BLD/TST: cross compat for pa/fp failing on osx & windows

[jreback]

xref #18662
closes #18714

[jreback]

this is not great because xfailing entire test. should figure out what actually is failing and only xfail that part.

[jreback]

I think fastparquet is using deps which cause an upgrade of numpy here.

package plan for installation in environment /Users/travis/miniconda3/envs/pandas:
The following NEW packages will be INSTALLED:
    arrow-cpp:      0.7.1-py35_2               conda-forge
    fastparquet:    0.1.3-py35_0               conda-forge
    feather-format: 0.3.1-py35_1               conda-forge
    llvmlite:       0.21.0-py35h3a66efe_0                 
    numba:          0.36.1-np113py35hd65e1c9_0            
    parquet-cpp:    1.3.0.post-2               conda-forge
    pyarrow:        0.7.1-py35_1               conda-forge
    thrift:         0.10.0-py35_0              conda-forge
The following packages will be UPDATED:
    numpy:          1.10.4-py35_nomkl_2                    [nomkl] --> 1.13.3-py35_nomklh6554414_0 [nomkl]

So easiest is to simply use a later version of numpy on this build. I will update.

[TomAugspurger]

I think this will fail until fastparquet does a release with dask/fastparquet#264

[jreback]

yeah, going to move to 0.22.0

[codecov]

Codecov Report

Merging #18715 into master will decrease coverage by 0.07%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #18715      +/-   ##
==========================================
- Coverage   91.64%   91.57%   -0.08%     
==========================================
  Files         154      153       -1     
  Lines       51430    51307     -123     
==========================================
- Hits        47131    46982     -149     
- Misses       4299     4325      +26

Flag	Coverage Δ
#multiple	89.43% <ø> (-0.06%)	⬇️
#single	40.71% <ø> (-0.24%)	⬇️

Impacted Files	Coverage Δ
pandas/io/gbq.py	25% <0%> (-58.34%)	⬇️
pandas/plotting/_converter.py	64.78% <0%> (-1.74%)	⬇️
pandas/util/_test_decorators.py	93.33% <0%> (-1.22%)	⬇️
pandas/core/reshape/melt.py	96.22% <0%> (-1.03%)	⬇️
pandas/util/testing.py	82.01% <0%> (-0.9%)	⬇️
pandas/core/indexes/category.py	97.2% <0%> (-0.31%)	⬇️
pandas/core/categorical.py	95.64% <0%> (-0.15%)	⬇️
pandas/io/formats/format.py	96.03% <0%> (-0.15%)	⬇️
pandas/core/dtypes/dtypes.py	95.14% <0%> (-0.14%)	⬇️
pandas/core/indexes/numeric.py	97.26% <0%> (-0.11%)	⬇️
... and 16 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b5f1e71...741ff7f. Read the comment docs.

[jreback]

I blew this away accid, so closing for now.