Try to break `graphty`

[lu-pl]

graphty should be severely stress tested, some code paths might need to be more defensive and/or raise confusing Polars-level errors that could be more meaningful.

The idea of this issue is to collect breakage cases with opaque and/or confusing exceptions and handle those cases more gracefully.

[lu-pl]

Update: Basic sad path testing is implemented and merged; more advanced sad tests are still required.

For example, there should be a check for crashing invalid grouping/aggregation settings early. Currently, some invalid configs cause somewhat opaque Polars exceptions.

E.g. the following model

class Model(BaseModel):
    model_config = ConfigDict(group_by="y")
    x: int
    y: list[int]

raises

polars.exceptions.DuplicateError: column with name 'y' has more than one occurrence

Resolved plan until failure:

    ---> FAILED HERE RESOLVING 'select' <---
DF ["x", "y"]; PROJECT */2 COLUMNS

This error occurred with the following context stack:
    [1] 'group_by'
    [2] 'select'

which is perfectly reasonable, but likely non-obvious/non-helpful to library users.

[lu-pl]

Update: Basic sad path testing is implemented and merged; more advanced sad tests are still required.

For example, there should be a check for crashing invalid grouping/aggregation settings early. Currently, some invalid configs cause somewhat opaque Polars exceptions.

E.g. the following model

class Model(BaseModel):
model_config = ConfigDict(group_by="y")
x: int
y: list[int]

raises

polars.exceptions.DuplicateError: column with name 'y' has more than one occurrence

Resolved plan until failure:

    ---> FAILED HERE RESOLVING 'select' <---
DF ["x", "y"]; PROJECT */2 COLUMNS

This error occurred with the following context stack:
    [1] 'group_by'
    [2] 'select'

which is perfectly reasonable, but likely non-obvious/non-helpful to library users.

Note: This can be caught in the else path of get_group_by_value by checking the group_by value against the model projection.

[lu-pl]

Another case:

Here the AliasChoice cannot be resolved from the data project.

data = [
    {"a": 1, "b": 3},
    {"a": 1, "b": 4},
    {"a": 2, "b": 6},
]

class Model(BaseModel):
    x: int = Field(validation_alias=AliasChoices("foo", "bar", "x"))
    y: int = Field(validation_alias=AliasChoices("foo", "bar", "b"))

This raises

ValueError: Unable to resolve AliasChoice for field 'x' of model 'Model1'.
None of computed aliases '['foo', 'bar', 'x']' in projection '{'a', 'b'}'.

from AliasMap_generate_alias_map.

Generailly, this is fine, I think, merely the raise sight might be a possible matter of discussion.

[lu-pl]

Another one:

class Model(BaseModel):
    model_config = ConfigDict(group_by="dne")

    x: int = Field(validation_alias=AliasChoices("foo", "bar", "a"))
    y: list[int] = Field(validation_alias=AliasChoices("foo", "bar", "b"))

This raises KeyError: 'dne' from a __getitem__ call in AliasMap.