DEV-9254 remove dataclassY by Boris-Filin · Pull Request #225 · zepben/ewb-sdk-python

Boris-Filin · 2026-06-17T05:03:42Z

Description

We remove the dataclassy from the SDK, migrating to using the actual Python dataclass decorator. (Note: The library has not been physically removed yet as some of the downstream repos rely on it. But it is no longer used anywhere in the SDK). This is motivated by lack of ongoing support for dataclassy, as well as no-one on the team wanting to deal with its mess of magic code.

Since the CIM classes rely on overwriting inits and have quite a few custom properties, default dataclass inits still cannot be used. A class DataclassBase was implemented to manually implement dataclass-like field initialisation.
Every class is now decorated with @zb_dataclass - a normal dataclass decorator with some of its defaults changed (eg slots=True)
Eq and Hash are now implemented in Identifiable and are based on the mrid
Str and Repr are also now in Identifiable
(QoL readability) now every abstract class (non-leaf in the hierarchy) is marked with metaclass=ABCMeta. This makes zero difference programmatically, but makes it clearer which part of the model you are in.
"Optimise imports" in PyCharm was ran on most of the files in the SDK to clean up residual imports from dataclassy, so import structures might be slightly changed. They should now be closer to PEP.

The changes and reasoning are described a lot more verbosely in src/zepben/ewb/dataclass_descriptors/MANIFESTO.md.

Associated tasks

No prerequisites.

Note that the code in this PR heavily anticipates the incoming nullable list change. Specifically, the BackedDescriptor class is currently only used in one place, but will be integral soon. Also, custom inits look a bit bulky with *args and **kwargs forwarding, but they will all be removed soon.

Test Steps

Run the test suite with the new tests; Ideally load a real model to make sure no side-effeects slipped through the cracks.

Checklist

If any of these are not applicable, strikethrough the line ~like this~. Do not delete it!. Let the reviewer decide if you should have done it.

Code

I have performed a self review of my own code (including checking issues raised when creating the PR).
I have added/updated unit tests for these changes, and if not I have explained why they are not necessary.
I have commented my code in any hard-to-understand or hacky areas.
I have handled all new warnings generated by the compiler or IDE.
I have rebased onto the target branch (usually main).

Security

When developing applications, use following guidelines for information security considerations:
* Access to applications should be protected with security keys/tokens or usernames and passwords;
* All sessions are encrypted if possible;
* All application input is sanitised before being acted on (ie SQL statements, etc);
* Log messages, and especially client-facing ones, must be handled securely and must not leak credentials information (internal URLs, passwords, tokens).

I have considered if this change impacts information security and made sure those impacts are handled.

Documentation

I have updated the changelog.
I have updated any documentation required for these changes.

Breaking Changes

I have considered if this is a breaking change and will communicate it with other team members by posting it on the Slack breaking-changes channel.

The internals of classes have changed, potentially leading to unknown side-effects (tests all pass). More importantly, You can no longer use positional arguments except for mrid (MyClass("mrid", 42) bad, MyClass("mrid", thing=42) good).

ryanjcoleman · 2026-06-17T05:03:48Z

Task linked: DEV-9254 Remove dataclassy from Python SDK

chestm007

I've had a quick look - looks good from what ive got through. Github is a flaming pile of... yeah... and is dying trying to render this PR. theres a few nit pick formatting issues, ie newlines, missing line at EOF, and all not being at the top of the file, all pretty minor, but need to be fixed. I'll pull locally and review on anything that isnt github.

charlta · 2026-06-18T05:53:59Z

Converted to draft while formatting fixes are put in place

Boris-Filin · 2026-06-22T02:43:30Z

Reset to Max's formatting changes. Allegedly inline with all of the required standards

chestm007

looks good

chestm007 · 2026-06-24T15:10:58Z

theres some minor formatting issues (3 lines between classes) - I'm not worried, ill adjust the linter to fix those once this goes in.

IMO, just change the type of exception thrown 👇🏿

and merge it in, adios dataclassy, good riddance!

charlta · 2026-06-24T23:41:18Z

Converted to draft until requested changes are fixed

Signed-off-by: Boris Filin <boris.filin@zepben.com>

charlta · 2026-06-24T23:44:08Z

+#  file, You can obtain one at https://mozilla.org/MPL/2.0/.
+
+
+from .dataclass_base import DataclassBase as DataclassBase, zb_dataclass as zb_dataclass


question: Any reason we are importing these into this package? We generally don't import into the containing package, so is this one special somehow, or is this a future route we are taking?

