The python object "db" was created by a module in global name space. By repacking the python object into a function that instantiates the db as a new object, the problem has been resolved.

However I was still surprise that db.disconnect() doesn't clear out completely.

Hi! According to current design, a Database object cannot be "unbinded" from the provider. The disconnect method just closes a DBAPI connection to a database server. The reason for that behavior is that some properties can be set only after we know specific database server. For example, I may be interested in knowing MyEntity._table_ property which contains a table name for this specific entity. This table name may be different for different databases: MYENTITY in Oracle and myentity in other databases. So, the hypothetical unbind method should also 'reset' all such properties to its initial state, and it may be hard to do. Currently we suggest to create db object inside a factory function, and make separate instance of db object for each database

Your suggestion is the solution I went for.

So far I'm very impressed with pony.

Thanks, please let us know what improvements you'd like to see in pony. This could help us to make it even better ;)

Will do :-)

Hi, I am getting very bad performance in the view of my web app, I have narrowed it down to code like this: in for loop: product = Product.get(lambda p: p.attr1 == x and p.attr2 == y) using sql_debug, I see it's doing a select query with LIMIT 2, which I suspect is making postgres scan the whole table in every loop, any idea why there's a LIMIT 2?

A lookup using the default limit 2 = 30ms, with limit 1 it's 10ms

from running the generated query manually in psql

product = Product.select(lambda p: p.attr1 == x and p.attr2 == y).limit(1)[? or [None])[0] is approx 3x faster than the get, correctly issues a limit 1

with a correct index, it's a lot faster, but still a 3x difference between limit 1 and limit 2

The get method assumes that no more than one object satisfies specified criteria. Pony attempts to retrieve more than one object in order to throw exception if multiple objects were found. Sometimes the query condition is wrong and a very big number of objects satisfy the criteria. Pony adds LIMIT 2 in order not to load all that objects to memory.

Does it take composite_keys into account? That could be an optimisation when you know the database restricts the possible results to be max 1

Or maybe a get parameter / get version that only does limit 1?

Hi, this form should take composite key into account: Product.get(attr1=x, attr2=y)

Hi, this form should take composite key into account: Product.get(attr1=x, attr2=y)

I tested this with a composite_key present on attr1 and attr2, it is still LIMIT 2

Hmm, that's strange. How do you define the composite key? Is it composite_key(attr1, attr2) in entity definition? Remember that composite_index is not sufficient, because composite_index is not unique

yes it is composite_key(attr1, attr2)

Looks like a bug... I'll try to fix it

This is with SQLite by the way

Ok, it should work with any database

Any update on this?

No. We'll fix it tomorrow

Suggestion: Add an Unsigned checkbox to the editor

Sorry, I fixed, it but we had hard refactoring and we needed time to test it all. I will ping then it would done(very soon)

Suggestion: Add an Unsigned checkbox to the editor

Hi Joery, Where do you suggest adding this?

@alexeymalashkevich

Good point, thanks

I tested this with a composite_key present on attr1 and attr2, it is still LIMIT 2

Hello. I just fixed this bug. Already released on GitHub.

Hi, great! Does it work with lambda style queries, or only **kwargs style get?

Only for kwargs.

It's too complex to analyse lambda. If you know that you are searching by unique key - kwargs is enough.

Agreed :) thanks!

Is a new release planned soon?

I think tomorrow.

Didn't have enough time to make a release today - documenting new features takes time

Same for LongUnicode

In Pony unicode and str mean the same (unicode), for easer porting of application code to Python3. You can use str in models with meaning "unicode"

In Pony unicode and str mean the same (unicode), for easer porting of application code to Python3. You can use str in models with meaning "unicode"

Oh okay!

class Example(db.Entity): test_attribute = Optional(int, size=8, sql_default=0, default=0) The editor generates the sql_default attribute as an integer, but shouldn't it be a string?

It should. Thanks for reporting, this is a bug

Hi guys. Do you have best practice or templates for new project which use pony orm ? I mean where I must to store models and where others?

I need your advice.

I start with a single file for a flask + pony project until it gets larger or I need to access the models outside the web app, then I move the models to a models.py

I once made the mistake of using a single models.py file for multiple applications. I needed just a few of the models in each. But when I called db.bind it created all of the models in my database (not just the ones I imported from models.py) and I ended up with clutter. So, don't do what I did :)

thanks

Is it possible to set a required field in before_insert? class Article(db.Entity): slug = Required(str, 200, unique=True) title = Required(str, 255) def before_insert(self): self.slug = self._slugify_url_unique(self.title)

use __init__()

Thanks will try.

def __init__(self, title): slug = xxx(title) super().__init__(title=title, slug=slug)

