BUG: Fix unwrap function for object arrays with large numbers. #28911
Conversation
Implemented special handling for object arrays in np.unwrap to correctly unwrap sequences containing very large numbers (e.g., >10^50). The fix properly implements period-based unwrapping logic for object arrays, ensuring consistent increments that respect the specified period parameter. Fixes numpy#27686.
| dd = diff(p, axis=axis) | ||
|
|
||
| if p.dtype == np.object_: | ||
| initial_value = p[0] |
There was a problem hiding this comment.
What if p is empty? Or p is a 2D array?
There was a problem hiding this comment.
the function should work properly if p is empty or a 2D array. my changes only affected the case where p is an object
There was a problem hiding this comment.
What happens for np.unwrap(np.array([], dtype=object))?
There was a problem hiding this comment.
you're right. it gives me IndexError: index 0 is out of bounds for axis 0 with size 0. I'll work on a solution for this
| initial_value = p[0] | ||
| unwrapped = [initial_value] | ||
| for val in p[1:]: | ||
| unwrapped.append(unwrapped[-1] + 1) |
There was a problem hiding this comment.
Appending in a for loop will be much slower than the current implementation for larger arrays.
Do you know the root cause of the problem for bigints? Perhaps we can modify the algorithm at that point. I checked that the first step (e.g. np.diff(p, axis=axis)) works fine
Implemented special handling for object arrays in np.unwrap to correctly unwrap sequences containing very large numbers (e.g., >10^50). The fix properly implements period-based unwrapping logic for object arrays, ensuring consistent increments that respect the specified period parameter. Fixes #27686.