mirror of
https://github.com/moparisthebest/SickRage
synced 2024-11-19 07:45:06 -05:00
590 lines
25 KiB
Python
590 lines
25 KiB
Python
# informix/base.py
|
|
# Copyright (C) 2005-2014 the SQLAlchemy authors and contributors <see AUTHORS file>
|
|
# coding: gbk
|
|
#
|
|
# This module is part of SQLAlchemy and is released under
|
|
# the MIT License: http://www.opensource.org/licenses/mit-license.php
|
|
|
|
"""
|
|
.. dialect:: informix
|
|
:name: Informix
|
|
|
|
.. note::
|
|
|
|
The Informix dialect functions on current SQLAlchemy versions
|
|
but is not regularly tested, and may have many issues and
|
|
caveats not currently handled.
|
|
|
|
"""
|
|
|
|
|
|
import datetime
|
|
|
|
from sqlalchemy import sql, schema, exc, pool, util
|
|
from sqlalchemy.sql import compiler, text
|
|
from sqlalchemy.engine import default, reflection
|
|
from sqlalchemy import types as sqltypes
|
|
|
|
RESERVED_WORDS = set(
|
|
["abs", "absolute", "access", "access_method", "acos", "active", "add",
|
|
"address", "add_months", "admin", "after", "aggregate", "alignment",
|
|
"all", "allocate", "all_rows", "alter", "and", "ansi", "any", "append",
|
|
"array", "as", "asc", "ascii", "asin", "at", "atan", "atan2", "attach",
|
|
"attributes", "audit", "authentication", "authid", "authorization",
|
|
"authorized", "auto", "autofree", "auto_reprepare", "auto_stat_mode",
|
|
"avg", "avoid_execute", "avoid_fact", "avoid_full", "avoid_hash",
|
|
"avoid_index", "avoid_index_sj", "avoid_multi_index", "avoid_nl",
|
|
"avoid_star_join", "avoid_subqf", "based", "before", "begin",
|
|
"between", "bigint", "bigserial", "binary", "bitand", "bitandnot",
|
|
"bitnot", "bitor", "bitxor", "blob", "blobdir", "boolean", "both",
|
|
"bound_impl_pdq", "buffered", "builtin", "by", "byte", "cache", "call",
|
|
"cannothash", "cardinality", "cascade", "case", "cast", "ceil", "char",
|
|
"character", "character_length", "char_length", "check", "class",
|
|
"class_origin", "client", "clob", "clobdir", "close", "cluster",
|
|
"clustersize", "cobol", "codeset", "collation", "collection",
|
|
"column", "columns", "commit", "committed", "commutator", "component",
|
|
"components", "concat", "concurrent", "connect", "connection",
|
|
"connection_name", "connect_by_iscycle", "connect_by_isleaf",
|
|
"connect_by_rootconst", "constraint", "constraints", "constructor",
|
|
"context", "continue", "copy", "cos", "costfunc", "count", "crcols",
|
|
"create", "cross", "current", "current_role", "currval", "cursor",
|
|
"cycle", "database", "datafiles", "dataskip", "date", "datetime",
|
|
"day", "dba", "dbdate", "dbinfo", "dbpassword", "dbsecadm",
|
|
"dbservername", "deallocate", "debug", "debugmode", "debug_env", "dec",
|
|
"decimal", "declare", "decode", "decrypt_binary", "decrypt_char",
|
|
"dec_t", "default", "default_role", "deferred", "deferred_prepare",
|
|
"define", "delay", "delete", "deleting", "delimited", "delimiter",
|
|
"deluxe", "desc", "describe", "descriptor", "detach", "diagnostics",
|
|
"directives", "dirty", "disable", "disabled", "disconnect", "disk",
|
|
"distinct", "distributebinary", "distributesreferences",
|
|
"distributions", "document", "domain", "donotdistribute", "dormant",
|
|
"double", "drop", "dtime_t", "each", "elif", "else", "enabled",
|
|
"encryption", "encrypt_aes", "encrypt_tdes", "end", "enum",
|
|
"environment", "error", "escape", "exception", "exclusive", "exec",
|
|
"execute", "executeanywhere", "exemption", "exists", "exit", "exp",
|
|
