########################################
*views* and *copies*
########################################

.. epigraph::

  Sometimes, even after five years using NumPy,
  I have to go with my gut feeling to predict if an operation will trigger a copy in the underlying data.
  **-- NumPy developer**

When NumPy returns an array, it may be the same array (a `view`) or a
new one (a `copy`).

There's nothing obvious to tell which you're getting, but the difference
can change how the code behaves. Returning a view is faster because there's no copying step. But a
view is the gift that keeps on giving -- if either side changes a cell, it silently changes for both.

- `Are there functions guaranteed to return one or the other?`_
- `Is there a rule to figure out which one Function X will return?`_
- `Can I turn a view into a copy?`_
- `Can I turn a copy into a view?`_
- `When can I ignore the difference?`_
- `Is there a way for code to tell me which one I've gotten?`_
- `Are there slick uses for one or the other?`_
- `What's the deal with Pandas' SettingWithCopyWarning?`_
- `Where can I find a more detailed explanation?`_