advanced patterns with np.unique and return_index on multi-dimensional arrays in NumPy 1.24.2
I've tried everything I can think of but I'm maintaining legacy code that I'm prototyping a solution and I need some guidance on Quick question that's been bugging me - I'm working with an scenario with `np.unique` when trying to find unique rows in a 2D array, specifically using the `return_index` option....... I expect it to return the indices of the first occurrences of unique rows, but it seems to be giving unexpected results. Here's my code snippet: ```python import numpy as np array = np.array([[1, 2], [1, 2], [3, 4], [5, 6], [3, 4]]) unique_rows, indices = np.unique(array, axis=0, return_index=True) print("Unique Rows:", unique_rows) print("Indices:", indices) ``` I anticipated that the indices would correspond to the first occurrences of each unique row. However, the output I'm getting is: ``` Unique Rows: [[1 2] [3 4] [5 6]] Indices: [0 2 3] ``` While the unique rows are as expected, I noticed that the indices do not reflect the actual first occurrences when printed. Specifically, if I inspect `array[indices]`, it returns: ``` [[1 2] [3 4] [5 6]] ``` While this matches the unique rows, the index `1` for the second unique row is skipped since it corresponds to a repeated row. This behavior is misleading to me since `return_index` implies that all unique items should be represented by their first occurrence index, including duplicates. Is there a better way to achieve what I want? Am I misunderstanding how `np.unique` is supposed to work with multi-dimensional arrays? Any insights would be much appreciated! How would you solve this? I'm working on a service that needs to handle this. For context: I'm using Python on Linux. The stack includes Python and several other technologies. Is there a simpler solution I'm overlooking? I'm coming from a different tech stack and learning Python. Thanks for any help you can provide! Thanks in advance!