"explain", "explicit", "express", "expression", "extdirectives",
|
|
"extend", "extent", "external", "fact", "false", "far", "fetch",
|
|
"file", "filetoblob", "filetoclob", "fillfactor", "filtering", "first",
|
|
"first_rows", "fixchar", "fixed", "float", "floor", "flush", "for",
|
|
"force", "forced", "force_ddl_exec", "foreach", "foreign", "format",
|
|
"format_units", "fortran", "found", "fraction", "fragment",
|
|
"fragments", "free", "from", "full", "function", "general", "get",
|
|
"gethint", "global", "go", "goto", "grant", "greaterthan",
|
|
"greaterthanorequal", "group", "handlesnulls", "hash", "having", "hdr",
|
|
"hex", "high", "hint", "hold", "home", "hour", "idslbacreadarray",
|
|
"idslbacreadset", "idslbacreadtree", "idslbacrules",
|
|
"idslbacwritearray", "idslbacwriteset", "idslbacwritetree",
|
|
"idssecuritylabel", "if", "ifx_auto_reprepare", "ifx_batchedread_table",
|
|
"ifx_int8_t", "ifx_lo_create_spec_t", "ifx_lo_stat_t", "immediate",
|
|
"implicit", "implicit_pdq", "in", "inactive", "increment", "index",
|
|
"indexes", "index_all", "index_sj", "indicator", "informix", "init",
|
|
"initcap", "inline", "inner", "inout", "insert", "inserting", "instead",
|
|
"int", "int8", "integ", "integer", "internal", "internallength",
|
|
"interval", "into", "intrvl_t", "is", "iscanonical", "isolation",
|
|
"item", "iterator", "java", "join", "keep", "key", "label", "labeleq",
|
|
"labelge", "labelglb", "labelgt", "labelle", "labellt", "labellub",
|
|
"labeltostring", "language", "last", "last_day", "leading", "left",
|
|
"length", "lessthan", "lessthanorequal", "let", "level", "like",
|
|
"limit", "list", "listing", "load", "local", "locator", "lock", "locks",
|
|
"locopy", "loc_t", "log", "log10", "logn", "long", "loop", "lotofile",
|
|
"low", "lower", "lpad", "ltrim", "lvarchar", "matched", "matches",
|
|
"max", "maxerrors", "maxlen", "maxvalue", "mdy", "median", "medium",
|
|
"memory", "memory_resident", "merge", "message_length", "message_text",
|
|
"middle", "min", "minute", "minvalue", "mod", "mode", "moderate",
|
|
"modify", "module", "money", "month", "months_between", "mounting",
|
|
"multiset", "multi_index", "name", "nchar", "negator", "new", "next",
|
|
"nextval", "next_day", "no", "nocache", "nocycle", "nomaxvalue",
|
|
"nomigrate", "nominvalue", "none", "non_dim", "non_resident", "noorder",
|
|
"normal", "not", "notemplatearg", "notequal", "null", "nullif",
|
|
"numeric", "numrows", "numtodsinterval", "numtoyminterval", "nvarchar",
|
|
"nvl", "octet_length", "of", "off", "old", "on", "online", "only",
|
|
"opaque", "opclass", "open", "optcompind", "optical", "optimization",
|
|
"option", "or", "order", "ordered", "out", "outer", "output",
|
|
"override", "page", "parallelizable", "parameter", "partition",
|
|
"pascal", "passedbyvalue", "password", "pdqpriority", "percaltl_cos",
|
|
"pipe", "pli", "pload", "policy", "pow", "power", "precision",
|
|
"prepare", "previous", "primary", "prior", "private", "privileges",
|
|
"procedure", "properties", "public", "put", "raise", "range", "raw",
|
|
"read", "real", "recordend", "references", "referencing", "register",
|
|
"rejectfile", "relative", "release", "remainder", "rename",
|
|
"reoptimization", "repeatable", "replace", "replication", "reserve",
|
|
"resolution", "resource", "restart", "restrict", "resume", "retain",
|
|
"retainupdatelocks", "return", "returned_sqlstate", "returning",
|
|
"returns", "reuse", "revoke", "right", "robin", "role", "rollback",
|
|
"rollforward", "root", "round", "routine", "row", "rowid", "rowids",
