Migrate to 2018 edition

[DCjanus]

There was function get_with_timeout, re-create a client each time this function is called, and "timeout" always 'default', I merged these two functions and reuse reqwest::Client to reuse TCP connection.

Other mirror change:

• move to 2018 edition
• work with rustfmt

Recommended review way: commit by commit.

[ordian]

Thanks! I have one concern wrt url handling (inline comment). Also could you merge with master to fix CI tests?

[DCjanus]

The purpose of this PR is to reduce the time of cargo upgrade, because of Chinese network block, every cargo upgrade takes a lot of time.

Maybe we should query local registry index rather than query HTTP API? I'd like to handle this.

[ordian]

The purpose of this PR is to reduce the time of cargo upgrade

I see. Does it help though, we still make the same amount of requests?

Maybe we should query local registry index rather than query HTTP API?

I think we should eventually use cargo offline mode

[DCjanus]

Does it help though, we still make the same amount of requests?

Less connection with less TCP handshake, that's do helpful, especially if the network is bad. But, never mind, you shouldn't spend too much effort on my special situation.

I think we should eventually use cargo offline mode

I am not very familiar with Cargo's offline mode. My current idea is to use the SourceConfigMap#load in the cargo crate to get the Source and use the query method to get all versions of a crate.

Of cause, update index in the begin.

But temporarily I don't known how to support the alternate registry

[ordian]

@DCjanus could you open a separate issue for querying local registry index/offline mode? Note, that we don't want to depend on cargo as a lib (because it's huge), but instead the plan is to eventually integrate cargo-edit into cargo (rust-lang/cargo#5586).

[DCjanus]

Thanks for your help!

[ordian]

bors r+

[bors]

Build succeeded

• continuous-integration/travis-ci/push
• continuous-integration/appveyor/branch