actually we are checking last_update dt on client and server and compare them. Also autosave do save only if there are any unsaved data.

Oh, but at least the intervall isn't changed, so it can trigger right after you cancel

I can never remember how to do this in pony - how do I select the object with the largest id for each day for the past n days?

select((s.created.date(), max(s.id), s) for s in FacebookDataSnapshot if s.created.date() >= datetime.date.today() - datetime.timedelta(days=7))[:]

This works, but then I have the ID in the results, can I do itt anotther way so it only returns the date and the object?

I'm no sure this is a correct query

Probably it should be something like

oh yeah it doesnt give one result per date

select(s for s in FacebookDataSnapshot if s.id in select(max_id for date, max_id in ( select((s.created.date(), max(s.id)) for s in FacebookDataSnapshot if s.created.date() >= datetime.date.today() - datetime.timedelta(days=7)) ))).order_by(lambda s: s.created.date)[:]

In SQL this query is not an easy one too

Note that I removed s from (s.created.date(), max(s.id)) to group objects by date

that makes sense! I am getting AttributeError: 'MethodMonad' object has no attribute 'getsql' though

ah fixed it

I forgot parentheses after the last date

removed the date from the order by

ok

Thank you!

Sure

What is the current way of supplying a pony orm model as pluggable ORM in libraries, next to e.g. redis or mongodb connectors? https://github.com/ponyorm/pony/issues/330 I don't like to start using sql alchemy just for that.

What is the current way of supplying a pony orm model as pluggable ORM in libraries, next to e.g. redis or mongodb connectors? https://github.com/ponyorm/pony/issues/330 I don't like to start using sql alchemy just for that.

https://github.com/luckydonald/telestate/tree/deploy2/telestate/contrib

Hi Luckydonald, I answered on GitHub: https://github.com/ponyorm/pony/issues/330#issuecomment-470153594

Hi 🙋‍♂

I am try use pony with fdb but show this error no module named fdb

And import fdb not show nothing error

I installed fdb

Hi, at this moment Pony doesn't work with fdb, it supports PostgreSQL, MySQL, SQLite and Oracle

😔

Thanks Alex 👍

What's fdb?

Firebird

Firebird

Interesting. Can someone enlighten me why it would be chosen as DB compared to mysql/postgres?

Can't really see any huge differences

I think it’s mostly used in enterprise. It’s a descent of interbase or something. Think it runs on more commercial unixes and mainframes and stuff compared to Postgres or whatever. Stuff like AIX, Solaris, HP-UX, IBM ZOS...

Same reasons you might pick innoDB vs another MySQL engine I guess. Some workloads are better suited to it. Don’t ask me which ones tho

I would guess there is not a lot of overlap between ORM users and fire base users :-P

I would guess there is not a lot of overlap between ORM users and fire base users :-P

So far only one of ~200

Hi there

Welcome!

The main benefit of upsert is atomicity. X.get(foo=1) or X(foo=1) sometimes can raise an exception, if another transaction was able to insert object in between

Hey, do we have upsert support now? I don't wanna get UnrepeatableReadError/OptimisticCheckError, but just create/update a entity. Especially in concurrent environments, the last should just win.

MongoDB has mongodb_table.replace_one( filter={'chat_id': chat_id, 'user_id': user_id}, replacement={ 'chat_id': chat_id, 'user_id': user_id, 'state': state_name, 'data': state_data, } )

Hi! I forget, what database are you using?

All of them, the user using the library should decide later.

Sorry, we don't have support of upserts yet But I agree this is a important feature Maybe I can add it next weekday

👍

This would be really great. Many SQL languages already have a concept of that, and for the ones that doesn't a regular get + set/create approach could be put in as placeholder, to at least get startet.

My current implementation while True: try: # updating existing state first State.get( chat_id=chat_id, user_id=user_id ).set( chat_id=chat_id, user_id=user_id, state=state_name, data=state_data, ) break # successful except Exception: pass # end try try: # creating a new state State( chat_id=chat_id, user_id=user_id, state=state_name, data=state_data, ) break # successful except: pass # end def # end while Most certainly this will get to a point in time where it creates/updates without raising any errors.

