"Slugify" Usernames

[cycomachead]

When we send emails, and eventually for the social site, the practice of encoding query strings is annoying. Mostly, it's visually unappealing, and it obscures the readability (and type-ability) of URLs and usernames.

My proposal is this:

• Create a new DB column, called "slug"
• Store a slug based on the username
• the username parameter in a URL can be a slug or username.
  • In most cases these are the same
• Add a unique index on the slug column

Creating slugs:

• pick a set of "safe" characters for usernames

• replace unsafe characters with a -.

• If existing usernames would lead to conflicting slugs, we will "hand-edit" them using _ as the separator for the second slug, and in the few cases where there's more than 2 conflicts, I'll get creative, I guess. (But it's a tiny number of users.)

• Lapis provide's a slugify method, but because it removes duplicate -, it leads to more conflicts, and it seems to filter out too many characters.

• The lapis slugify method seems prone to returning an empty string in some cases, which seems wrong.

[cycomachead]

The username slug is also important for discourse which turns spaces into _. Discourse, annoyingly, normalizes unicode characters. Though I'd like not to do that.