In the meantime, If you really need a solution right now, I’ve made https://github.com/luckydonald/pony_up which might help you. It is more of a manual approach of writing migrations, but executes them automatically, and thus it is still better then executing SQL queries yourself.

In the meantime, If you really need a solution right now, I’ve made https://github.com/luckydonald/pony_up which might help you. It is more of a manual approach of writing migrations, but executes them automatically, and thus it is still better then executing SQL queries yourself.

Permanent link to the luckydonald/pony_up project you mentioned. (?)

Hi! I have been working recently with PonyORM and it is awesome! 🙂 I was wondering, how is the GraphQL development going? I have read some posts on Github, Stackoverflow, about plans of its integration (I think the oldest I found was of 2015), but couldn't find anything in the documentation

Hi Carlos! We don't develop GraphQL integration right now

Maybe we'll return to it later

got it! Thanks, Alexander! 👍🏼

Hi guys, just meeting pony. A question about Pony and ORM's in general. Can Pony help me to generate the query detached from the database? I mean I need to generate complex sql agg query for Apache Spark but I do not want to concatentate SQL strings. Any help_

existing_review_data = {k: v for k, v in existing_review.to_dict().items() if (logic) }

existing_review is a pony object

ProgrammingError: can't adapt type 'TrackedDict'

I am eventually using that data to construct a new pony object

is TrackedDict the object produced by to_dict() ? Should I explicitly call dict() on it to fix this?

From looking at pony source, I think I should call to_dict().get_untracked() ?

Who raises ProgrammingError?

pony/orm/dbapiprovider.py - wrap_dbapi_exceptions

Can you pm me full traceback?

yep

And what is existing_review_data?

do you mean existing_review ?

No, I mean existing_review_data. Like, are you trying to send this object to database somehow?

It is a dict which gets modified then passed to the constructor of ReviewEdit which is a pony model, as **kwargs

Does ReviewEdit have JSON attr?

Sorry, that’s wrong, it is used to compare against another similar dict which is used as kwargs

Does ReviewEdit have JSON attr?

yes

product_attributes = Optional(Json)

Hi Matthew! TrackedDict is returned for json attrs. Its purpose is to notify object when the value of attribute (or some of its subcomponent) has changed, so object will know it should be saved to database on commit. TrackedDict is instance of a normal dict, so if a library which try to "adapt" TrackedDict value (whatever it means) should work with TrackedDict instance without problems. But probably psycopg2 library uses a check like type(value) is dict instead of isinstance(value, dict). But I'm wondering why you are trying to pass a dict as a parameter value to SQL query, it looks strange. Yes, you need to call get_untracked, but not on obj.to_dict(), but on obj.some_json_attr_value. Maybe you need to use some helper like def adapt_dict(d): result = {} for k, v in d.items(): if isinstance(v, TrackedDict): v = v.get_untracked() result[k] = v retrn result

thank you, that makes sense, I will try it

Maybe obj.to_dict() should always do .get_untracked() to get copy of JSON value which is ubinded from the object

Hi guys, just meeting pony. A question about Pony and ORM's in general. Can Pony help me to generate the query detached from the database? I mean I need to generate complex sql agg query for Apache Spark but I do not want to concatentate SQL strings. Any help_

Hi Argenis! Database object has last_sql property to obtain last sql query text which was executed, but currently there is no method for getting sql without executing a query. I think we should add it

existing_review_data = {k: v for k, v in existing_review.to_dict().items() if k not in forbidden_keys} existing_review_data = adapt_pony_dict(existing_review_data)

The exception is thrown on tthe first line, before the function you provided can run

hmm

existing_review_data = {k: v for k, v in adapt_pony_dict(existing_review.to_dict()).items() if k not in forbidden_keys}

This also fails

This line throws the exception:

x = existing_review.to_dict()

By the way, the exception is only being thrown in a tiny minority of cases

The only JSON field in existing_review value is {}

Looks like a bug in pony, I'll look into it

When I try existing_review.to_json() I get:

PermissionError: The current user None which belongs to groups ['anybody'] has no rights to see the object Review[15498384] on the frontend

Don't use to_json, it was experimental undocumented method and need to be rewritten

ah!

What's your dbms?

Postgres

If in ipython I load the Review directly then call to_dict(), it works

The problem is not in .to_dict(). .to_dict() calls flush(), and the error happens inside flush

JSON attribute content represented in two ways. For database it is a string with serialized JSON. For Python user it is a dict. Pony automatically converts dict to serialized string when it saves objects to database. But in your case for some reason the dict was passed to database without conversion to string. I'll investigate it

Thank you!

It would be good if you discover reproducible example of how it happens

It seems to only happen with a minority of jobs of that type

but it consistently happens for at least that one object

I suspect this object was previously pickled/unpickled

I don’t think it has been

Maybe you did something unusual with this object before calling to_dict

The only thing I do is set a value on the object if thatt value is None

if x.y is None: x.y = rue

and set another attribute to None

#if existing_review.review_verified is None: # existing_review.review_verified = review_verified #existing_review.deleted = None

when commented out, it works

#existing_review.deleted = None

this is the code that when commentted out makes it work

I think the lines that you’ve commented out just manifests the error. They modify the object, and then to_dict internally calls flush() trying to save the modified object, and flush fails to save corrupted object which leads to exception. But object state corruption happens before the lines that you’ve commented out.

The only line before those lines which involve that object is:

existing_review = models.Review.get(amazon_review_id=amazon_review_id, product=product)

Does Pony ever issue an UPDATE SQL query when doing Entity.get() ?

I know that sounds strange, but that seems to be happening for me

and then failing

If you have some unsaved changes, Pony will save them before issuing SELECT to be sure that SELECT sees these changes

Ah, so if I end up “get” ing the same object twice, and the object was modified the first time, that could explain it

Do you observe some incorrect behavior?

I am ultimately getting this error:

ProgrammingError: operator does not exist: json = unknown LINE 13: AND "facebook_tasks" = '["VIEW_MONETIZATION_INSIGHTS", "AN...

So the right part is not casted to ::json. Ok, It seems I need to fix it

Do you need some more context from me?

I think no

🙂

So I was processing the same object twice due to a bug in my code. Now the JSON error no longer appears, but I guess it is still a bug?

Yes, I think it is a bug and I'll fix it

Can I somehow use pony with python3.8?