How do I safely join relative url segments? - ruby

Question

I'm trying to find a robust method of joining partial url path segments together. Is there a quick way to do this?
I tried the following:
puts URI::join('resource/', '/edit', '12?option=test')
I expect:
resource/edit/12?option=test
But I get the error:
`merge': both URI are relative (URI::BadURIError)
I have used File.join() in the past for this but something does not seem right about using the file library for urls.

Answer 1

URI's api is not neccearily great.
URI::join will work only if the first one starts out as an absolute uri with protocol, and the later ones are relative in the right ways... except I try to do that and can't even get that to work.
This at least doesn't error, but why is it skipping the middle component?
URI::join('http://somewhere.com/resource', './edit', '12?option=test')
I think maybe URI just kind of sucks. It lacks significant api on instances, such as an instance #join or method to evaluate relative to a base uri, that you'd expect. It's just kinda crappy.
I think you're going to have to write it yourself. Or just use File.join and other File path methods, after testing all the edge cases you can think of to make sure it does what you want/expect.
edit 9 Dec 2016 I figured out the addressable gem does it very nicely.
base = Addressable::URI.parse("http://example.com")
base + "foo.html"
# => #<Addressable::URI:0x3ff9964aabe4 URI:http://example.com/foo.html>
base = Addressable::URI.parse("http://example.com/path/to/file.html")
base + "relative_file.xml"
# => #<Addressable::URI:0x3ff99648bc80 URI:http://example.com/path/to/relative_file.xml>
base = Addressable::URI.parse("https://example.com/path")
base + "//newhost/somewhere.jpg"
# => #<Addressable::URI:0x3ff9960c9ebc URI:https://newhost/somewhere.jpg>
base = Addressable::URI.parse("http://example.com/path/subpath/file.html")
base + "../up-one-level.html"
=> #<Addressable::URI:0x3fe13ec5e928 URI:http://example.com/path/up-one-level.html>

Answer 2

Have uri as URI::Generic or subclass of thereof
uri.path += '/123'
Enjoy!
06/25/2016 UPDATE for skeptical folk
require 'uri'
uri = URI('http://ioffe.net/boris')
uri.path += '/123'
p uri
Outputs
<URI::HTTP:0x2341a58 URL:http://ioffe.net/boris/123>
Run me

Answer 3

The problem is that resource/ is relative to the current directory, but /edit refers to the top level directory due to the leading slash. It's impossible to join the two directories without already knowing for certain that edit contains resource.
If you're looking for purely string operations, simply remove the leading or trailing slashes from all parts, then join them with / as the glue.

Answer 4

The way to do it using URI.join is:
URI.join('http://example.com', '/foo/', 'bar')
Pay attention to the trailing slashes. You can find the complete documentation here:
http://www.ruby-doc.org/stdlib-1.9.3/libdoc/uri/rdoc/URI.html#method-c-join

Answer 5

As you noticed, URI::join won't combine paths with repeated slashes, so it doesn't fit the part.
Turns out it doesn't require a lot of Ruby code to achieve this:
module GluePath
def self.join(*paths, separator: '/')
paths = paths.compact.reject(&:empty?)
last = paths.length - 1
paths.each_with_index.map { |path, index|
_expand(path, index, last, separator)
}.join
end
def self._expand(path, current, last, separator)
if path.start_with?(separator) && current != 0
path = path[1..-1]
end
unless path.end_with?(separator) || current == last
path = [path, separator]
end
path
end
end
The algorithm takes care of consecutive slashes, preserves start and end slashes, and ignores nil and empty strings.
puts GluePath::join('resource/', '/edit', '12?option=test')
outputs
resource/edit/12?option=test

