`transform()` fails on large partitioned points with duplicate index

[enric-bazz]

With the recent unreleased spatialdata version, transform() fails when a points element comes from a large parquet file that gets partitioned.

Reproduce

import numpy as np
import pandas as pd
import dask.dataframe as dd
import spatialdata as sd

n = 40_000_000  # number of points

df = pd.DataFrame({
    "x": 100 * np.random.rand(n),
    "y": 100 * np.random.rand(n),
})

df.to_parquet("/tmp/points_file.parquet")

ddf = dd.read_parquet("/tmp/points_file.parquet")

points = sd.models.PointsModel.parse(ddf)

sdata = sd.SpatialData.init_from_elements({"transcripts": points})

sd.transform(sdata, to_coordinate_system='global')

Error

ValueError: cannot reindex on an axis with duplicate labels

Traceback (most recent call last):

  File "<ipython-input-1>", line 1
    sd.transform(sdata, to_coordinate_system='global')

  File ".../spatialdata/_core/operations/transform.py", line 288, in _
    return data.transform_to_coordinate_system(target_coordinate_system=to_coordinate_system)

  File ".../spatialdata/_core/spatialdata.py", line 880, in transform_to_coordinate_system
    transformed = sdata.transform_element_to_coordinate_system(...)

  File ".../spatialdata/_core/spatialdata.py", line 819, in transform_element_to_coordinate_system
    transformed = transform(element, to_coordinate_system=target_coordinate_system, maintain_positioning=maintain_positioning)

  File ".../spatialdata/_core/operations/transform.py", line 466, in _
    new_col = pd.Series(new_ax.data.flatten().compute(), index=transformed.index)

  File ".../pandas/core/indexes/base.py", line 4436, in reindex
    raise ValueError("cannot reindex on an axis with duplicate labels")

Note: This is the root cause of this issue I encountered previously.

Environment

• spatialdata version: 0.7.3a1.dev9+g094b86905
• dask: 2026.1.1
• pandas: 2.3.3

[melonora]

@enric-bazz thanks for reporting, I am currently a bit tight on time but will try to get to it.