pkgsrc/www/ruby-rest-client/ALTERNATIVES
taca 2672960f84 Update ruby-rest-client to 2.0.0.
pkgsrc change: Add pkg_alternatives support.

# 2.0.0

This release is largely API compatible, but makes several breaking changes.

- Drop support for Ruby 1.9
- Allow mime-types as new as 3.x (requires ruby 2.0)
- Respect Content-Type charset header provided by server. Previously,
  rest-client would not override the string encoding chosen by Net::HTTP. Now
  responses that specify a charset will yield a body string in that encoding.
  For example, `Content-Type: text/plain; charset=EUC-JP` will return a String
  encoded with `Encoding::EUC_JP`. (#361)
- Change exceptions raised on request timeout. Instead of
  `RestClient::RequestTimeout` (which is still used for HTTP 408), network
  timeouts will now raise either `RestClient::Exceptions::ReadTimeout` or
  `RestClient::Exceptions::OpenTimeout`, both of which inherit from
  `RestClient::Exceptions::Timeout`. For backwards compatibility, this still
  inherits from `RestClient::RequestTimeout` so existing uses will still work.
  This may change in a future major release. These new timeout classes also
  make the original wrapped exception available as `#original_exception`.
- Unify request exceptions under `RestClient::RequestFailed`, which still
  inherits from `ExceptionWithResponse`. Previously, HTTP 304, 401, and 404
  inherited directly from `ExceptionWithResponse` rather than from
  `RequestFailed`. Now _all_ HTTP status code exceptions inherit from both.
- Rename the `:timeout` request option to `:read_timeout`. When `:timeout` is
  passed, now set both `:read_timeout` and `:open_timeout`.
- Change default HTTP Accept header to `*/*`
- Use a more descriptive User-Agent header by default
- Drop RC4-MD5 from default cipher list
- Only prepend http:// to URIs without a scheme
- Fix some support for using IPv6 addresses in URLs (still affected by Ruby
  2.0+ bug https://bugs.ruby-lang.org/issues/9129, with the fix expected to be
  backported to 2.0 and 2.1)
- `Response` objects are now a subclass of `String` rather than a `String` that
  mixes in the response functionality. Most of the methods remain unchanged,
  but this makes it much easier to understand what is happening when you look
  at a RestClient response object. There are a few additional changes:
  - Response objects now implement `.inspect` to make this distinction clearer.
  - `Response#to_i` will now behave like `String#to_i` instead of returning the
    HTTP response code, which was very surprising behavior.
  - `Response#body` and `#to_s` will now return a true `String` object rather
    than self. Previously there was no easy way to get the true `String`
    response instead of the Frankenstein response string object with
    AbstractResponse mixed in.
  - Response objects no longer accept an extra request args hash, but instead
    access request args directly from the request object, which reduces
    confusion and duplication.
- Handle multiple HTTP response headers with the same name (except for
  Set-Cookie, which is special) by joining the values with a comma space,
  compliant with RFC 7230
- Rewrite cookie support to be much smarter and to use cookie jars consistently
  for requests, responses, and redirection in order to resolve long-standing
  complaints about the previously broken behavior: (#498)
  - The `:cookies` option may now be a Hash of Strings, an Array of
    HTTP::Cookie objects, or a full HTTP::CookieJar.
  - Add `RestClient::Request#cookie_jar` and reimplement `Request#cookies` to
    be a wrapper around the cookie jar.
  - Still support passing the `:cookies` option in the headers hash, but now
    raise ArgumentError if that option is also passed to `Request#initialize`.
  - Warn if both `:cookies` and a `Cookie` header are supplied.
  - Use the `Request#cookie_jar` as the basis for `Response#cookie_jar`,
    creating a copy of the jar and adding any newly received cookies.
  - When following redirection, also use this same strategy so that cookies
    from the original request are carried through in a standards-compliant way
    by the cookie jar.
- Don't set basic auth header if explicit `Authorization` header is specified
- Add `:proxy` option to requests, which can be used for thread-safe
  per-request proxy configuration, overriding `RestClient.proxy`
- Allow overriding `ENV['http_proxy']` to disable proxies by setting
  `RestClient.proxy` to a falsey value. Previously there was no way in Ruby 2.x
  to turn off a proxy specified in the environment without changing `ENV`.
- Add actual support for streaming request payloads. Previously rest-client
  would call `.to_s` even on RestClient::Payload::Streamed objects. Instead,
  treat any object that responds to `.read` as a streaming payload and pass it
  through to `.body_stream=` on the Net:HTTP object. This massively reduces the
  memory required for large file uploads.
- Changes to redirection behavior: (#381, #484)
  - Remove `RestClient::MaxRedirectsReached` in favor of the normal
    `ExceptionWithResponse` subclasses. This makes the response accessible on
    the exception object as `.response`, making it possible for callers to tell
    what has actually happened when the redirect limit is reached.
  - When following HTTP redirection, store a list of each previous response on
    the response object as `.history`. This makes it possible to access the
    original response headers and body before the redirection was followed.
  - Follow redirection consistently, regardless of whether the HTTP method was
    passed as a symbol or string. Under the hood rest-client now normalizes the
    HTTP request method to a lowercase string.
- Add `:before_execution_proc` option to `RestClient::Request`. This makes it
  possible to add procs like `RestClient.add_before_execution_proc` to a single
  request without global state.
- Run tests on Travis's beta OS X support.
- Make `Request#transmit` a private method, along with a few others.
- Refactor URI parsing to happen earlier, in Request initialization.
- Improve consistency and functionality of complex URL parameter handling:
  - When adding URL params, handle URLs that already contain params.
  - Add new convention for handling URL params containing deeply nested arrays
    and hashes, unify handling of null/empty values, and use the same code for
    GET and POST params. (#437)
  - Add the RestClient::ParamsArray class, a simple array-like container that
    can be used to pass multiple keys with same name or keys where the ordering
    is significant.
- Add a few more exception classes for obscure HTTP status codes.
- Multipart: use a much more robust multipart boundary with greater entropy.
- Make `RestClient::Payload::Base#inspect` stop pretending to be a String.
- Add `Request#redacted_uri` and `Request#redacted_url` to display the URI
  with any password redacted.

# 2.0.0.rc1

Changes in the release candidate that did not persist through the final 2.0.0
release:
- RestClient::Exceptions::Timeout was originally going to be a direct subclass
  of RestClient::Exception in the release candidate. This exception tree was
  made a subclass of RestClient::RequestTimeout prior to the final release.
2016-10-18 16:08:35 +00:00

1 line
52 B
Text

bin/restclient @PREFIX@/bin/restclient@RUBY_SUFFIX@