Hmm now I have class Article(db.Entity): _table_ = 'articles' slug = Required(str, 200, unique=True) title = Required(str, 255) description = Required(LongStr) body = Required(LongStr) created_at = Required(datetime, 0, default=datetime.utcnow) updated_at = Required(datetime, 0, default=datetime.utcnow) author = Required(User, reverse="articles") favorited = Set(User, reverse="favorites") comments = Set("Comment") tag_list = Set("Tag") def _unique_check(self, text, uids): if text in uids: return False return not self.exists(slug=text) def _slugify_url_unique(self, text): slugify_url_unique = UniqueSlugify( unique_check=self._unique_check, uids=['feed'] ) slugify_url_unique.to_lower = True slugify_url_unique.stop_words = ('a', 'an', 'the') slugify_url_unique.max_length = 200 return slugify_url_unique(text) def __init__( self, title, description, body, author, tag_list ): slug = self._slugify_url_unique(title) super().__init__( slug=slug, title=title, description=description, body=body, author=author, tag_list=tag_list )But getting AttributeError: 'Article' object has no attribute 'exists'. Shouldn't it have that attribute?

No, this method is defined for entity class only. You can write self.__class__.exists(...) instead. Maybe in the future we'll make it a usual classmethod which can be called both from class and from object

@akozlovsky Why does def before_insert(self): self.slug = self._slugify_url_unique(self.title)doesn't work?

What exception did you get?

ValueError: Attribute Article.slug is required

before_insert works before the INSERT command is sent to the database, not before the __init__ code is executed. When the attribute is specified as required, Pony does not allow to create an instance of object without that attribute defined. We can add something like option delayed_check=True to allow temporary creation of object without all required attributes, but I'm not sure this is a good practice. I think you can create make_article function (or Article.make classmethod) which will create a slug before creating an Article instance: class Article(db.Entity): ... @classmethod def make(cls, **kwargs): if `slug` not in kwargs: kwargs['slug'] = cls._slugify_url_unique(kwargs['title']) return cls(**kwargs) With this approach, you need to make _slugify_url_unique and _unique_check classmethods too

And would it be possible to set a function as default value?

Sure, you can do something like: from datetime import datetime from uuid import UUID, uuid4 class MyObject(db.Entity): created_at = Required(datetime, default=datetime.now) guid = Required(UUID, unique=True, default=uuid4)

But not with an argument, right?

Sure. With argument it will be just one specific value

def __init__(self, title): slug = xxx(title) super().__init__(title=title, slug=slug)

Thanks again! Using __init__() works for me.

I either just use Optional, or set a fake initial value to be overwritten in after_insert

Yeah thought about that. A reason why you use after_insert not before_insert?

oh that was just an example :) depends on the usecase

As slug is unique I think I cannot use Optional. Maybe use UUID as default.

that would work

Just for case it is possible for unique attribute to be Optional, because in all databases except MS SQL Server NULL values are not considered to be unique

So it is possible to have multiple NULL values in unique column

Oh so I could set default to null and make it Optional.

yes

So this sounds actually the least 'hacky'. ?

@akozlovsky Getting TypeError: Default value for non-nullable attribute Article.slug cannot be set to Nonewhen setting class Article(db.Entity): _table_ = 'articles' slug = Optional(str, 200, unique=True, default=None) title = Required(str, 255)From the documentaion unique optional fields should be nullable: > If an optional string attribute is used as a unique key or as a part of a unique composite key, it will always have nullable set to True automatically.

I'm getting also pony.orm.core.TransactionIntegrityError: Object Article[new:1] cannot be stored in the database. IntegrityError: NOT NULL constraint failed: articles.slugfor SQLITE when not using before_insert. When adding nullable=True and using def before_insert(self): if not self.slug: self.slug = self._slugify_url_unique(self.title)everything works fine! ?

^^'

Which is the proper way to separate the models into different files? I have users.py with the following content: from pony.orm import * from datetime import datetime db = Database() class User(db.Entity): uuid = PrimaryKey(int, auto=True) name = Required(str) token = Required(str) registered_at = Required(datetime) active = Required(bool) And the main.py with this: from pony.orm import * from models.users import User def main(): db = Database() db.bind(provider='mysql', host='127.0.0.1', port=3306, user='root', passwd='example', db='example') sql_debug(True) db.generate_mapping(create_tables=True) if __name__ == '__main__': main() But when I execute the main function it does not create the tables and it does not throw any exception. The code works if I put it into the same file.

You have to import db from users. Not create it a second time.

Will try tomorrow morning and see if it works. Thank you :D

hi all! Is pony has a on_save or before_save methods for calling on pony entities creation? i want to do some actions before pony save data

hi Nickolai https://docs.ponyorm.com/api_reference.html#entity-hooks

Hey - At work were having a discussion about whether to bind and execute pony calls in functional manner, or, whether to keep the bindings and store them in a dictionary. We're connecting to multiple databases as multiple users, so the diversity cannot be avoided. Any thoughts on performance here? A) Use and drop the bindings - the users wont notice. B) Keep and maintain the bindings? it matters.

Is it in a long running process?

Could be.

Let's say yes.