Low-hanging fruit fixes on bad redirects in dumper

When reviewing https://github.com/mdn/stumptown-renderer/pull/499/files I noticed a bunch of things the repairUri has to do. Every time. Things that cause unnecessary trouble that could be solved once and for all in the dumper.

For example, a lot of _redirects.txt files look like this:

# FROM-URL      TO-URL
/th/docs/en     /
...

and that can become:

# FROM-URL      TO-URL
/th/docs/en     /en-US/docs/
...

Another example is there’s 27 uses of a _ in the locale code in content/files/zh-tw/_redirects.txt. It uses zh_tw instead of zh-TW which is what what’s in VALID_LOCALES.

In content/files/zh-cn/_redirects.txt there are a lot of these:

/zh-CN/docs/AJAX:Getting_Started        /cn/docs/AJAX/Getting_Started

Just like we once used to called en-US just en, I guess at some point we used to call zh-CN just cn. There’s also a lot of redirects to things like /pt/docs/... which needs to be replaced with pt-BR or pt-PT but which one?! Perhaps it can be deduced from the locale in the MySQL wiki_document row.

[peterbe]

Here’s what I think we should do:

1. Wait for https://github.com/mdn/stumptown-renderer/pull/499 to land
2. Work on this issue and test it with the repairUri function to see how much less it had to do in the builder.
3. Revisit how repairUri is used in the builder, in a new issue, to make it much much stricter so that faulty fatfingered PR builds where people edit the _redirects.txt files break the build before it makes things worse.

Good plan?

[peterbe]

But what about things like:

/zh-CN/docs/About_the_MDC       /Project:cn/docs/%E5%85%B3%E4%BA%8E

that shouldn’t have been included in the _redirects.txt file, right?