for rowid in matching_rowids.keys():

                txn = None

                try:

                    for column in columns: