openEXO / cloud-kepler

Cloud Kepler is a cloud enabled Kepler Planet searching pipeline
8 stars 3 forks source link

Pandas version 0.14.0 is currently not compatible with the code #35

Open emprice opened 10 years ago

zonca commented 10 years ago

can you paste the error?

zonca commented 10 years ago

tested on my machine with pandas 0.14.0 and it works fine. Not sure what could be the problem. Please reopen the issue with more details.

emprice commented 10 years ago

@zonca Take a look at the output from these two builds on Travis. Note that there are no errors, and Python isn't crashing. The problem is that the output is not consistent between the two, even though the input data is the same. The only change of any significance between these two is that I changed the version of pandas to be 0.13.0 instead of 0.14.0, which can be confirmed by comparing the two commits.

Travis with pandas 0.14.0: https://travis-ci.org/openEXO/cloud-kepler/builds/28409173 Travis with pandas 0.13.0: https://travis-ci.org/openEXO/cloud-kepler/builds/28440080

Both builds "pass" because there is no explicit failure. However, if you look at the output itself, you'll see that every transit event with the higher version of pandas is marked as straddling, and the only way for the code to check for that is if the event is not recovered with the input parameters within tolerance. This also happens every time, which is clearly wrong.

I have demonstrated on at least three machines (the Travis remote environment, my Linux laptop, and my Mac laptop) that this behavior can be reproduced simply by changing the version of pandas.

zonca commented 10 years ago

I see, I've no idea what is wrong, we need to debug the code line by line. I have no time right now, I want to fix the hadoop run first.