(which is checked for in line 228)

{it could of course also be me who is an idiot, as I perhaps should use absolute paths instead...}

I've also faced with that magic in attempts to pass is_path parameter to sqlite

(it can be used to get some extended features of in-memory databases like shared cache)

Hmmm.... I'm not too fond of that. In all cases the write should be to os.getcwd(). Not to an import.

Hi, this is long-standing behavior. The reason for it is mod_wsgi. Initially mod_wsgi was the most popular way to use Pony with "real" web-server. And in mod_wsgi the current working directory is set to some obscure path. So, the typical error was, the user creates a sample application to use from mod_wsgi, and specifyes in it db = Database('sqlite', 'mydatabase.sqlite', create_db=True) When the script is running locally, using CherryPy or Flask, it creates a database file in a current working directory, that is, typically the same directory where the script placed. A user populate the database with some example data, and then run the application under mod_wsgi. But then he encounters a strange bug: the database is empty! Although, the user sees that the populated database lies right here.

So relative filenames were just unusable for that reason. And because of this, some frameworks, like Django, just forbid to use relative path in settings: https://stackoverflow.com/questions/28635405/why-absolute-paths-in-django-settings

We had that option too, it was possible to raise an error when the user provide a relative path to the database, in order to prevent that obscure error. But then we thought, if relative path in its usual meaning cannot be used, why not allow them, but with the different meaning, "relative to the file where the database object is defined". Sometimes it is useful.

And, you always can specify absolute path to be exact.

Maybe it is possible to change that behavior now, but (1) it breaks backward compatibility, (2) "real" relative path can still be unreliable, for example if you have multiple scripts placed in different directories which all import the same db object, and (3) "relative to db object" semantics is quite handy sometimes

Hi - Sorry for the silence: My team has settled on using absolute paths for good. 👍 (bug report cancelled 😊👍)

Is there a way to add a computed property to an entity that will be included by default in to_dict?

I've just overridden to_dict in my entity to take care of it, but I was wondering if there was a better way.

I think sqlalchemy calls this "hybrid attributes"...

Hi, we want to add hybrid attributes soon, but this task turns out to be surprisingly difficult. Right now you need to override to_dict

Okay, thanks!

Hi all

I ve using Pony for my small project and I am very happy to use it especially because the syntax is very pythonic way.

You asked info about lastest release, idk why did you delete your message but if you care... Here you can read info about it https://github.com/ponyorm/pony/releases/tag/0.7.3

I am going to check it right now , thanks..

Or here https://blog.ponyorm.com/2017/10/23/pony-orm-release-0-7-3/

Thanks for your fast response

No problem

hi.. im new to pony, now im doing validation before inserting data into the database, what your suggestion for doing the validation, is it before the insert using entity_hook, or plain validation before initiating new entity instance?

What kind of an application is it? If it's a web app, I validate in my Forms (WTForms or Django Forms)

In general, I validate outside of the model

Sometimes after validation it won't make sense to make an instance of the model, for example if a required piece of data is missing or invalid

an API, using falcon..

so do it in the code before creating the model instance

Ok

and thanks..

no problem

Not sure if it's a good idea but you can use class method of model to create instance. Class method allows you keep validation code not so far from model and don't initialize instance if data is wrong.

yes, I've​ been thinking something like that, but haven't really know how to do it properly, since I use schematics library to do the validations duplicating pony entities into schemas is not ideal I think.. Maybe I might try that one..

https://github.com/ponyorm/pony/issues/315 hmmm

Have joined the framework benchmark with Morepath using Pony (on Python 3).

https://www.techempower.com/benchmarks/previews/round15/#section=data-r15&hw=ph&test=db&l=zijzen

Difficult to compare as Morepath is the only framework which uses Pony. But looks promising. Especially the tests using the database where Pony is involved.

Hmm, I see Morepath, but I don't see Pony in this list. Do you mean that all Morepath results are with Pony?

Yes at least the ones which uses the database.

So, if I read correctly, Morepath+Pony is approximately two times faster than Django

Depending which test but yeah a big difference. BTW json and plain text tests doesn't use the database. The others do.

Especially when you take into account that it's Python 3 and compare it with the PY3 versions of the other frameworks it looks quite amazing.

That's great!

Yeah. In fact, in terms of speed, I think Pony still has some room for growth ) The current code was written just from "theoretical" point of view, and profiling on real application may find some local bottlenecks which can be optimized even more

But even now it looks quite good

great to hear thta\

BTW the source code of the benchmark is here : https://github.com/TechEmpower/FrameworkBenchmarks/tree/master/frameworks/Python/morepath

Hi! I am new to Pony, and got stuck with MySQL. I have a table of users, and datetime field in it, and pony's mysql connector raises exception during parsing '0000-00-00 00:00:00', which is returned for NULL datetime fields lib/python2.7/site-packages/pony/orm/dbproviders/mysql.py", line 334, in str2datetime return datetime(*imap(int, s.split())) ValueError: year is out of range

How can I make Pony treat this value as None ?