Answer 6

Use this code:
File.join('resource/', '/edit', '12?option=test').
gsub(File::SEPARATOR, '/').
sub(/^\//, '')
# => resource/edit/12?option=test
example with empty strings:
File.join('', '/edit', '12?option=test').
gsub(File::SEPARATOR, '/').
sub(/^\//, '')
# => edit/12?option=test
Or use this if possible to use segments like resource/, edit/, 12?option=test and where http: is only a placeholder to get a valid URI. This works for me.
URI.
join('http:', 'resource/', 'edit/', '12?option=test').
path.
sub(/^\//, '')
# => "resource/edit/12"

Answer 7

A not optimized solution. Note that it doesn't take query params into account. It only handles paths.
class URL
def self.join(*str)
str.map { |path|
new_path = path
# Check the first character
if path[0] == "/"
new_path = new_path[1..-1]
end
# Check the last character
if path[-1] != "/"
new_path += "/"
end
new_path
}.join
end
end

Answer 8

This question is nearly a decade old, yet it seems that there is no perfect solution posted.
A handful of posted answers fail to handle multiple //, e.g. stuff like path = path[1..-1] if path.start_with?('/')
Answers that simply call File.join(*paths) seem to be the accepted "Ruby way," yet they fail in cases where you pass a URI object, e.g. File.join(URI.join('some/path')) fails with TypeError: no implicit conversion of URI::Generic into String.
Below is what I ended up using:
module UrlHelper
def self.join(*paths)
# yes, Ruby's stdlib really does lack a functional join method for URLs
File.join(*paths.map(&:to_s))
end
end

Answer 9

You can use File.join('resource/', '/edit', '12?option=test')

Answer 10

I improved #Maximo Mussini's script to make it works gracefully:
SmartURI.join('http://example.com/subpath', 'hello', query: { token: secret })
=> "http://example.com/subpath/hello?token=secret"
https://gist.github.com/zernel/0f10c71f5a9e044653c1a65c6c5ad697
require 'uri'
module SmartURI
SEPARATOR = '/'
def self.join(*paths, query: nil)
paths = paths.compact.reject(&:empty?)
last = paths.length - 1
url = paths.each_with_index.map { |path, index|
_expand(path, index, last)
}.join
if query.nil?
return url
elsif query.is_a? Hash
return url + "?#{URI.encode_www_form(query.to_a)}"
else
raise "Unexpected input type for query: #{query}, it should be a hash."
end
end
def self._expand(path, current, last)
if path.starts_with?(SEPARATOR) && current != 0
path = path[1..-1]
end
unless path.ends_with?(SEPARATOR) || current == last
path = [path, SEPARATOR]
end
path
end
end

Answer 11

You can use this:
URI.join('http://exemple.com', '/a/', 'b/', 'c/', 'd')
=> #<URI::HTTP http://exemple.com/a/b/c/d>
URI.join('http://exemple.com', '/a/', 'b/', 'c/', 'd').to_s
=> "http://exemple.com/a/b/c/d"
See: http://ruby-doc.org/stdlib-2.4.1/libdoc/uri/rdoc/URI.html#method-c-join-label-Synopsis

Answer 12

My understanding of URI::join is that it thinks like a web browser does.
To evaluate it, point your mental web browser to the first parameter, and keep clicking links until you browse to the last parameter.
For example, URI::join('http://example.com/resource/', '/edit', '12?option=test'), you would browse like this:
http://example.com/resource/, click a link to /edit (a file at the root of the site)
http://example.com/edit, click a link to 12?option=test (a file in the same directory as edit)
http://example.com/12?option=test
If the first link were /edit/ (with a trailing slash), or /edit/foo, then the next link would be relative to /edit/ rather than /.
This page possibly explains it better than I can: Why is URI.join so counterintuitive?

Answer 13

This is my simple take on this problem, just splitting up all the path segments and join them together again. This only works if you're only working with relative path segments, but if that's all you want to do this is handy.
def join_paths *paths
paths.map{|p| p.split('/')}
.flatten
.reject(&:empty?)
.compact
.join('/')
end
Then you can use it like so:
join_paths 'foo/', '/bar', 'a/b/c', 'd' #=> "foo/bar/a/b/c/d"