|
|
"rows", "row_count", "rpad", "rtrim", "rule", "sameas", "samples",
|
|
"sampling", "save", "savepoint", "schema", "scroll", "seclabel_by_comp",
|
|
"seclabel_by_name", "seclabel_to_char", "second", "secondary",
|
|
"section", "secured", "security", "selconst", "select", "selecting",
|
|
"selfunc", "selfuncargs", "sequence", "serial", "serial8",
|
|
"serializable", "serveruuid", "server_name", "session", "set",
|
|
"setsessionauth", "share", "short", "siblings", "signed", "sin",
|
|
"sitename", "size", "skall", "skinhibit", "skip", "skshow",
|
|
"smallfloat", "smallint", "some", "specific", "sql", "sqlcode",
|
|
"sqlcontext", "sqlerror", "sqlstate", "sqlwarning", "sqrt",
|
|
"stability", "stack", "standard", "start", "star_join", "statchange",
|
|
"statement", "static", "statistics", "statlevel", "status", "stdev",
|
|
"step", "stop", "storage", "store", "strategies", "string",
|
|
"stringtolabel", "struct", "style", "subclass_origin", "substr",
|
|
"substring", "sum", "support", "sync", "synonym", "sysdate",
|
|
"sysdbclose", "sysdbopen", "system", "sys_connect_by_path", "table",
|
|
"tables", "tan", "task", "temp", "template", "test", "text", "then",
|
|
"time", "timeout", "to", "today", "to_char", "to_date",
|
|
"to_dsinterval", "to_number", "to_yminterval", "trace", "trailing",
|
|
"transaction", "transition", "tree", "trigger", "triggers", "trim",
|
|
"true", "trunc", "truncate", "trusted", "type", "typedef", "typeid",
|
|
"typename", "typeof", "uid", "uncommitted", "under", "union",
|
|
"unique", "units", "unknown", "unload", "unlock", "unsigned",
|
|
"update", "updating", "upon", "upper", "usage", "use",
|
|
"uselastcommitted", "user", "use_hash", "use_nl", "use_subqf",
|
|
"using", "value", "values", "var", "varchar", "variable", "variance",
|
|
"variant", "varying", "vercols", "view", "violations", "void",
|
|
"volatile", "wait", "warning", "weekday", "when", "whenever", "where",
|
|
"while", "with", "without", "work", "write", "writedown", "writeup",
|
|
"xadatasource", "xid", "xload", "xunload", "year"
|
|
])
|
|
|
|
|
|
class InfoDateTime(sqltypes.DateTime):
|
|
|
|
def bind_processor(self, dialect):
|
|
def process(value):
|
|
if value is not None:
|
|
if value.microsecond:
|
|
value = value.replace(microsecond=0)
|
|
return value
|
|
return process
|
|
|
|
|
|
class InfoTime(sqltypes.Time):
|
|
|
|
def bind_processor(self, dialect):
|
|
def process(value):
|
|
if value is not None:
|
|
if value.microsecond:
|
|
value = value.replace(microsecond=0)
|
|
return value
|
|
return process
|
|
|
|
def result_processor(self, dialect, coltype):
|
|
def process(value):
|
|
if isinstance(value, datetime.datetime):
|
|
return value.time()
|
|
else:
|
|
return value
|
|
return process
|
|
|
|
colspecs = {
|
|
sqltypes.DateTime: InfoDateTime,
|
|
sqltypes.TIMESTAMP: InfoDateTime,
|
|
sqltypes.Time: InfoTime,
|
|
}
|
|
|
|
|
|
ischema_names = {
|
|
0: sqltypes.CHAR, # CHAR
|
|
1: sqltypes.SMALLINT, # SMALLINT
|
|
2: sqltypes.INTEGER, # INT
|
|
3: sqltypes.FLOAT, # Float
|
|
3: sqltypes.Float, # SmallFloat
|
|
5: sqltypes.DECIMAL, # DECIMAL
|
|
6: sqltypes.Integer, # Serial
|
|
7: sqltypes.DATE, # DATE
|
|
8: sqltypes.Numeric, # MONEY
|
|
10: sqltypes.DATETIME, # DATETIME
|
|
11: sqltypes.LargeBinary, # BYTE
|
|
12: sqltypes.TEXT, # TEXT
|
|
13: sqltypes.VARCHAR, # VARCHAR
|
|
15: sqltypes.NCHAR, # NCHAR
|
|
16: sqltypes.NVARCHAR, # NVARCHAR
|
|
17: sqltypes.Integer, # INT8
|
|
18: sqltypes.Integer, # Serial8
|
|
