Reduce warnings in pytest output #10168
Conversation
bdice
added
improvement
Codecov Report
@@ Coverage Diff @@
## branch-22.04 #10168 +/- ##
================================================
+ Coverage 10.42% 10.47% +0.05%
================================================
Files 119 122 +3
Lines 20603 20487 -116
================================================
- Hits 2148 2147 -1
+ Misses 18455 18340 -115
Continue to review full report at Codecov.
|
| @@ -29,8 +29,37 @@ | |||
|
|
|||
| @pytest.fixture(params=dtypes, ids=dtypes) | |||
| def pandas_input(request): | |||
There was a problem hiding this comment.
This fixture was throwing a lot of warnings about unsafe casting -- in particular, things like creating a Series with dtype int8 from data that could exceed 127 (the max random value was 1000). Instead, I implemented a function that creates random values based on the type.
|
pytest output on Warnings on this branch (as of 99006ce, results from a local build): Looks like this PR cuts the warning count from 19524 warnings to 3041. That's a good start! |
Co-authored-by: Vyas Ramasubramani <[email protected]>
| def random_ints(dtype, size): | ||
| dtype_min = np.iinfo(dtype).min | ||
| dtype_max = np.iinfo(dtype).max | ||
| return rng.integers(dtype_min, dtype_max, size=size, dtype=dtype) | ||
|
|
||
| try: | ||
| dtype = np.dtype(dtype) | ||
| except TypeError: | ||
| if dtype == "category": | ||
| data = random_ints(np.int64, size) | ||
| else: | ||
| raise | ||
| else: | ||
| if dtype.kind == "b": | ||
| data = rng.choice([False, True], size=size) | ||
| elif dtype.kind in ("m", "M"): | ||
| # datetime or timedelta | ||
| data = random_ints(np.int64, size) | ||
| elif dtype.kind == "U": | ||
| # Unicode strings of integers like "12345" | ||
| data = random_ints(np.int64, size).astype(dtype.str) | ||
| elif dtype.kind == "f": | ||
| # floats in [0.0, 1.0) | ||
| data = rng.random(size=size, dtype=dtype) | ||
| else: | ||
| data = random_ints(dtype, size) |
There was a problem hiding this comment.
Seems like this could be more generally useful. Which also makes me wonder... does something like this already exist in cudf, and can we re-use that?
There was a problem hiding this comment.
That's definitely a good question. I plan to investigate that further and strive for de-duplication as I work through some more warnings. Lots of the warnings are affected by dtypes!
|
@gpucibot merge |
This PR reduces some warnings in pytest output.
The pyarrow warning that I silenced was occurring thousands of times during our tests but should be fixed by #9686, so I marked it with a TODO.
I just tackled a few files for now, aiming for around 100 LOC changed for ideal review size.