add info method to dataset #1176
Conversation
|
Nice! Why pass a buffer into that function though? Why not return a string and the user do what they want with it? |
|
Good question. I tried to merge the output format from |
|
I never noticed that DataFrame.info() doesn't actually have a return value. It does seems strange to write to a buffer taken as an argument but I can see one reason why it sort of makes sense -- otherwise you get quotation marks printed around the returned string. |
|
Yes interesting, I didn't know that either. |
|
Thanks @jhamman ! About the name: what about just |
|
I'm also thinking |
| @@ -147,6 +147,10 @@ Enhancements | |||
| plots (:issue:`897`). See :ref:`plotting.figsize` for more details. | |||
| By `Stephan Hoyer <https://github.com/shoyer>`_ and | |||
| `Fabien Maussion <https://github.com/fmaussion>`_. | |||
| - New :py:meth:`~Dataset.attr_info` method to summarize ``Dataset`` variables | |||
| and attributes. The method produces a stirng output similar to what the | |||
|
Tests are passing on Python 2 now. @shoyer - how do you feel about |
|
I'm pretty happy with |
| lines.append('\t:{k} = {v} ;'.format(k=k, v=v)) | ||
| lines.append('}') | ||
|
|
||
| formatting._put_lines(buf, lines) |
There was a problem hiding this comment.
I would probably just use buf.write(u'\n'.join(lines)) here, which will work as long as you ensure all elements in lines are the same (unicode/str) type.
| lines.append('xarray.Dataset {') | ||
| lines.append('dimensions:') | ||
| for name, size in self.dims.items(): | ||
| lines.append('\t{name} = {size} ;'.format(name=name, size=size)) |
There was a problem hiding this comment.
I think these probably should all be unicode (using u literals), otherwise this will break for non-ASCII characters on Python 2. Take a look at what things look like in formatting.py.
There was a problem hiding this comment.
It would also be good to add a test for attributes with non-ASCII values.
| @@ -12,6 +12,10 @@ | |||
| import dask.array as da | |||
| except ImportError: | |||
| pass | |||
| try: | |||
| from io import StringIO | |||
There was a problem hiding this comment.
we might put this in in pycompat instead
| lines.append(u'\t:{k} = {v} ;'.format(k=k, v=v)) | ||
| lines.append(u'}') | ||
|
|
||
| lines = [ensure_valid_repr(line) for line in lines] |
There was a problem hiding this comment.
I don't think we need this. I think you can write unicode to string buffers, even on Python 2 (at least it works for sys.stdout and StringIO).
There was a problem hiding this comment.
Without it, we get:
UnicodeEncodeError: 'ascii' codec can't encode character u'\xae' in position 363: ordinal not in range(128)
I've always had a tough time getting the string/bytes/unicode stuff straight so I'm open to other ideas here.
There was a problem hiding this comment.
Ah, the issue is that you're using StringIO from cStringIO, which only handles bytes, not unicode. Instead, use StringIO from io, on both Python 2 and 3 (no need for the separate compatibility module even).
|
@shoyer --- green. |
I don't know if this is exactly what we want but here's an idea that emulates
ncdump -h. I'm sure people will have thoughts on the implementation and output so I'll just throw this first cut up and let people discuss.closes #1150
fixes #244
xref: #1044, #820