43: sqltypes.String, # LVARCHAR
|
|
-1: sqltypes.BLOB, # BLOB
|
|
-1: sqltypes.CLOB, # CLOB
|
|
}
|
|
|
|
|
|
class InfoTypeCompiler(compiler.GenericTypeCompiler):
|
|
def visit_DATETIME(self, type_):
|
|
return "DATETIME YEAR TO SECOND"
|
|
|
|
def visit_TIME(self, type_):
|
|
return "DATETIME HOUR TO SECOND"
|
|
|
|
def visit_TIMESTAMP(self, type_):
|
|
return "DATETIME YEAR TO SECOND"
|
|
|
|
def visit_large_binary(self, type_):
|
|
return "BYTE"
|
|
|
|
def visit_boolean(self, type_):
|
|
return "SMALLINT"
|
|
|
|
|
|
class InfoSQLCompiler(compiler.SQLCompiler):
|
|
|
|
def default_from(self):
|
|
return " from systables where tabname = 'systables' "
|
|
|
|
def get_select_precolumns(self, select):
|
|
s = ""
|
|
if select._offset:
|
|
s += "SKIP %s " % select._offset
|
|
if select._limit:
|
|
s += "FIRST %s " % select._limit
|
|
s += select._distinct and "DISTINCT " or ""
|
|
return s
|
|
|
|
def visit_select(self, select, asfrom=False, parens=True, **kw):
|
|
text = compiler.SQLCompiler.visit_select(self, select, asfrom, parens, **kw)
|
|
if asfrom and parens and self.dialect.server_version_info < (11,):
|
|
#assuming that 11 version doesn't need this, not tested
|
|
return "table(multiset" + text + ")"
|
|
else:
|
|
return text
|
|
|
|
def limit_clause(self, select):
|
|
return ""
|
|
|
|
def visit_function(self, func, **kw):
|
|
if func.name.lower() == 'current_date':
|
|
return "today"
|
|
elif func.name.lower() == 'current_time':
|
|
return "CURRENT HOUR TO SECOND"
|
|
elif func.name.lower() in ('current_timestamp', 'now'):
|
|
return "CURRENT YEAR TO SECOND"
|
|
else:
|
|
return compiler.SQLCompiler.visit_function(self, func, **kw)
|
|
|
|
def visit_mod_binary(self, binary, operator, **kw):
|
|
return "MOD(%s, %s)" % (self.process(binary.left, **kw),
|
|
self.process(binary.right, **kw))
|
|
|
|
|
|
class InfoDDLCompiler(compiler.DDLCompiler):
|
|
|
|
def visit_add_constraint(self, create):
|
|
preparer = self.preparer
|
|
return "ALTER TABLE %s ADD CONSTRAINT %s" % (
|
|
self.preparer.format_table(create.element.table),
|
|
self.process(create.element)
|
|
)
|
|
|
|
def get_column_specification(self, column, **kw):
|
|
colspec = self.preparer.format_column(column)
|
|
first = None
|
|
if column.primary_key and column.autoincrement:
|
|
try:
|
|
first = [c for c in column.table.primary_key.columns
|
|
if (c.autoincrement and
|
|
isinstance(c.type, sqltypes.Integer) and
|
|
not c.foreign_keys)].pop(0)
|
|
except IndexError:
|
|
pass
|
|
|
|
if column is first:
|
|
colspec += " SERIAL"
|
|
else:
|
|
colspec += " " + self.dialect.type_compiler.process(column.type)
|
|
default = self.get_column_default_string(column)
|
|
if default is not None:
|
|
colspec += " DEFAULT " + default
|
|
|
|
if not column.nullable:
|
|
colspec += " NOT NULL"
|
|
|
|
return colspec
|
|
|
|
def get_column_default_string(self, column):
|
|
if (isinstance(column.server_default, schema.DefaultClause) and
|
|
isinstance(column.server_default.arg, basestring)):
|
|
if isinstance(column.type, (sqltypes.Integer, sqltypes.Numeric)):
|
|
return self.sql_compiler.process(text(column.server_default.arg))
|
|
|
|
return super(InfoDDLCompiler, self).get_column_default_string(column)
|
|
|
|
### Informix wants the constraint name at the end, hence this ist c&p from sql/compiler.py
|
|
def visit_primary_key_constraint(self, constraint):
|
|
if len(constraint) == 0:
|
|
return ''
|
|
text = "PRIMARY KEY "
|
|
text += "(%s)" % ', '.join(self.preparer.quote(c.name, c.quote)
|