Use custom datetime type class

Hello everybody, anyone know how to make a custom autoincrement like primary key? E.g. RCRD_001, RCRD_002, RCRD_003, etc. Thank you so much.

Hi Juan. What database do you use?

/effect@bonbot

/eduvote@bonbot

/whereami@bonbot

/whereami@bonbot

/whereami@bonbot

Hi Juan. What database do you use?

I'm using MariaDB

I think it is better to have integer primary key, because it is smaller and more efficient. But if you really need a string key, you can probalby declare it as a secondary key using nullable UNIQUE column and initialize it using an AFTER INSERT trigger

I think it is better to have integer primary key, because it is smaller and more efficient. But if you really need a string key, you can probalby declare it as a secondary key using nullable UNIQUE column and initialize it using an AFTER INSERT trigger

I was thinking something like this, but I though that using ponyorm I could create a custom primary key. Thank you for your answer.

With pony you can have a string primary key, but if it includes some "counter" you need to process it somewhere. You can store the counter value in a separate table, and use to construct primary key in Pony: class Counter(db.Entity): entity_name = PrimaryKey(str) value = Required(int, default=0) class MyEntity(db.Entity): number = PrimaryKey(str) foo = Required(str) bar = Required(str) @classmethod def make(cls, **kwargs): cls_name = cls.__name__ counter = Counter.get(entity_name=cls_name) \ or Counter(entity_name=cls_name) counter.value += 1 number = 'RCRD_%d' % counter.value return cls(number=number, **kwargs) ... with db_session: obj = MyEntity.make(foo='foo', bar='bar') But it will be less efficient then using integer autoincrement primary keys

@bonbot Who are you?

Use custom datetime type class

Sorry, I'm not sure if this will fix the problem? str2datetime is assigned into the conv, which is converters I guess, like this conv[FIELD_TYPE.TIMESTAMP] = str2datetime conv[FIELD_TYPE.DATETIME] = str2datetime I can't see if custom datetime class can somehow affect this I've even tried using str, instead of datetime class, but didn't succeed. It raises exception way before getting to my object

Sorry, I'm not sure if this will fix the problem? str2datetime is assigned into the conv, which is converters I guess, like this conv[FIELD_TYPE.TIMESTAMP] = str2datetime conv[FIELD_TYPE.DATETIME] = str2datetime I can't see if custom datetime class can somehow affect this I've even tried using str, instead of datetime class, but didn't succeed. It raises exception way before getting to my object

Use str with sql_type="db specific". After define in entity property date. And implement in converter to datetime

Hi Bulat, thanks for reporting, I think we need to fix it

Hi Bulat, thanks for reporting, I think we need to fix it

👍

That would be great!

Hi! I am new to Pony, and got stuck with MySQL. I have a table of users, and datetime field in it, and pony's mysql connector raises exception during parsing '0000-00-00 00:00:00', which is returned for NULL datetime fields lib/python2.7/site-packages/pony/orm/dbproviders/mysql.py", line 334, in str2datetime return datetime(*imap(int, s.split())) ValueError: year is out of range

Fixed. Should work now. This date will represent as None.

Good morning - Just FYI: We're using pony with SQLite for unittesting and postgresql in production. It's awesome. 10 x 👍 hashtag @dematic.com

Of mere curiosity: Are anyone working on implementing database triggers into db.Entity?

I'm glad you like it 😎

At this moment we don't working on database triggers. But you can create them manually and use them with pony

Good morning - Just FYI: We're using pony with SQLite for unittesting and postgresql in production. It's awesome. 10 x 👍 hashtag @dematic.com

That means you will not use anything particular to Postgres right? e.g. JSON & JSONB

The problem with triggers is they may change data in the database and pony will not know about that, so object already loaded into memory may become out of date. It is possible in pony to mark attribute as volatile, that means it can be changed by trigger, so pony re-fetch attribute value after the object was changed

@ifindam We decided about 3 months ago to use ponyorm for all database interaction - including the database definition - which is stored in a "model.py" that is importable for all other modules.

@ifindam We decided about 3 months ago to use ponyorm for all database interaction - including the database definition - which is stored in a "model.py" that is importable for all other modules.

Cool 👍

By starting in pony, we can create unittests with: from model import get_db where: db = get_db(as_test=True) leads to the loading of an sqlite db with necessary test data loaded into memory.

db = get_db() leads to loading of the postgres production defaults.

We use nosetests' general principles with: def setup(): db global db = get_db(as_test=True) def teardown(): pass # drops the db from memory all code inbetween can then run with the "in memory" db variable shared.