issue_comments

Go for it!

Is this being worked on?

If not, I can send a PR today since I have some code ready that might help add this functionality.

Will work on this issue in the next three weeks if no one else does till then.

Anyways so yes, I agree that these would definitely be more readable than the equivalent operations with isel().

One virtue of tail() is particular is that we can define array.tail(x=0) to consistently return an array with an x dimension of size 0. This requires special case logic with slicing, since slice(None, -0) is the same as slice(None, 0).

I think this is still worth doing. I actually prefer @shoyer's original method names and API. I also think this would be a good first issue.

In order to maintain a list of currently relevant issues, we mark issues as stale after a period of inactivity

If this issue remains relevant, please comment here or remove the stale label; otherwise it will be marked as closed automatically

seems like for, e.g., head, you can pass either a single dimension or multiple ones (e.g., either as **kwargs or a dictionary) and use those as the start dimension.

that said, about naming conventions, i think for tensors the most common convention is definitely slice() (which is implemented as isel). head/tail can be implemented in terms of slice().

e.g.: ds.slice(dim1=3, dim2=(1,4), dim3=(1,None,5)) -- or -- ds.slice({'dim1': 3, 'dim2': (1,4), 'dim3': (1, None, 5)})

head/tail/whatever are easy calls to this and you can have that in the documentation. as a result, people won't get confused because they understand slice.

On Tue, Feb 10, 2015 at 4:14 PM, Stephan Hoyer notifications@github.com wrote:

@ebrevdo https://github.com/ebrevdo yes, I think we probably need to stick to the pandas/numpy convention for the meaning of take.

My inspiration for these names was the head() and tail() methods in pandas, which are quite convenient. But it's not entirely clear how/if these generalize cleanly to N-dimensions. I suppose take_first and take_last could be an improvement over head/tail.

— Reply to this email directly or view it on GitHub https://github.com/xray/xray/issues/319#issuecomment-73812481.

@ebrevdo yes, I think we probably need to stick to the pandas/numpy convention for the meaning of take.

My inspiration for these names was the head() and tail() methods in pandas, which are quite convenient. But it's not entirely clear how/if these generalize cleanly to N-dimensions. I suppose take_first and take_last could be an improvement over head/tail.

Clojure conventions: .take, .take_last get the first n and last n pandas/ndarray conventions: .take([3,4,5]) selects rows 3,4,5.

probably want to be consistent with one of these.

On Tue, Feb 10, 2015 at 3:28 PM, Stephan Hoyer notifications@github.com wrote:

These would be shortcuts for isel/slice syntax: - ds.head(time=5) -> ds.isel(time=slice(5)): select the first five time values - ds.tail(time=5) -> ds.isel(time=slice(-5, None)): select the last five time values - ds.subsample(time=5) -> ds.isel(time=slice(None, None, 5)): select every 5th time value

— Reply to this email directly or view it on GitHub https://github.com/xray/xray/issues/319.