How to print all columns in SQLAlchemy ORM
Question:
Using SQLAlchemy, I am trying to print out all of the attributes of each model that I have in a manner similar to:
SELECT * from table;
However, I would like to do something with each models instance information as I get it. So far the best that I’ve been able to come up with is:
for m in session.query(model).all():
print [getattr(m, x.__str__().split('.')[1]) for x in model.__table__.columns]
# additional code
And this will give me what I’m looking for, but it’s a fairly roundabout way of getting it. I was kind of hoping for an attribute along the lines of:
m.attributes
# or
m.columns.values
I feel I’m missing something and there is a much better way of doing this. I’m doing this because I’ll be printing everything to .CSV files, and I don’t want to have to specify the columns/attributes that I’m interested in, I want everything (there’s a lot of columns in a lot of models to be printed).
Answers:
Take a look at SQLAchemy’s metadata reflection feature.
A Table object can be instructed to load information about itself from the corresponding database schema object already existing within the database. This process is called reflection.
You may be interested in what I came up with to do this.
from sqlalchemy.orm import class_mapper
import collections
# structure returned by get_metadata function.
MetaDataTuple = collections.namedtuple("MetaDataTuple",
"coltype, colname, default, m2m, nullable, uselist, collection")
def get_metadata_iterator(class_):
for prop in class_mapper(class_).iterate_properties:
name = prop.key
if name.startswith("_") or name == "id" or name.endswith("_id"):
continue
md = _get_column_metadata(prop)
if md is None:
continue
yield md
def get_column_metadata(class_, colname):
prop = class_mapper(class_).get_property(colname)
md = _get_column_metadata(prop)
if md is None:
raise ValueError("Not a column name: %r." % (colname,))
return md
def _get_column_metadata(prop):
name = prop.key
m2m = False
default = None
nullable = None
uselist = False
collection = None
proptype = type(prop)
if proptype is ColumnProperty:
coltype = type(prop.columns[0].type).__name__
try:
default = prop.columns[0].default
except AttributeError:
default = None
else:
if default is not None:
default = default.arg(None)
nullable = prop.columns[0].nullable
elif proptype is RelationshipProperty:
coltype = RelationshipProperty.__name__
m2m = prop.secondary is not None
nullable = prop.local_side[0].nullable
uselist = prop.uselist
if prop.collection_class is not None:
collection = type(prop.collection_class()).__name__
else:
collection = "list"
else:
return None
return MetaDataTuple(coltype, str(name), default, m2m, nullable, uselist, collection)
I believe this is the easiest way:
print [cname for cname in m.__dict__.keys()]
EDIT: The answer above me using sqlalchemy.inspection.inspect() seems to be a better solution.
I use this because it’s slightly shorter:
for m in session.query(*model.__table__.columns).all():
print m
Building on Rodney L’s answer:
model = MYMODEL
columns = [m.key for m in model.__table__.columns]
This is an old post, but I ran into a problem with the actual database column names not matching the mapped attribute names on the instance. We ended up going with this:
from sqlalchemy import inspect
inst = inspect(model)
attr_names = [c_attr.key for c_attr in inst.mapper.column_attrs]
Hope that helps somebody with the same problem!
