wonky test: TestGoogle:test_get_multi1

[jreback]

https://travis-ci.org/pydata/pandas/jobs/9095624

cc @cpcloud @jtratner

[cpcloud]

should we just remove all equality testing in these tests then?

[jreback]

it looks like a legit test, but if one of the series say has NaN in it (because google decided to not give you data for some reason) this will fail....a possible outcome

I guess skip

[cpcloud]

skip if not equal? that makes the test sort of invalid...raise on any are nans maybe? what exactly were you thinking?

[jreback]

this test shouldn't fail....what I am saying is it was prob a fluke and maybe some/all nan's got returned in one of the series, more of a remote data error...maybe just leave it...I haven't seen before, you?

[cpcloud]

nah i haven't seen this exact error, but similar eq comp failures. maybe only check equality if no nans?

[cpcloud]

personally, i think all the comparisons should be removed (i.e., just check to see that they succeed in the call or throw RemoteDataError)...but that might be a little too drastic.

[jreback]

the problem here is that something was returned just not clear exactly what (could be nan, partially invalid, 0, who knows)

maybe just skip for now and mark for review in 0.13

[cpcloud]

ok marking for 0.13

[cpcloud]

similar yahoo failures

https://travis-ci.org/jtratner/pandas/jobs/9314450

cc @jtratner

[jtratner]

@cpcloud weird thought: what if the network decorator defaulted to running a test a second time if it fails with any non-IO error? (at least that last travis issue looks like it could have come from a bad request...).

[cpcloud]

@jtratner good idea...want to submit a pr for that 😄 ?

[jtratner]

@cpcloud sure, I can try setting this up this weekend.

[jreback]

@cpcloud these look all fixed?

[cpcloud]

closed by #5073