Selecting from Index with duplicates raises or returns incorrect results

[phofl]

from dask_expr import from_pandas

df = pd.DataFrame({"a": [1, 2, 3], "bb": 1}, index=["a", "a", "b"])
ddf = from_pandas(df)

ddf.a["b"].compute()

This raises

Traceback (most recent call last):
  File "/Users/patrick/Library/Application Support/JetBrains/PyCharm2023.1/scratches/dask_epr.py", line 11, in <module>
    print(ddf.a["b"].compute())
          ^^^^^^^^^^^^^^^^^^^^
  File "/Users/patrick/mambaforge/envs/dask-expr/lib/python3.11/site-packages/dask/base.py", line 314, in compute
    (result,) = compute(self, traverse=False, **kwargs)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/patrick/mambaforge/envs/dask-expr/lib/python3.11/site-packages/dask/base.py", line 583, in compute
    collections, repack = unpack_collections(*args, traverse=traverse)
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/patrick/mambaforge/envs/dask-expr/lib/python3.11/site-packages/dask/base.py", line 474, in unpack_collections
    repack_dsk[out] = (tuple, [_unpack(i) for i in args])
                              ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/patrick/mambaforge/envs/dask-expr/lib/python3.11/site-packages/dask/base.py", line 474, in <listcomp>
    repack_dsk[out] = (tuple, [_unpack(i) for i in args])
                               ^^^^^^^^^^
  File "/Users/patrick/mambaforge/envs/dask-expr/lib/python3.11/site-packages/dask/base.py", line 435, in _unpack
    if is_dask_collection(expr):
       ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/patrick/mambaforge/envs/dask-expr/lib/python3.11/site-packages/dask/base.py", line 186, in is_dask_collection
    return x.__dask_graph__() is not None
           ^^^^^^^^^^^^^^^^^^
  File "/Users/patrick/PycharmProjects/dask_dev/dask-expr/dask_expr/collection.py", line 86, in __dask_graph__
    out = out.simplify()
          ^^^^^^^^^^^^^^
  File "/Users/patrick/PycharmProjects/dask_dev/dask-expr/dask_expr/expr.py", line 227, in simplify
    out = expr._simplify_down()
          ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/patrick/PycharmProjects/dask_dev/dask-expr/dask_expr/expr.py", line 836, in _simplify_down
    assert a == b
           ^^^^^^
AssertionError

Process finished with exit code 1

while

ddf.a["a"].compute()

returns

a    1
a    2
b    3
Name: a, dtype: int64

[mrocklin]

It seems to me like the underlying issue here is that we're abusing Projection to also mean row selection. Probably we should add some row selection operation and then change __getitem__ to operate differently based on if the meta is series-like or dataframe-like. Thoughts?

[rjzamora]

I'm +1 on distinguishing between column projection and row selection at the Expr level (if possible).

[phofl]

Yep that makes the most sense to me as well!