I used the setup in dataclassY as an example. Their code is

# __init__.py from .decorator import dataclass, make_dataclass from .dataclass import DataClass, Internal from .functions import fields, values, as_dict, as_tuple, replace

I guess I landed on the one example that is not representative... In that sense, it is an archaic, not futuristic approach.

With that said, I think for a library that is semantically standalone (not tightly tied to CIM by itself), it might make sense to have this sort of endpoint.

Your call - say the word and I'll change it.

we normally have imports in the top level package, just not the sub packages. i.e. __init__.py in zepben.ewb

charlta · 2026-06-26T01:31:36Z

@@ -12,8 +12,6 @@
 from zepben.ewb.model.cim.iec61970.base.wires.junction import Junction

 # noinspection PyArgumentList


nit: This should be removed also. I am pretty sure it will start applying incorrectly to things undeer it unintentionally.

charlta · 2026-06-29T06:04:57Z


 from datetime import datetime
 from typing import Optional
+from dataclasses import dataclass


nit: excess blank line after removal of old import

charlta · 2026-06-29T06:06:55Z

    from zepben.ewb.model.cim.iec61970.base.wires.power_electronics_connection import PowerElectronicsConnection


+


nit: extra blank line. There are quite a few of these across the files.

charlta · 2026-06-29T06:12:28Z

+    # We are using class names instead of identities because dataclasses create their own class instances
+    assert expected == all_equipment_container_classes, "Should be checking all EquipmentContainer subclasses"
+
+    junction = Junction('j1')


question: Why did this mrid change?

charlta · 2026-06-29T06:18:02Z

    :return: A set of all concrete implementations of `cls` under `package`
    """
-    y = set()
+    y = {}


question: Why did this need to change? There is a lot more code here than before, and no description of why this change was needed.

charlta · 2026-06-29T06:20:02Z

 ## [1.4.0] - UNRELEASED
 ### Breaking Changes
-* None.
+* CIM classes now have to be instantiated with keyword arguments. The only accepted positional argument is mrid (or equivalent identifier)


question: Are these the only breaking changes?

Were all object before hashed just by their mrid? If not, then this is a major/breaking change for adding objects to a hashable collection.

Boris-Filin requested review from charlta and chestm007 and removed request for chestm007 June 17, 2026 05:04

Boris-Filin force-pushed the DEV-9254-remove-dataclassY branch from f53b427 to f331425 Compare June 17, 2026 05:09

chestm007 reviewed Jun 17, 2026

View reviewed changes

charlta marked this pull request as draft June 18, 2026 05:53

Boris-Filin force-pushed the DEV-9254-remove-dataclassY branch 2 times, most recently from 3bc224d to b996101 Compare June 22, 2026 02:35

Boris-Filin marked this pull request as ready for review June 22, 2026 02:42

chestm007 requested changes Jun 24, 2026

View reviewed changes

charlta marked this pull request as draft June 24, 2026 23:41

Boris-Filin force-pushed the DEV-9254-remove-dataclassY branch 10 times, most recently from a8b3c2d to 62b1d78 Compare June 25, 2026 06:19

Remove dataclassY

b2f9274

Signed-off-by: Boris Filin <boris.filin@zepben.com>

Boris-Filin force-pushed the DEV-9254-remove-dataclassY branch from 064460c to b2f9274 Compare June 26, 2026 01:41

charlta requested changes Jun 29, 2026

View reviewed changes

		# file, You can obtain one at https://mozilla.org/MPL/2.0/.


		from .dataclass_base import DataclassBase as DataclassBase, zb_dataclass as zb_dataclass

		@@ -12,8 +12,6 @@
		from zepben.ewb.model.cim.iec61970.base.wires.junction import Junction

		# noinspection PyArgumentList

Uh oh!

Conversation

Boris-Filin commented Jun 17, 2026

Description

Associated tasks

Test Steps

Checklist

Code

Security

Documentation

Breaking Changes

Uh oh!

ryanjcoleman commented Jun 17, 2026

Uh oh!

chestm007 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

charlta commented Jun 18, 2026

Uh oh!

Boris-Filin commented Jun 22, 2026

Uh oh!

chestm007 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

chestm007 commented Jun 24, 2026

Uh oh!

charlta commented Jun 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants