Generation for memoized calls does not work for inherited methods and does not support dynamic binding

[lars-reimann]

Describe the bug

Running the pipeline below produces the following logs:

2024-04-08 20:12:18.772 [info] Launching Pipeline (653341ab-09b7-438d-950b-935b2ca7d4ad): file:///c%3A/Users/Lars/OneDrive/Desktop/test/test.sdstest - myPipeline
2024-04-08 20:12:18.773 [info] [Runner] Sending message to python server: {"type":"program","id":"653341ab-09b7-438d-950b-935b2ca7d4ad","data":{"code":{"test":{"gen_test":"# Imports ----------------------------------------------------------------------\r\n\r\nimport safeds_runner\r\nfrom safeds.data.tabular.containers import Table\r\nfrom safeds.data.tabular.transformation import OneHotEncoder, TableTransformer\r\n\r\n# Pipelines --------------------------------------------------------------------\r\n\r\ndef myPipeline():\r\n    titanic = safeds_runner.memoized_call(\"safeds.data.tabular.containers.Table.from_csv_file\", Table.from_csv_file, ['C:/Users/Lars/OneDrive/Desktop/test/titanic.csv'], [safeds_runner.file_mtime('C:/Users/Lars/OneDrive/Desktop/test/titanic.csv')])\r\n    safeds_runner.save_placeholder('titanic', titanic)\r\n    encoder = safeds_runner.memoized_call(\"safeds.data.tabular.transformation.OneHotEncoder.fit\", OneHotEncoder.fit, [safeds_runner.memoized_call(\"safeds.data.tabular.transformation.OneHotEncoder\", OneHotEncoder, [], []), titanic, ['sex']], [])\r\n    safeds_runner.save_placeholder('encoder', encoder)\r\n    transformedTable = safeds_runner.memoized_call(\"safeds.data.tabular.transformation.TableTransformer.transform\", TableTransformer.transform, [encoder, titanic], [])\r\n    safeds_runner.save_placeholder('transformedTable', transformedTable)\r\n","gen_test_myPipeline":"from .gen_test import myPipeline\r\n\r\nif __name__ == '__main__':\r\n    myPipeline()\r\n"}},"main":{"modulepath":"test","module":"test","pipeline":"myPipeline"}}}
2024-04-08 20:12:21.994 [info] [Runner] Message received: '{"type": "placeholder_type", "id": "653341ab-09b7-438d-950b-935b2ca7d4ad", "data": {"name": "titanic", "type": "Table"}}'
2024-04-08 20:12:21.994 [info] [Runner] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): titanic of type Table
2024-04-08 20:12:21.994 [info] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): titanic of type Table
2024-04-08 20:12:21.994 [info] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): titanic of type Table
2024-04-08 20:12:21.999 [info] [Runner] Message received: '{"type": "placeholder_type", "id": "653341ab-09b7-438d-950b-935b2ca7d4ad", "data": {"name": "encoder", "type": "OneHotEncoder"}}'
2024-04-08 20:12:21.999 [info] [Runner] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): encoder of type OneHotEncoder
2024-04-08 20:12:21.999 [info] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): encoder of type OneHotEncoder
2024-04-08 20:12:21.999 [info] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): encoder of type OneHotEncoder
2024-04-08 20:12:22.002 [info] [Runner] Message received: '{"type": "placeholder_type", "id": "653341ab-09b7-438d-950b-935b2ca7d4ad", "data": {"name": "transformedTable", "type": "Null"}}'
2024-04-08 20:12:22.002 [info] [Runner] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): transformedTable of type Null
2024-04-08 20:12:22.002 [info] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): transformedTable of type Null
2024-04-08 20:12:22.002 [info] Placeholder was calculated (653341ab-09b7-438d-950b-935b2ca7d4ad): transformedTable of type Null
2024-04-08 20:12:22.005 [info] [Runner] Message received: '{"type": "runtime_progress", "id": "653341ab-09b7-438d-950b-935b2ca7d4ad", "data": "done"}'
2024-04-08 20:12:22.005 [info] [Runner] Runner-Progress (653341ab-09b7-438d-950b-935b2ca7d4ad): done
2024-04-08 20:12:22.005 [info] Runner-Progress (653341ab-09b7-438d-950b-935b2ca7d4ad): done
2024-04-08 20:12:22.005 [info] Runner-Progress (653341ab-09b7-438d-950b-935b2ca7d4ad): done

Particularly note the part {"name": "transformedTable", "type": "Null"}. transformedTable should instead be a Table. This happens because the assignment of the placeholder is translated to

transformedTable = safeds_runner.memoized_call("safeds.data.tabular.transformation.TableTransformer.transform", TableTransformer.transform, [encoder, titanic], [])

There are two issues with this:

1. The memoization key is wrong. Because of this, any invocation of transform on any subclass of TableTransformer is treated the same.
2. The callable TableTransformer.transform is taken from the abstract superclass. It returns None since it has no implementation.

To Reproduce

Run this code (replace the path to the CSV file):

package test

pipeline myPipeline {
    val titanic = Table.fromCsvFile("C:/Users/Lars/OneDrive/Desktop/test/titanic.csv");
    val encoder = OneHotEncoder().fit(titanic, ["sex"]);
    val transformedTable = encoder.transform(titanic);
}

Expected behavior

In the DSL, we only know the static types of objects. We, thus, cannot compute memoization keys for instance methods at this point. This must be deferred to the runtime when memoized_call is evaluated. Likewise, we should let Python handle the dynamic binding.

For global functions and static methods, we can compute the memoization keys in the DSL, but it seems like we should fully move this responsibility to the runner if possible.

We could, for example, introduce a new runner method memoized_dynamic_call, which gets

1. the receiver,
2. the name of the method to call,
3. the parameters,
4. the hidden parameter.

Based on the dynamic type of the receiver and the method name, it can then derive its memoization key.

We could then rename memoized_call to memoized_static_call to be explicit.

Screenshots (optional)

No response

Additional Context (optional)

No response

[lars-reimann]

@WinPlay02 I'd say this one has a little higher priority than #1012, since there is no workaround. But it also seems more involved than #1012.

[lars-reimann]

🎉 This issue has been resolved in version 0.11.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• npm package (@latest dist-tag)
• v0.11.0
• GitHub release

Your semantic-release bot 📦🚀