issue_comments

Hi All. I stumble across the same issue trying to convert a 5000 column dataframe to xarray (it was never going to happen...). I found a workaround and I am posting the test below. Hope it helps.

```python import xarray as xr import pandas as pd import numpy as np

xr.version

'0.15.1'

pd.version

'1.0.5'

df = pd.DataFrame(np.random.randn(200, 500))

%%time one = df.to_xarray()

CPU times: user 29.6 s, sys: 60.4 ms, total: 29.6 s
Wall time: 29.7 s

%%time dic={} for name in df.columns: dic.update({name:(['index'],df[name].values)})

two = xr.Dataset(dic, coords={'index': ('index', df.index.values)})

CPU times: user 17.6 ms, sys: 158 µs, total: 17.8 ms
Wall time: 17.8 ms

one.equals(two)

True

```

I am adding here a comment to keep it alive. In fact, this is more complicated than it seems because in combining files with duplicate times one has to choose how to merge i.e keep first, keep last or even a combination of the two.

Thanks @shoyer. Your approach works better (one line) plus is consistent with the xarray-pandas shared paradigm. Unfortunately, I can't spare the time to do the PR right now. I haven't done it before for xarray and it will require some time overhead. Maybe someone with more experience can oblige.

To partially answer my issue, I came up with the following post-processing option

1. get the index of the duplicate coordinate values val,idx = np.unique(arr.time, return_index=True)

2. trim the dataset arr = arr.isel(time=idx)

Maybe this can be integrated somehow...

I am also interested. In terms of the table from @jhamman I am in principle ok with. However, there could be an option to refer to the original attrs in order to provide provenance even on operations like reduce and arithmetic. The idea here is reproducibility and tractability. Maybe an 'origin' attribute?