|
for c in constraint)
|
|
text += self.define_constraint_deferrability(constraint)
|
|
|
|
if constraint.name is not None:
|
|
text += " CONSTRAINT %s" % self.preparer.format_constraint(constraint)
|
|
return text
|
|
|
|
def visit_foreign_key_constraint(self, constraint):
|
|
preparer = self.dialect.identifier_preparer
|
|
remote_table = list(constraint._elements.values())[0].column.table
|
|
text = "FOREIGN KEY (%s) REFERENCES %s (%s)" % (
|
|
', '.join(preparer.quote(f.parent.name, f.parent.quote)
|
|
for f in constraint._elements.values()),
|
|
preparer.format_table(remote_table),
|
|
', '.join(preparer.quote(f.column.name, f.column.quote)
|
|
for f in constraint._elements.values())
|
|
)
|
|
text += self.define_constraint_cascades(constraint)
|
|
text += self.define_constraint_deferrability(constraint)
|
|
|
|
if constraint.name is not None:
|
|
text += " CONSTRAINT %s " % \
|
|
preparer.format_constraint(constraint)
|
|
return text
|
|
|
|
def visit_unique_constraint(self, constraint):
|
|
text = "UNIQUE (%s)" % (', '.join(self.preparer.quote(c.name, c.quote) for c in constraint))
|
|
text += self.define_constraint_deferrability(constraint)
|
|
|
|
if constraint.name is not None:
|
|
text += "CONSTRAINT %s " % self.preparer.format_constraint(constraint)
|
|
return text
|
|
|
|
|
|
class InformixIdentifierPreparer(compiler.IdentifierPreparer):
|
|
|
|
reserved_words = RESERVED_WORDS
|
|
|
|
|
|
class InformixDialect(default.DefaultDialect):
|
|
name = 'informix'
|
|
|
|
max_identifier_length = 128 # adjusts at runtime based on server version
|
|
|
|
type_compiler = InfoTypeCompiler
|
|
statement_compiler = InfoSQLCompiler
|
|
ddl_compiler = InfoDDLCompiler
|
|
colspecs = colspecs
|
|
ischema_names = ischema_names
|
|
preparer = InformixIdentifierPreparer
|
|
default_paramstyle = 'qmark'
|
|
|
|
def initialize(self, connection):
|
|
super(InformixDialect, self).initialize(connection)
|
|
|
|
# http://www.querix.com/support/knowledge-base/error_number_message/error_200
|
|
if self.server_version_info < (9, 2):
|
|
self.max_identifier_length = 18
|
|
else:
|
|
self.max_identifier_length = 128
|
|
|
|
def _get_table_names(self, connection, schema, type, **kw):
|
|
schema = schema or self.default_schema_name
|
|
s = "select tabname, owner from systables where owner=? and tabtype=?"
|
|
return [row[0] for row in connection.execute(s, schema, type)]
|
|
|
|
@reflection.cache
|
|
def get_table_names(self, connection, schema=None, **kw):
|
|
return self._get_table_names(connection, schema, 'T', **kw)
|
|
|
|
@reflection.cache
|
|
def get_view_names(self, connection, schema=None, **kw):
|
|
return self._get_table_names(connection, schema, 'V', **kw)
|
|
|
|
@reflection.cache
|
|
def get_schema_names(self, connection, **kw):
|
|
s = "select owner from systables"
|
|
return [row[0] for row in connection.execute(s)]
|
|
|
|
def has_table(self, connection, table_name, schema=None):
|
|
schema = schema or self.default_schema_name
|
|
cursor = connection.execute(
|
|
"""select tabname from systables where tabname=? and owner=?""",
|
|
table_name, schema)
|
|
return cursor.first() is not None
|
|
|
|
@reflection.cache
|
|
def get_columns(self, connection, table_name, schema=None, **kw):
|
|
schema = schema or self.default_schema_name
|
|
c = connection.execute(
|
|
"""select colname, coltype, collength, t3.default, t1.colno from
|
|
syscolumns as t1 , systables as t2 , OUTER sysdefaults as t3
|
|
where t1.tabid = t2.tabid and t2.tabname=? and t2.owner=?