What would the best current possible way to get around the failures due to concurrency and multiple instances?

How often do you need this operation in transactions?

I mean, what percentage of transactions uses this operation?

I am building a state managment system which will load at a telegram (tele-)flask request, and store the new state afterwards. class State(db.Entity): user_id = Required(int) chat_id = Required(int) state = Required(str) data = Required(Json) orm.PrimaryKey(user_id, chat_id) # end class This particular library (telestate) does only that, but a user could do his own stuff in between. Does that provide answer to your question?

Will single db_session mix this update and arbitrary user updates?

Hmmm.... Here is a pseudocode mock of how it'd work def process_incomming_telegram_update(self, update): state = self.load_state_for_update(update) for func in userfuncs: global state result = func.process_update(update) telegram.send(result) # end for self.save_state_for_update(update, state) I have a dedicated load_state_for_update and save_state_for_update which I'll probably decorate with @orm.db_session. Therefore I'd have separate sessions to my understanding.

Then you may have special entity UpdateLock (probably with a single instance), and in your transaction before getting or creating State instance you can do ul = UpdateLock.select().for_update().first() <call get_or_create> It should fix concurrency problems, but all State updates will be sequential

Btw, found related issue #131.

For future reference, that's https://github.com/luckydonald/telestate/blob/deploy2/telestate/contrib/pony.py

For future reference, that's https://github.com/luckydonald/telestate/blob/deploy2/telestate/contrib/pony.py

Permanent link to the file telestate/contrib/pony.py mentioned. (?)

Btw, if you wanna get rid of the "user joined" service notifications, you can add @NoServiceBot with deletion permissions

We want to detect spam

Okey :)

Does Pony work on top of pg8000 or only psycopg2?

Hi guys!

A quick simple question - should I manage @db_session or just cover my entire codebase with it?

I've created a Stack Overflow question about this with more details, if needed, though https://stackoverflow.com/questions/55168799/should-i-minimize-the-usage-of-db-session-in-pony-orm

effectively exiting a session (going out of with scope) will commit all your pending transactions to the db

so maybe that informs the question a little bit

depends what you want your writes to look like

ah, I saw your SO question

well - do you want to block execution of the process until the DB write is completed?

Yes, I understand that, that's why I included an explicit commit() in the contra-example

I am asking rather about the performance/resource implications, for example many concurrent db_sessions might deplete the available connections

But I have a strong feeling that I'm prematurely optimizing :D

I guess I don’t understand your question beyond that - I guess general practice of scope minimization would apply, e.g., if you don’t have an explicit reason to be inside a block, don’t be

db_session blocks included?

Does Pony work on top of pg8000 or only psycopg2?

Anyone know this one?

The question beyond that is the db_session might hold a connection to the database busy (not returning it to the connection pool) I applied this principle, but encountered an issue where pytest-ponyorm assumes db_session everywhere while my code doesn't, so I started to consider adding db_session everyhere to my Flask application as described in the Pony documentation: https://docs.ponyorm.org/integration_with_flask.html#integration-with-flask

Ah. Don’t have any advice then, sorry

No problem, thank you very much :)

The question beyond that is the db_session might hold a connection to the database busy (not returning it to the connection pool) I applied this principle, but encountered an issue where pytest-ponyorm assumes db_session everywhere while my code doesn't, so I started to consider adding db_session everyhere to my Flask application as described in the Pony documentation: https://docs.ponyorm.org/integration_with_flask.html#integration-with-flask

In next release of pytest-ponyorm you'll be able to disable the auto added db_session to each test. Comi'g soon

@Jim I'm the author of that PR :))

Lol

At one point after submitting the PR, I've suddenly realized that I might be introducing an anti-pattern to the use of Pony Maybe all/most of the code should be under the db_session anyways :D As in pony orm documentation they just offer a quick solution to cover all of the Flask endpoints automatically, with no encouragement for covering only the general parts that need it