This document is a list of user visible feature changes made between releases except for bug fixes.
Note that each entry is kept so brief that no reason behind or reference information is supplied with. For a full list of changes with all sufficient information, see the ChangeLog file.
Added keyword arguments.
Added %i and %I for symbol list creation (similar to %w and %W).
Default source encoding is changed to UTF-8. (was US-ASCII)
No warning for unused variables starting with '_'
added method:
added ARGF#codepoints and ARGF#each_codepoint
, like the corresponding methods for IO
.
added method:
added Array#bsearch
for binary search.
incompatible changes:
random parameter of Array#shuffle!
and Array#sample
now will be called with one argument, maximum value.
when given Range
arguments, Array#values_at
now returns nil for each value that is out-of-range.
added method:
added Enumerable#lazy
method for lazy enumeration.
added method:
added Enumerator#size
for lazy size evaluation.
extended method:
Enumerator.new
accept an argument for lazy size evaluation.
new class Enumerator::Lazy
for lazy enumeration
aliased method:
ENV.to_h
is a new alias for ENV.to_hash
incompatible changes:
Fiber#resume
cannot resume a fiber which invokes “Fiber#transfer”.
extended method:
File.fnmatch?
now expands braces in the pattern if File::FNM_EXTGLOB option is given.
improvements:
introduced the bitmap marking which suppresses to copy a memory page with Copy-on-Write.
introduced the non-recursive marking which avoids unexpected stack overflow.
added method:
added GC::Profiler.raw_data
which returns raw profile data for GC
.
added method:
added Hash#to_h
as explicit conversion method, like Array#to_a
.
extended method:
Hash#default_proc=
can be passed nil to clear the default proc.
deprecated methods:
IO#lines, bytes, chars and codepoints are deprecated.
added method:
added Kernel#Hash conversion method like Array() or Float().
added Kernel#__dir__ which returns the absolute path of the directory of the file from which this method is called.
added Kernel#caller_locations
which returns an array of frame information objects.
extended method:
Kernel#warn
accepts multiple args in like puts.
Kernel#caller
accepts second optional argument `n' which specify required caller size.
Kernel#to_enum and enum_for accept a block for lazy size evaluation.
incompatible changes:
system() and exec() close non-standard file descriptors (The default of :close_others option is changed to true by default.)
respond_to? against a protected method now returns false unless the second argument is true.
__callee__ has returned to the original behavior, and now returns the called name but not the original name in an aliased method.
Kernel#inspect does not call to_s anymore (it used to call redefined to_s).
added method:
added LoadError#path
method to return the file name that could not be loaded.
added method:
added Module#prepend
which is similar to Module#include
, however a method in the prepended module overrides the corresponding method in the prepending module.
added Module.prepended
and Module.prepend_features
, similar to included and append_features.
added Module#refine
, which extends a class or module locally. [experimental]
extended method:
Module#define_method
accepts a UnboundMethod
from a Module
.
Module#const_get
accepts a qualified constant string, e.g. Object.const_get(“Foo::Bar::Baz”)
Mutex
added method:
added Mutex#owned? which returns the mutex is held by current thread or not. [experimental]
incompatible changes:
Mutex#lock, Mutex#unlock, Mutex#try_lock, Mutex#synchronize and Mutex#sleep are no longer allowed to be used from trap handler and raise a ThreadError
in such case.
Mutex#sleep may spurious wakeup. Check after wakeup.
added method:
added nil.to_h which returns {}
new low level class to hold weak references to objects.
incompatible change:
removed Proc#== and eql? so two procs are == only when they are the same object.
added method:
added getsid for getting session id (unix only).
added method:
added Range#size
for lazy size evaluation.
added Range#bsearch
for binary search.
RubyVM
(MRI specific)
added RubyVM::InstructionSequence.of
to get the instruction sequence from a method or a block.
added RubyVM::InstructionSequence#path
, absolute_path, label, base_label and first_lineno to retrieve information from where the instruction sequence was defined.
added Environment variables to specify stack usage:
RUBY_THREAD_VM_STACK_SIZE: vm stack size used at thread creation. default: 128KB (32bit CPU) or 256KB (64bit CPU).
RUBY_THREAD_MACHINE_STACK_SIZE: machine stack size used at thread creation. default: 512KB or 1024KB.
RUBY_FIBER_VM_STACK_SIZE: vm stack size used at fiber creation. default: 64KB or 128KB.
RUBY_FIBER_MACHINE_STACK_SIZE: machine stack size used at fiber creation. default: 256KB or 512KB.
These variables are checked only at launched time.
added constant DEFAULT_PARAMS to get above default parameters.
added method:
added Signal.signame
which returns signal name
incompatible changes:
Signal.trap
raises ArgumentError
when :SEGV, :BUS, :ILL, :FPE, :VTALRM are specified.
added method:
added String#b
returning a copied string whose encoding is ASCII-8BIT.
change return value:
String#lines
now returns an array instead of an enumerator.
String#chars
now returns an array instead of an enumerator.
String#codepoints
now returns an array instead of an enumerator.