|
|
and t3.tabid = t2.tabid and t3.colno = t1.colno
|
|
order by t1.colno""", table_name, schema)
|
|
|
|
pk_constraint = self.get_pk_constraint(connection, table_name, schema, **kw)
|
|
primary_cols = pk_constraint['constrained_columns']
|
|
|
|
columns = []
|
|
rows = c.fetchall()
|
|
for name, colattr, collength, default, colno in rows:
|
|
name = name.lower()
|
|
|
|
autoincrement = False
|
|
primary_key = False
|
|
|
|
if name in primary_cols:
|
|
primary_key = True
|
|
|
|
# in 7.31, coltype = 0x000
|
|
# ^^-- column type
|
|
# ^-- 1 not null, 0 null
|
|
not_nullable, coltype = divmod(colattr, 256)
|
|
if coltype not in (0, 13) and default:
|
|
default = default.split()[-1]
|
|
|
|
if coltype == 6: # Serial, mark as autoincrement
|
|
autoincrement = True
|
|
|
|
if coltype == 0 or coltype == 13: # char, varchar
|
|
coltype = ischema_names[coltype](collength)
|
|
if default:
|
|
default = "'%s'" % default
|
|
elif coltype == 5: # decimal
|
|
precision, scale = (collength & 0xFF00) >> 8, collength & 0xFF
|
|
if scale == 255:
|
|
scale = 0
|
|
coltype = sqltypes.Numeric(precision, scale)
|
|
else:
|
|
try:
|
|
coltype = ischema_names[coltype]
|
|
except KeyError:
|
|
util.warn("Did not recognize type '%s' of column '%s'" %
|
|
(coltype, name))
|
|
coltype = sqltypes.NULLTYPE
|
|
|
|
column_info = dict(name=name, type=coltype, nullable=not not_nullable,
|
|
default=default, autoincrement=autoincrement,
|
|
primary_key=primary_key)
|
|
columns.append(column_info)
|
|
return columns
|
|
|
|
@reflection.cache
|
|
def get_foreign_keys(self, connection, table_name, schema=None, **kw):
|
|
schema_sel = schema or self.default_schema_name
|
|
c = connection.execute(
|
|
"""select t1.constrname as cons_name,
|
|
t4.colname as local_column, t7.tabname as remote_table,
|
|
t6.colname as remote_column, t7.owner as remote_owner
|
|
from sysconstraints as t1 , systables as t2 ,
|
|
sysindexes as t3 , syscolumns as t4 ,
|
|
sysreferences as t5 , syscolumns as t6 , systables as t7 ,
|
|
sysconstraints as t8 , sysindexes as t9
|
|
where t1.tabid = t2.tabid and t2.tabname=? and t2.owner=? and t1.constrtype = 'R'
|
|
and t3.tabid = t2.tabid and t3.idxname = t1.idxname
|
|
and t4.tabid = t2.tabid and t4.colno in (t3.part1, t3.part2, t3.part3,
|
|
t3.part4, t3.part5, t3.part6, t3.part7, t3.part8, t3.part9, t3.part10,
|
|
t3.part11, t3.part11, t3.part12, t3.part13, t3.part4, t3.part15, t3.part16)
|
|
and t5.constrid = t1.constrid and t8.constrid = t5.primary
|
|
and t6.tabid = t5.ptabid and t6.colno in (t9.part1, t9.part2, t9.part3,
|
|
t9.part4, t9.part5, t9.part6, t9.part7, t9.part8, t9.part9, t9.part10,
|
|
t9.part11, t9.part11, t9.part12, t9.part13, t9.part4, t9.part15, t9.part16) and t9.idxname =
|
|
t8.idxname
|
|
and t7.tabid = t5.ptabid""", table_name, schema_sel)
|
|
|
|
def fkey_rec():
|
|
return {
|
|
'name': None,
|
|
'constrained_columns': [],
|
|
'referred_schema': None,
|
|
'referred_table': None,
|
|
'referred_columns': []
|
|
}
|
|
|
|
fkeys = util.defaultdict(fkey_rec)
|
|
|
|
rows = c.fetchall()
|
|
for cons_name, local_column, \
|
|
remote_table, remote_column, remote_owner in rows:
|
|
|
|
rec = fkeys[cons_name]
|
