Add quick start readme

[liufuyang]

As discussed here #778, I think this adding a README like this here can make beginners start with this lib much easier.

All the code in the doc is runnable in https://play.integer32.com/ playground :)

Fixes #778

[quietlychris]

I'm not a contributor to this crate, but having used ndarray in a limited capacity on some projects, everything in here looks good to me. I also found that it did a really good job at explaining some of the particular rough edges for beginners (like compilation errors due to type inference issues when calling Array::zeros()) that otherwise take some time to figure out.

[liufuyang]

@quietlychris Thank you for the input, I am glad it can help. And this is just a starting point of a doc like this. I think if we have something like it then we can slowly make it even better, if see some frequent asked questions.

For now I feel a doc like this, could help attract new users, and combining other two docs it will probably be good enough for most cases when one wants to actually using this package :)

[LukeMathWalker]

Suggested change

	More info checkout https://docs.rs/ndarray/latest/ndarray/struct.ArrayBase.html#arithmetic-operations
	For more info checkout https://docs.rs/ndarray/latest/ndarray/struct.ArrayBase.html#arithmetic-operations

[LukeMathWalker]

This can be confusing: clone() will perform a deep copy of the array elements only if you are cloning an owned array (i.e. Array).
Cloning an ArrayView does not clone the underlying elements - you are just cloning the view reference (as it happens in Rust when cloning a & reference).

[liufuyang]

Aha, I see. Let me add your explanation here then. Thank you very much :)

[bluss]

I'd avoid the term "Deep Copy" altogether, since it's a bit confusing in Rust terms. Proposed heading is Copying Arrays

[LukeMathWalker]

I left a bunch of style-related suggestions and a comment on wording, but overall it looks really good @liufuyang. Thanks for working on this!

It would be ideal to include a link to this quickstart guide in the README and in the landing page of the documentation, to make it discoverable.

[liufuyang]

Thank you very much for the good input. I will see whether I can find time this weekend to address your inputs :)

[liufuyang]

@LukeMathWalker Thank you for the corrections. I have updated as you suggested 👍

[liufuyang]

I added a few more lines showing usage of from_elem, as discussion mentioned here #778

Perhaps we can merge this soon? @jturner314 @LukeMathWalker :) ?

[bluss]

both links can link to the latest version - use latest, * or similar for the version?

[bluss]

Deep copy is not strictly true. When using clone it's always the shallowest copy that is allowed by ownership rules. That means that an array of Rc<T> elements, will .clone() each element (i.e increase its refcount, not "deep copy")

[bluss]

You could explain that .clone() also copies the array's elements, it creates an independently owned array of the same type. Just another nice way the type system can tell you clearly what's happening.

[bluss]

Can we say just robust code, not talk down on numpy here - we don't have to make this comparison

[bluss]

Thanks for working on this

[liufuyang]

@bluss Thank you for the input. I have updated the part you commented. Please take a look again? Thank you.

[bluss]

Great, we should merge this. It looks like you'd want to rebase/squash this before merge. Do you? Otherwise we can just merge it in

[bluss]

I have edited the PR description to put in the literal Fixes #778, so that the issue is closed when this is merged. link with keyword

[liufuyang]

@bluss Thank you. I think you can just use that Squash and merge button? :)

Something like this, then there is only a single commit for this in master branch

[bluss]

Ok, let's do that. I usually prefer to keep history if it's neat, but it should even keep all the authors. Let's go. Thanks again. (Otherwise a manual squash and cleanup is what I usually do.)

[bluss]

I want to add that the preferred merge strategy of this repo is Merge commit, then squash. Not using rebase.