sa default # 49

[vir-mir]

Hi!
I offer a patch for the default fields.

Do you think it makes sense for life?

@asvetlov

Issues #49

[asvetlov]

Wow! Very impressive.
Let me sleep on it and make careful review.

[codecov-io]

Current coverage is 92.26% (diff: 100%)

Merging #206 into master will increase coverage by 0.07%

@@             master       #206   diff @@
==========================================
  Files            12         12          
  Lines          1319       1331    +12   
  Methods           0          0          
  Messages          0          0          
  Branches        164        166     +2   
==========================================
+ Hits           1216       1228    +12   
  Misses           74         74          
  Partials         29         29          

Powered by Codecov. Last update d4cef36...b1ae7be

[vir-mir]

• Combed code.
• Add processing scalar, sequence and fix set value None where default not null.

[vir-mir]

@asvetlov any news?

[pomidoroshev]

@asvetlov Чё как?

[Eduard90]

We need default values!

[jettify]

Can we avoid doing additional queries, by compiling nextval ({}) into statement like:

INSERT INTO table VALUES(nexval('table_id_seq')); 

just curious...

[vir-mir]

sqlalchemy makes a queries to the db.
Of course, we can make queries in advance...
such behavior we should document it explicitly

[vir-mir]

Sorry!
I tested.

tbl = sa.Table('sa_tbl4', meta,
               sa.Column('id', sa.Integer, nullable=False, primary_key=True),
               sa.Column('id_sequence', sa.Integer, nullable=False,
                         default=sa.Sequence('id_sequence_seq')),
               sa.Column('name', sa.String(255), nullable=False,
                         default='default test'),
               sa.Column('count', sa.Integer, default=100, nullable=None),
               sa.Column('date', sa.DateTime, default=datetime.datetime.now),
               sa.Column('count_str', sa.Integer,
                         default=sa.func.length('abcdef')),
               sa.Column('is_active', sa.Boolean, default=True))

ins = tbl.insert().values()
ins.bind = engine
print(str(ins))

INSERT INTO sa_tbl4 (id_sequence, name, count, date, count_str, is_active) VALUES (nextval('id_sequence_seq'), %(name)s, %(count)s, %(date)s, length(%(length_1)s), %(is_active)s) RETURNING sa_tbl4.id

that is, everything works correctly. I will soon do commit

[jettify]

@asvetlov what do you think?

[jettify]

@vir-mir thank you very much for this PR! Lets give @asvetlov chance to review on weekends, otherwise I will merge it on Monday.

[asvetlov]

Looks good.
@jettify please merge

[serg666]

Guys, what about this comment ?

[Eduard90]

I have problems with UUID (PostgreSQL) column type:

  File "/usr/lib/python3.5/uuid.py", line 137, in __init__
    hex = hex.replace('urn:', '').replace('uuid:', '')
AttributeError: 'UUID' object has no attribute 'replace'

Table:

InvoiceTable = NewTable('invoice', metadata,
                        Column('id', UUID, primary_key=True),
                        Column('affiliate_id', UUID),
                        Column('status', String, index=True, default=StatusEnum.not_approved.name),
                        Column('created_at', DateTime, default=datetime.now),
                        Column('deleted', Boolean, default=False),
                        Column('is_paid', Boolean, default=False),
                        )

Primary key (id) is UUID and "by default" it must generate new UUID on insert.
I have "monkey patch" for _execute method (class SAConnection) but it's very bad code ))

[vir-mir]

@Eduard90

http://docs.sqlalchemy.org/en/latest/core/custom_types.html#backend-agnostic-guid-type

class UUID(sa.types.TypeDecorator):
    impl = sa.String

    def __init__(self, length=255, *args, **kwargs):
        super().__init__(length=255, *args, **kwargs)

    def process_bind_param(self, value, dialect=None):
        if value and isinstance(value, uuid.UUID):
            return value.bytes
        elif value and not isinstance(value, uuid.UUID):
            raise ValueError('value %s is not a valid uuid.UUID' % value)
        else:
            return None

    def process_result_value(self, value, dialect=None):
        if value:
            return uuid.UUID(bytes=value)
        else:
            return None

    def is_mutable(self):
        return False

meta = sa.MetaData()
tbl = sa.Table('sa_tbl4', meta,
               sa.Column('id', UUID(), primary_key=True, default=uuid.uuid4))

...

yield from conn.execute(tbl.insert().values())
yield from conn.execute(tbl.insert().values())
yield from conn.execute(tbl.insert().values())

[jettify]

@vir-mir thanks! should we put this snippet to the FAQ?

[vir-mir]

@serg666 #56 (comment)

[vir-mir]

@jettify all say that it is necessary to use an approach TypeDecorator

[Eduard90]

@vir-mir Yes, i know about this way. But ... with this custom type Alembic generate not correct migrations.
This problem in Alembic maybe. But i want to use UUID and migrations now )))

And sqlalchemy has "builtin" type:
from sqlalchemy.dialects.postgresql import UUID
I think it's more correct than custom type. Or not?

[Eduard90]

@vir-mir How i can talk with you directly? Skype, Telegram, etc?

[vir-mir]

This is problem alembic

  File "../site-packages/sqlalchemy/sql/schema.py", line 79, in _init_items
    item._set_parent_with_dispatch(self)
AttributeError: 'UUID' object has no attribute '_set_parent_with_dispatch'

you can use
from sqlalchemy.dialects.postgresql import UUID

But to make it work you need to add a parameter default

from sqlalchemy.dialects.postgresql import UUID
import uuid

meta = sa.MetaData()
tbl = sa.Table('sa_tbl4', meta,
              sa.Column('id', UUID(as_uuid=True), primary_key=True, default=uuid.uuid4))

...

yield from conn.execute(tbl.insert().values())
yield from conn.execute(tbl.insert().values())
yield from conn.execute(tbl.insert().values())

My skype virmir49