Run more complex code test cases using waltz

[killercup]

I've recently finished (what I consider) a usable version of waltz, a tool to extract code blocks from markdown and save them as files. This allows you to do something like literate programming with multiple files. (I wrote a bit about it here.)

I think this might be a nice addition to the rustdoc tests mdbook already supports. The main difference is that you need to specify a file name for the code block like this:

```rust,file=examples/foobar.rs
fn main() {
    println!("Hello world!");
}
```

This syntax is valid CommonMark and works on Github as well.

(You might also need to do some pre-processing to remove the leading of # some lines of code, if you support that kind of thing. waltz' main library function just takes an iterator of pulldown-cmark events, so this should be easy to do.)

[azerupi]

Interesting!
At this time, the test command in mdBook is quite hacky and only useful for rustdoc. But it could become quite useful if we make it more powerful and usable in different situations.

It would indeed be pretty cool to be able to write larger code blocks into chunks and extract them into a file to test. That file could even be loaded back into the book with the playpen feature to have a recap and make it runnable! 👍

How does it currently play with rustdoc? If we use both this tool and rustdoc on the same markdown, are there going to be conflicts with the formats? For example rustdoc interpreting a partial snippet and therefore failing?

I would have to think a little bit about the possibilities, but consider me intrigued 😉

/cc @steveklabnik

[killercup]

It would indeed be pretty cool to be able to write larger code blocks into chunks and extract them into a file to test.

Waltz currently writes m code blocks to m files. I opened an issue for concat though: killercup/waltz#5

[azerupi]

Oh sorry! I assumed that was the main functionality, my bad.
That would be a nice feature then :)

[steveklabnik]

I would have to think a little bit about the possibilities, but consider me intrigued 😉

Same 😄

[killercup]

My original idea was to use this for a guide like Diesel's, where the guide builds up to a working application. The I/O Project chapter might be good candidate as well.

FYI, I published new versions of the lib and binary that concats code blocks with the same file name.