David Tolnay

The sajson library has an interesting [single allocation mode](https://github.com/chadaustin/sajson#single) where it pre-allocates a buffer of size one word per byte in the input document. Justified in [this blog post](https://chadaustin.me/2013/01/sajson-why-the-parse-tree-is-big-enough/). Then...

Unit test with an io::Write that returns Ok(0)

2

This would have caught #178. Let's run all unit tests also against a writer that does not always succeed.

https://docs.rs/serde-xml-rs/0.2.1/serde_xml_rs/ needs a more helpful landing page.

I thought I would pass along some feedback from https://users.rust-lang.org/t/rust-serialization-to-xml/12212: > Serde documentation about xml sucks … :-/ I think they may have been comparing [`serde-xml-rs`](https://docs.rs/serde-xml-rs/0.2.0/serde_xml_rs/) against [`serde_json`](https://docs.rs/serde_json/1.0.2/serde_json/).

Where possible.

> Traverse DOM and count the number of JSON types, total length of string, and total numbers of elements/members in array/objects.

Similar to [memorystat.h](https://github.com/miloyip/nativejson-benchmark/blob/bfbb28f22f40a37bf4f81072abc89567819ac530/src/memorystat.h).

I came here to suggest/ask for exactly this behavior, but happened to notice it's already been landed in master (quite a while ago) in https://github.com/nvzqz/static-assertions-rs/commit/f0b707eb491bbd213472d1e6f2fe3b1dc05e53dc. The most recent release to...