Update README to flag issue with optional routes.

[JonRowe]

The original example would also match GET /posts_anything as GET /posts with a format of "_anything" because of the optional format separator, forcing the separator is kind of an ugly example so switch to using a trailing slash which should be recognised either way (I think). It'd be nice to be able to create optional groups e.g. (.:format)?

[kgrz]

I feel the description quite doesn't fit here since it looks like a path fragment rather than an extension. Perhaps this should be reworded?

[kytrinyx]

Additionally, the proposed change would not match "GET /posts", but "GET /post/".

How about something that would be more typically used as a path fragment? Perhaps

get '/posts/?:year?' do
  # matches "GET /posts" and any year "GET /posts/2009", "GET /posts/2010", etc
end

[JonRowe]

get /posts/:year? would be ok but the the /? means the problem would remain, as it'd match posts_2005 as a :year of _2005

[kytrinyx]

Ah, right. Damn, I missed that.

OK, so if we do get '/posts/:year?, then the comment would have to say that it `matches "GET /posts/", which sucks as a route, since people typing things in would tend to leave off the trailing slash.

[JonRowe]

I believe most HTTP servers automatically mismatch the trailing ‘/‘, which is how the /folder/index.html trick works as a /folder url (e.g. with middleman), or at least thats what I was lead to believe when I picked the example :)

[kytrinyx]

This is what I get:

$ cat app.rb 
require 'sinatra'

get '/hello/' do
  'Hello world!'
end

$ ruby app.rb

$ curl -v 'localhost:4567/hello/'
HTTP/1.1 200 OK

$ curl -v 'localhost:4567/hello/'
HTTP/1.1 404 Not Found

[JonRowe]

Ah it's true webbrick doesn't but nginx and apache fronting it would, tricky.

[kytrinyx]

Oh, wow. I didn't realize that nginx/apache would treat these differently. Yeah, that is tricky :/

Thin behaves the same way as webrick. I didn't check puma and unicorn.

[kytrinyx]

@mhutter left a comment in #1023 (comment) which I think is the right way to go for fixing the README. What do you think @JonRowe ?

[mhutter]

Sooo question here, is it really expected for the . in strings to behave like a Regex-.?

[JonRowe]

I actually tried that route at fist @kytrinyx but couldn't make it work, which is why I went down the rewording the readme to avoid the issue entirely route.

[mhutter]

I would actually leave the README as-is - It's only an example.

Or replace by a completely generic scenario, such as

get '/hello/:name?' do
  name = params[:name] || "World!"
  "Hello, #{name}"
end

But then again this example will behave differently depending on the webserver...

[JonRowe]

Thats what this PR is about, changing the example to a more correct if more generic example, just to prevent confusion as in #1023

[JonRowe]

Anything I can do to move this along?

[zzak]

We have to update the translations too, let me see if I can fix this.