function with db_session

I tried setting db_session(strict=True) and ussing the context manager (with db_session)

Transactional ORMs doesnt work good with async. Pony does not support async.

Transactional ORMs doesnt work good with async. Pony does not support async.

My simple solution has been to just put all the DB code on it's own thread and talk to that thread via zeromq when I need database access.

Thats good

I think

Works great for everything I've needed thus far. With the DB code all in it's own private thread, it has nothing to contend with for access, and no deadlocks of any kind, thanks to the magic of zeromq messaging.

Just send off commands to the DB thread, and await responses. All very straightforward.

Might be a good practice. Maybe we should write some article about all these things.

I've never been against more helpful documentation bein' added to the web. :)

In truth, I have plans to add myself a proper blog to the web soon documenting exactly such fun things I've learned while exploring Python. :)

Python's prolly the most enjoyable programming language I've ever worked with, so it's been a lotta fun discovering the various toys I have available to me. Pony ORM's been one of the better ones, and ZeroMQ too. Still wrapping my brain around all the new async stuff tho.

Async programming requires a different way of planning one's code than I'd been previously used to.

In truth, I have plans to add myself a proper blog to the web soon documenting exactly such fun things I've learned while exploring Python. :)

Great! 🌟🌟🌟 Please share the link once you post it

Surely will do. Prolly gonna use Nikola (https://getnikola.com/) to generate the site/blog. Big fan of static HTML. :)

Transactional ORMs doesnt work good with async. Pony does not support async.

Hmm, thank you anyway xD

I googled and couldn’t find much: What’s the suggested way to use pony with type annotations? How about mypy?

We didn't think much about type annotation support yet

Maybe we need now

Running mypy on a file with some pony models gives a bunch of errors about imports, which currently I think means any project with pony in it won’t work with mypy

in general types seem neat, and will probably make python code a lot more solid

Maybe we need now

Check out pydantic, it might help with input

Ok, thanks

Doesnt it conflict with py2?

Maybe we need now

Yeah, to keep it "pythonic" as you claim, that's really be needed.

Doesnt it conflict with py2?

I think py2 required type hints in comments, whereas modern py3 versions can have “native” hints

I mean that pony supports also python2 and it uses the same code as python 3.

So probably we’ll add type annotations after we dump Python2 support. At this moment it probably still used in some scientific code

👋🏻 It's possible to create an object with its sub-objects? for example: Command (name = "new_command", rules = [Rule (text = "rule")], messages = [Message (text = "msg")])

yes

But actually in this code Rule or Message are created before Command, not at the same time. If Rule.command or Message.command is required attributes, you need to create Command before and these objects later

👋🏻 It's possible to create an object with its sub-objects? for example: Command (name = "new_command", rules = [Rule (text = "rule")], messages = [Message (text = "msg")])

insted of: `cmd = Command(name="cmd") cmd.rules.create(text="rule") cmd.messages.create(text="msg")`

if Rule.command and Message.command are optional, you can create objects in any order. You show two options, and the third option is cmd = Command(...) Rule(..., command=cmd) Message(..., command=cmd)

Ok, I'll just change to optional

🙏🏻

If it really is required, it is better to leave it as required

Right. Another question, is there a one-way relationship?

No, but on the other side you can add some Set attribute with arbitrary name (Also, in my experience, this "unnecessary" attribute may turn out to be quite useful to express queries in a simple and compact way)

No, but on the other side you can add some Set attribute with arbitrary name (Also, in my experience, this "unnecessary" attribute may turn out to be quite useful to express queries in a simple and compact way)

Can you explain a little more?

Like, you think that you want to have one-way relationship Person->House. You cannot do it in Pony (right now), so you need to add reverse attribute to House too. In most cases it wil be Set attribute, but sometimes it may be Optional one. so you need to define class Person(db.Entity): name = Required(str) house = Optional("House") class House(db.Entity): address = Required(str) persons = Set("Person") If you are not going to actually use House.persons attribute it may be named arbitrarily. But it may be useful for many kinds of queries, like instead of select(p for p in Person if p.house == h) you can just use h.persons

Note that House.persons is a "virtual" attribute which does not have corresponding table column

Ok, thanks!😀

Hi! Been looking into alternatives for peewee, as I was having issues with it when the database connection drops. How does pony handle that situation? I try execute a query, but the database is not available at the moment. In peewee, this would always happen eventually - the connection would drop, even if it can be immediately reestablished and any queries would fail. Can I set pony up to automatically reconnect and retry? Maybe even keep retrying periodically until a timeout is reached?

Hi! Pony automatically reconnects under the hood if a database connection closes unexpectedly. Pony checks the state of the current transaction. If the transaction is just started, or up to this point you performed only SELECTs, Pony will reconnect silently. But if at least one UPDATE was executed, then an exception is raised instead, because we lost the state of transaction after the connection was closed

Sounds sensible to me. Very nice to hear :)

You are welcome :)

Just to confirm: This is only within a transaction, right? If I do individual statements outside a transaction or start a new transaction the "update was performed" thing is irrelevant?

