在SqlAlchemy中反映数据库时缺少外键关系

Question

我正在尝试使用SqlAlchemy（0.5.8）以声明方式与旧数据库进行接口并使用反射。我的测试代码如下所示：

from sqlalchemy import * 
from sqlalchemy.orm import create_session 
from sqlalchemy.ext.declarative import declarative_base 

Base = declarative_base() 
engine = create_engine('oracle://schemaname:pwd@SID') 
meta = MetaData(bind=engine) 

class CONSTRUCT(Base): 
    __table__ = Table('CONSTRUCT', meta, autoload=True) 

class EXPRESSION(Base): 
    __table__ = Table('EXPRESSION', meta, autoload=True) 

session = create_session(bind=engine) 

现在，当我尝试使用这两个表（由外键约束在底层Oracle架构中定义）之间的连接运行查询：

print session.query(EXPRESSION).join(PURIFICATION) 

...没有喜悦：

sqlalchemy.exc.ArgumentError: Can't find any foreign key relationships between 'EXPRESSION' and 'PURIFICATION' 

但是：

>>> EXPRESSION.epiconstruct_pkey.property.columns 
[Column(u'epiconstruct_pkey', OracleNumeric(precision=10, scale=2, asdecimal=True, 
length=None), ForeignKey(u'construct.pkey'), table=<EXPRESSION>, nullable=False)] 

>>> CONSTRUCT.pkey.property.columns 
[Column(u'pkey', OracleNumeric(precision=38, scale=0, asdecimal=True, length=None), 
table=<CONSTRUCT>, primary_key=True, nullable=False)] 

这清楚地表明反射拾取了外键。

我哪里错了？

Answer 1

用Eclipse调试脚本+ SqlAlchemy代码后，我发现表/列的列表内部保持小写。因此，EXPRESSION.foreignkey和expression.foreignkey之间绝无可能匹配。因此错误信息。

深挖入SQLAlchemy的文档（http://www.sqlalchemy.org/docs/reference/dialects/oracle.html#identifier-casing）我然后发现如下：

“在Oracle，使用大写文本数据字典代表所有不区分大小写的标识符名称的SQLAlchemy另一方面认为全低。案例标识符名称不区分大小写Oracle模式在模式级通信期间将所有不区分大小写的标识符转换为或来自这两种格式，例如表和索引的反映。在SQLAlchemy侧使用大写名称表示区分大小写的标识符，以及SQLAlchemy会引用名称 - 这将导致与从Oracle收到的数据字典数据不匹配，所以除非标识符名称已真正创建为区分大小写（即usi ng引用的名称），所有小写名称都应该用在SQLAlchemy方面。“

所以我的代码工作，如果它看起来像这样（的差别仅仅是个案的变化）：

from sqlalchemy import * 
from sqlalchemy.orm import create_session 
from sqlalchemy.ext.declarative import declarative_base 

Base = declarative_base() 
engine = create_engine('oracle://EPIGENETICS:sgc04lab@ELN') 
meta = MetaData(bind=engine) 

class construct(Base): 
    __table__ = Table('construct', meta, autoload=True) 

class expression(Base): 
    __table__ = Table('expression', meta, autoload=True) 

class purification(Base): 
    __table__ = Table('purification', meta, autoload=True) 

session = create_session(bind=engine) 
print session.query(expression).join(purification,expression) 

...它吐出：

SELECT expression.pkey AS expression_pkey, expression.cellline AS expression_cellline, expression.epiconstruct_pkey AS expression_epiconstruct_pkey, expression.elnexp AS expression_elnexp, expression.expression_id AS expression_expression_id, expression.expressioncomments AS expression_expressioncomments, expression.cellmass AS expression_cellmass, expression.datestamp AS expression_datestamp, expression.person AS expression_person, expression.soluble AS expression_soluble, expression.semet AS expression_semet, expression.scale AS expression_scale, expression.purtest AS expression_purtest, expression.nmrlabelled AS expression_nmrlabelled, expression.yield AS expression_yield 
FROM expression JOIN purification ON expression.pkey = purification.epiexpression_pkey JOIN expression ON expression.pkey = purification.epiexpression_pkey 

结案。