COMPAT, TST: allow numpy array comparisons with complex dtypes #13392
Conversation
|
Didn't really think I needed a |
|
this needs perf testing, not sure how much that path is hit (so I don't think its an issue). but pls run them. |
jreback
added
Testing
| return np.array_equal(left, right) | ||
| except AttributeError: | ||
| # see gh-13388 | ||
| if not is_dtype_equal(left.dtype, right.dtype): |
There was a problem hiding this comment.
add a comment here of what this is doing / catching
|
@jreback : Will quickly run benchmarks, but as this is just a compatibility issue (bug exists until |
Current coverage is 84.22%@@ master #13392 diff @@
==========================================
Files 138 138
Lines 50756 50763 +7
Methods 0 0
Messages 0 0
Branches 0 0
==========================================
+ Hits 42756 42757 +1
- Misses 8000 8006 +6
Partials 0 0
|
Traces back to bug in NumPy v1.7.1 in which the 'array_equivalent' method could not compare NumPy arrays with complicated dtypes. As pandas relies on this function to check NumPy array equality during testing, this commit adds a fallback method for doing so. Closes pandas-devgh-13388. [ci skip]
gfyoung
force-pushed
the
np-array-compare-compat
branch
from
230616e to
fe87955
Compare
|
@jreback : Travis is passing on all machines again. Ready to merge if there are no other concerns. |
|
benchmarks ok? |
|
No obvious changes in the benchmarks AFAICT and as expected. |
|
ty |
Traces back to bug in NumPy
v1.7.1in which thearray_equivalentmethod could not compare NumPy arrays with complicated dtypes. Aspandasrelies on this function to check NumPy array equality during testing, this commit adds a fallback method for doing so.Closes #13388.