If dropped connection had some uncommitted inserts/updates/deletes, then we cannot reconnect silently. Otherwise, we can. If you work in interactive mode outside of db_session, transaction is still started on first insert/update/delete, and will continue until explicit commit()/rollback()

Hmm, I guess if it's wrapped in a transaction a failed query would, at the very least, cause the commit to throw an exception? To be fair, my main concern isn't with failed queries mid-transaction, but I'm curious now.

with db_session: s = Student[123] #1 dept = s.group.department #2 s.scholarship += 100 #3 some_students = select(s for s in Student if s.scholarship > 500) #4 some_students[0].group = Group[456] #5 #1 This is the first SELECT in this db_session, if we are trying to execute it and see broken connection, we can reconnect silently without any problem #2 Additional SELECT to fetch department. If we are trying to execute it and see broken connection, we still can reconnect silently #3 We perform some updates in memory, but these updates are not saved to database yet. Pony doesn't talk with database here #4 When Pony sees that query, it also sees there are some unsaved changes. So it tries to save changes before issuing SELECT in order to get correct result from the query #4.1 If Pony tries to do first UPDATE and sees broken connection, it still can reconnect silently #4.2 after that point (on any subsequent UPDATE/SELECT) Pony cannot reconnect silently, because the first UPDATE was rolled back by database when connection was dropped, and a new connection will not have that first UPDATE applied #5 the same as 4.2, it is too late to restore connection silently, Pony needs to throw an exception here, because some updates were lost

It is seldom that connection drops in the middle of transaction, usually it happens between transactions, and then the connection will be restored at the start of the next transaction

Hence my comment that I'm not too worried about it dropping mid-transaction. You say it cannot fail silently, that means it will just throw an exception, right? A failure in one transaction will not affect any subsequent transactions?

yes, after we start a new transaction after exception was raised Pony will open a new connection instead of dropped one

Awesome. I am really liking this. I'm gonna give pony a go, I think. It will definitely beat peewee, which I used previously and where I was told that queries failing due to a dropped connection (between transactions) is intentional...

Hi, is there an alternative of CREATE TYPE in Pony ORM?

Hi, what type do you want to create?

hi, any way to define a custom dump/load method for a json column?

For what purpose?

I have json columns on tables where i save info from dicts

currently i'm using rapidjson, with some specific way to match Dateeimes/decimal/uuids fields

if i can specifc my custom loader/dumps i can save double conversion

> currently i'm using rapidjson is it for c++?

yep, there is a bind for python

python-rapidjson

Do you write from c++ directly to database and then read from Python or what?

https://github.com/python-rapidjson/python-rapidjson

https://github.com/python-rapidjson/python-rapidjson

Permanent link to the python-rapidjson/python-rapidjson project you mentioned. (?)

i'm using this because is very strict/configurable with json conversion, a part of very perfomance

Currently'm sanitazing the information in this way: try: dat_str = rapidjson.dumps(stint.get('datum'), uuid_mode=rapidjson.UM_CANONICAL, number_mode=rapidjson.NM_DECIMAL, datetime_mode=rapidjson.DM_ISO8601) dat_dict = rapidjson.loads(dat_str) except: self.logger.exception("Error converting close datum: {}".format(stint.get('datum'))) dat_dict = { 'error': 'Error converting close datum' } finally: update_current_act['datum'] = dat_dict

but at the end, i'm parsing and reparsing the json

update_current_act is the db.Entity and datum is a json Column

What database are you using?

postgresql

Right now it is not customizable, you can open an issue to allow specifying JSON serializer

Hi guys, we just released PonyORM 0.7.12 ## Features * CockroachDB support added * CI testing for SQLite, PostgreSQL & CockroachDB (see test result statuses here: https://github.com/ponyorm/pony/blob/orm/README.md) ## Bugfixes * Fix translation of getting array items with negative indexes * Fix string getitem translation for slices and negative indexes * PostgreSQL DISTINCT bug fixed for queries with ORDER BY clause * Fix date difference syntax in PostgreSQL * Fix casting json to dobule in PostgreSQL * Fix count by several columns in PostgreSQL * Fix PostgreSQL MIN and MAX expressions on boolean columns * Fix determination of interactive mode in PyCharm * Fix column definition when sql_default is specified: DEFAULT should be before NOT NULL * Relax checks on updating in-memory cache indexes (don't throw CacheIndexError on valid cases) * Fix deduplication logic for attribute values

MySQL and Oracle tests builds will be published as well.

Hi, what type do you want to create?

Sorry, I was not interested in messages. For example in this link CREATE TYPE bug_status AS ENUM ('new', 'open', 'closed'); CREATE TABLE bug ( id serial, description text, status bug_status );

At this moment it is not supported, instead you can use int values with names defined in Python class bug_status: new = 1 open = 2 closed = 3 class Bug(db.Entity): id = PrimaryKey(int, auto=True) desctiption = Required(str) status = Required(int, default=bug_status.new) I hope we can add Enum support soon

Ok, thank you

Hello guys, I am sorry, but I will ask the question right from the start - is there any way of going from pony code to diagram? or only from diagram to code?