Put this together and found it helpful:
from sqlalchemy import create_engine
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy.orm import sessionmaker
engine = create_engine('mysql+pymysql://testuser:password@localhost:3306/testdb')
DeclarativeBase = declarative_base()
metadata = DeclarativeBase.metadata
metadata.bind = engine
# configure Session class with desired options
Session = sessionmaker()
# associate it with our custom Session class
Session.configure(bind=engine)
# work with the session
session = Session()
And then:
d = {k: metadata.tables[k].columns.keys() for k in metadata.tables.keys()}
Example output print(d)
:
{'orderdetails': ['orderNumber', 'productCode', 'quantityOrdered', 'priceEach', 'orderLineNumber'],
'offices': ['addressLine1', 'addressLine2', 'city', 'country', 'officeCode', 'phone', 'postalCode', 'state', 'territory'],
'orders': ['comments', 'customerNumber', 'orderDate', 'orderNumber', 'requiredDate', 'shippedDate', 'status'],
'products': ['MSRP', 'buyPrice', 'productCode', 'productDescription', 'productLine', 'productName', 'productScale', 'productVendor', 'quantityInStock'],
'employees': ['employeeNumber', 'lastName', 'firstName', 'extension', 'email', 'officeCode', 'reportsTo', 'jobTitle'],
'customers': ['addressLine1', 'addressLine2', 'city', 'contactFirstName', 'contactLastName', 'country', 'creditLimit', 'customerName', 'customerNumber', 'phone', 'postalCode', 'salesRepEmployeeNumber', 'state'],
'productlines': ['htmlDescription', 'image', 'productLine', 'textDescription'],
'payments': ['amount', 'checkNumber', 'customerNumber', 'paymentDate']}
OR and then:
from sqlalchemy.sql import text
cmd = "SELECT * FROM information_schema.columns WHERE table_schema = :db ORDER BY table_name,ordinal_position"
result = session.execute(
text(cmd),
{"db": "classicmodels"}
)
result.fetchall()
Probably the shortest solution (see the recent documentation):
from sqlalchemy.inspection import inspect
columns = [column.name for column in inspect(model).c]
The last line might look more readable, if rewrite it in three lines:
table = inspect(model)
for column in table.c:
print column.name
I’m using SQL Alchemy v 1.0.14 on Python 3.5.2
Assuming you can connect to an engine with create_engine(), I was able to display all columns using the following code. Replace “my connection string” and “my table name” with the appropriate values.
from sqlalchemy import create_engine, MetaData, Table, select
engine = create_engine('my connection string')
conn = engine.connect()
metadata = MetaData(conn)
t = Table("my table name", metadata, autoload=True)
columns = [m.key for m in t.columns]
columns
the last row just displays the column names from the previous statement.
print repr(model.__table__)
Or just the columns:
print str(list(model.__table__.columns))
Using SQLAlchemy, I am trying to print out all of the attributes of each model that I have in a manner similar to:
SELECT * from table;
However, I would like to do something with each models instance information as I get it. So far the best that I’ve been able to come up with is:
for m in session.query(model).all():
print [getattr(m, x.__str__().split('.')[1]) for x in model.__table__.columns]
# additional code
And this will give me what I’m looking for, but it’s a fairly roundabout way of getting it. I was kind of hoping for an attribute along the lines of:
m.attributes
# or
m.columns.values
I feel I’m missing something and there is a much better way of doing this. I’m doing this because I’ll be printing everything to .CSV files, and I don’t want to have to specify the columns/attributes that I’m interested in, I want everything (there’s a lot of columns in a lot of models to be printed).
Take a look at SQLAchemy’s metadata reflection feature.
A Table object can be instructed to load information about itself from the corresponding database schema object already existing within the database. This process is called reflection.
You may be interested in what I came up with to do this.
from sqlalchemy.orm import class_mapper
import collections
# structure returned by get_metadata function.
MetaDataTuple = collections.namedtuple("MetaDataTuple",
"coltype, colname, default, m2m, nullable, uselist, collection")
def get_metadata_iterator(class_):
for prop in class_mapper(class_).iterate_properties:
name = prop.key
if name.startswith("_") or name == "id" or name.endswith("_id"):
continue
md = _get_column_metadata(prop)
if md is None:
continue
yield md
def get_column_metadata(class_, colname):
prop = class_mapper(class_).get_property(colname)
md = _get_column_metadata(prop)
if md is None:
raise ValueError("Not a column name: %r." % (colname,))
return md
def _get_column_metadata(prop):
name = prop.key
m2m = False
default = None
nullable = None
uselist = False
collection = None
proptype = type(prop)
if proptype is ColumnProperty:
coltype = type(prop.columns[0].type).__name__
try:
default = prop.columns[0].default
except AttributeError:
default = None
else:
if default is not None:
default = default.arg(None)
nullable = prop.columns[0].nullable
elif proptype is RelationshipProperty:
coltype = RelationshipProperty.__name__
m2m = prop.secondary is not None
nullable = prop.local_side[0].nullable
uselist = prop.uselist
if prop.collection_class is not None:
collection = type(prop.collection_class()).__name__
else:
collection = "list"
else:
return None
return MetaDataTuple(coltype, str(name), default, m2m, nullable, uselist, collection)
I believe this is the easiest way:
print [cname for cname in m.__dict__.keys()]
EDIT: The answer above me using sqlalchemy.inspection.inspect() seems to be a better solution.