|
rec['name'] = cons_name
|
|
local_cols, remote_cols = \
|
|
rec['constrained_columns'], rec['referred_columns']
|
|
|
|
if not rec['referred_table']:
|
|
rec['referred_table'] = remote_table
|
|
if schema is not None:
|
|
rec['referred_schema'] = remote_owner
|
|
|
|
if local_column not in local_cols:
|
|
local_cols.append(local_column)
|
|
if remote_column not in remote_cols:
|
|
remote_cols.append(remote_column)
|
|
|
|
return fkeys.values()
|
|
|
|
@reflection.cache
|
|
def get_pk_constraint(self, connection, table_name, schema=None, **kw):
|
|
schema = schema or self.default_schema_name
|
|
|
|
# Select the column positions from sysindexes for sysconstraints
|
|
data = connection.execute(
|
|
"""select t2.*
|
|
from systables as t1, sysindexes as t2, sysconstraints as t3
|
|
where t1.tabid=t2.tabid and t1.tabname=? and t1.owner=?
|
|
and t2.idxname=t3.idxname and t3.constrtype='P'""",
|
|
table_name, schema
|
|
).fetchall()
|
|
|
|
colpositions = set()
|
|
|
|
for row in data:
|
|
colpos = set([getattr(row, 'part%d' % x) for x in range(1, 16)])
|
|
colpositions |= colpos
|
|
|
|
if not len(colpositions):
|
|
return {'constrained_columns': [], 'name': None}
|
|
|
|
# Select the column names using the columnpositions
|
|
# TODO: Maybe cache a bit of those col infos (eg select all colnames for one table)
|
|
place_holder = ','.join('?' * len(colpositions))
|
|
c = connection.execute(
|
|
"""select t1.colname
|
|
from syscolumns as t1, systables as t2
|
|
where t2.tabname=? and t1.tabid = t2.tabid and
|
|
t1.colno in (%s)""" % place_holder,
|
|
table_name, *colpositions
|
|
).fetchall()
|
|
|
|
cols = reduce(lambda x, y: list(x) + list(y), c, [])
|
|
return {'constrained_columns': cols, 'name': None}
|
|
|
|
@reflection.cache
|
|
def get_indexes(self, connection, table_name, schema, **kw):
|
|
# TODO: schema...
|
|
c = connection.execute(
|
|
"""select t1.*
|
|
from sysindexes as t1 , systables as t2
|
|
where t1.tabid = t2.tabid and t2.tabname=?""",
|
|
table_name)
|
|
|
|
indexes = []
|
|
for row in c.fetchall():
|
|
colnames = [getattr(row, 'part%d' % x) for x in range(1, 16)]
|
|
colnames = [x for x in colnames if x]
|
|
place_holder = ','.join('?' * len(colnames))
|
|
c = connection.execute(
|
|
"""select t1.colname
|
|
from syscolumns as t1, systables as t2
|
|
where t2.tabname=? and t1.tabid = t2.tabid and
|
|
t1.colno in (%s)""" % place_holder,
|
|
table_name, *colnames
|
|
).fetchall()
|
|
c = reduce(lambda x, y: list(x) + list(y), c, [])
|
|
indexes.append({
|
|
'name': row.idxname,
|
|
'unique': row.idxtype.lower() == 'u',
|
|
'column_names': c
|
|
})
|
|
return indexes
|
|
|
|
@reflection.cache
|
|
def get_view_definition(self, connection, view_name, schema=None, **kw):
|
|
schema = schema or self.default_schema_name
|
|
c = connection.execute(
|
|
"""select t1.viewtext
|
|
from sysviews as t1 , systables as t2
|
|
where t1.tabid=t2.tabid and t2.tabname=?
|
|
and t2.owner=? order by seqno""",
|
|
view_name, schema).fetchall()
|
|
|
|
return ''.join([row[0] for row in c])
|
|
|
|
def _get_default_schema_name(self, connection):
|
|
return connection.execute('select CURRENT_ROLE from systables').scalar()
|