I use this because it’s slightly shorter:
for m in session.query(*model.__table__.columns).all():
print m
Building on Rodney L’s answer:
model = MYMODEL
columns = [m.key for m in model.__table__.columns]
This is an old post, but I ran into a problem with the actual database column names not matching the mapped attribute names on the instance. We ended up going with this:
from sqlalchemy import inspect
inst = inspect(model)
attr_names = [c_attr.key for c_attr in inst.mapper.column_attrs]
Hope that helps somebody with the same problem!
Put this together and found it helpful:
from sqlalchemy import create_engine
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy.orm import sessionmaker
engine = create_engine('mysql+pymysql://testuser:password@localhost:3306/testdb')
DeclarativeBase = declarative_base()
metadata = DeclarativeBase.metadata
metadata.bind = engine
# configure Session class with desired options
Session = sessionmaker()
# associate it with our custom Session class
Session.configure(bind=engine)
# work with the session
session = Session()
And then:
d = {k: metadata.tables[k].columns.keys() for k in metadata.tables.keys()}
Example output print(d)
:
{'orderdetails': ['orderNumber', 'productCode', 'quantityOrdered', 'priceEach', 'orderLineNumber'],
'offices': ['addressLine1', 'addressLine2', 'city', 'country', 'officeCode', 'phone', 'postalCode', 'state', 'territory'],
'orders': ['comments', 'customerNumber', 'orderDate', 'orderNumber', 'requiredDate', 'shippedDate', 'status'],
'products': ['MSRP', 'buyPrice', 'productCode', 'productDescription', 'productLine', 'productName', 'productScale', 'productVendor', 'quantityInStock'],
'employees': ['employeeNumber', 'lastName', 'firstName', 'extension', 'email', 'officeCode', 'reportsTo', 'jobTitle'],
'customers': ['addressLine1', 'addressLine2', 'city', 'contactFirstName', 'contactLastName', 'country', 'creditLimit', 'customerName', 'customerNumber', 'phone', 'postalCode', 'salesRepEmployeeNumber', 'state'],
'productlines': ['htmlDescription', 'image', 'productLine', 'textDescription'],
'payments': ['amount', 'checkNumber', 'customerNumber', 'paymentDate']}
OR and then:
from sqlalchemy.sql import text
cmd = "SELECT * FROM information_schema.columns WHERE table_schema = :db ORDER BY table_name,ordinal_position"
result = session.execute(
text(cmd),
{"db": "classicmodels"}
)
result.fetchall()
Probably the shortest solution (see the recent documentation):
from sqlalchemy.inspection import inspect
columns = [column.name for column in inspect(model).c]
The last line might look more readable, if rewrite it in three lines:
table = inspect(model)
for column in table.c:
print column.name
I’m using SQL Alchemy v 1.0.14 on Python 3.5.2
Assuming you can connect to an engine with create_engine(), I was able to display all columns using the following code. Replace “my connection string” and “my table name” with the appropriate values.
from sqlalchemy import create_engine, MetaData, Table, select
engine = create_engine('my connection string')
conn = engine.connect()
metadata = MetaData(conn)
t = Table("my table name", metadata, autoload=True)
columns = [m.key for m in t.columns]
columns
the last row just displays the column names from the previous statement.
print repr(model.__table__)
Or just the columns:
print str(list(model.__table__.columns))