String#bytes
now returns an array instead of an enumerator.
added method:
added Struct#to_h
returning values with keys corresponding to the instance variable names.
added method:
added Thread#thread_variable_get
for getting thread local variables (these are different than Fiber
local variables).
added Thread#thread_variable_set
for setting thread local variables.
added Thread#thread_variables
for getting a list of the thread local variable keys.
added Thread#thread_variable?
for testing to see if a particular thread variable has been set.
added Thread.handle_interrupt
as well as instance and singleton methods pending_interrupt? for asynchronous handling of exceptions
added Thread#backtrace_locations
which returns similar information of Kernel#caller_locations
.
new class Thread::Backtrace::Location
to hold backtrace location information. These are returned by Thread#backtrace_locations
and Kernel#caller_locations
.
incompatible changes:
Thread#join
and Thread#value
now raises a ThreadError
if target thread is the current or main thread.
change return value:
Time#to_s
now returns US-ASCII encoding instead of BINARY.
new class. This class is replacement of set_trace_func. Easy to use and efficient implementation.
toplevel
added method:
added main.define_method which defines a global function.
added main.using, which imports refinements into the current file or eval string. [experimental]
See above.
These methods no longer return an Enumerator
, although passing a block is still supported for backwards compatibility.
Code like str.lines.with_index(1) { |line, lineno| … } no longer works because str.lines returns an array. Replace lines with each_line in such cases.
IO#lines
IO#chars
IO#codepoints
IO#bytes
ARGF#lines
ARGF#chars
ARGF#bytes
StringIO#lines
StringIO#chars
StringIO#codepoints
StringIO#bytes
Zlib::GzipReader#lines
Zlib::GzipReader#bytes
These methods are deprecated in favor of each_line, each_byte, each_char and each_codepoint.
Proc#==
These methods were removed. Two procs are == only when they are the same object.
Fixnum
Bignum
Fixnums, Bignums and Floats are frozen.
See above.
Merge Onigmo. github.com/k-takata/Onigmo
The :close_others option is true by default for system() and exec(). Also, the close-on-exec flag is set by default for all new file descriptors. This means file descriptors doesn't inherit to spawned process unless explicitly requested such as system(…, fd=>fd).
Kernel#respond_to? against a protected method now returns false unless the second argument is true.
Kernel#respond_to_missing?
Kernel#initialize_clone
Kernel#initialize_dup
These methods are now private.
See above.
Mutex#lock, Mutex#unlock, Mutex#try_lock, Mutex#synchronize and Mutex#sleep
See above.
cgi
Add HTML5 tag maker.
CGI#header
has been renamed to CGI#http_header
and aliased to CGI#header
.
When HTML5 tagmaker called, overwrite CGI#header
, CGI#header
function is to create a <header> element.
Removed CSV::dump and CSV::load to protect users from dangerous serialization vulnerability
iconv
Iconv has been removed. Use String#encode
instead.
io/console
new methods:
added IO#cooked
which sets the terminal to cooked mode within the given block.
added IO#cooked!
which sets the terminal to cooked.
extended method:
io/wait
new features:
added IO#wait_writable
method.
added IO#wait_readable
method as alias of IO#wait
.
json
updated to 1.7.7.
net/http
new features:
Proxies are now automatically detected from the http_proxy environment variable. See Net::HTTP::new
for details.
gzip and deflate compression are now requested for all requests by default. See Net::HTTP
for details.
SSL sessions are now reused across connections for a single instance. This speeds up connection by using a previously negotiated session.
Requests may be created from a URI
which sets the request_uri and host header of the request (but does not change the host connected to).
Responses contain the URI
requested which allows easier implementation of redirect following.
new methods:
extended method:
Net::HTTP#connect
uses local_host and local_port if specified.
net/imap
new methods:
Net::IMAP.default_port
Net::IMAP.default_imap_port
Net::IMAP.default_tls_port
Net::IMAP.default_ssl_port
Net::IMAP.default_imaps_port
objspace
new method:
openssl
Consistently raise an error when trying to encode nil values. All instances of OpenSSL::ASN1::Primitive
now raise TypeError
when calling to_der on an instance whose value is nil. All instances of OpenSSL::ASN1::Constructive
raise NoMethodError
in the same case. Constructing such values is still permitted.
TLS 1.1 & 1.2 support by setting OpenSSL::SSL::SSLContext#ssl_version to :TLSv1_2, :TLSv1_2_server, :TLSv1_2_client or :TLSv1_1, :TLSv1_1_server :TLSv1_1_client. The version being effectively used can be queried with OpenSSL::SSL#ssl_version. Furthermore, it is also possible to blacklist the new TLS versions with OpenSSL::SSL::OP_NO_TLSv1_1 and OpenSSL::SSL::OP_NO_TLSv1_2.
Added OpenSSL::SSL::SSLContext#renegotiation_cb
. A user-defined callback may be set which gets called whenever a new handshake is negotiated. This also allows to programmatically decline (client) renegotiation attempts.
Support for “0/n” splitting of records as BEAST mitigation via OpenSSL::SSL::OP_DONT_INSERT_EMPTY_FRAGMENTS.
The default options for OpenSSL::SSL::SSLContext
have changed to OpenSSL::SSL::OP_ALL & ~OpenSSL::SSL::OP_DONT_INSERT_EMPTY_FRAGMENTS instead of OpenSSL::SSL::OP_ALL only. This enables the countermeasure for the BEAST attack by default.
OpenSSL
requires passwords for decrypting PEM-encoded files to be at least four characters long. This led to awkward situations where an export with a password with fewer than four characters was possible, but accessing the file afterwards failed. OpenSSL::PKey::RSA
, OpenSSL::PKey::DSA
and OpenSSL::PKey::EC
therefore now enforce the same check when exporting a private key to PEM with a password - it has to be at least four characters long.
SSL/TLS support for the Next Protocol Negotiation extension. Supported with OpenSSL
1.0.1 and higher.
OpenSSL::OPENSSL_FIPS allows client applications to detect whether OpenSSL
is FIPS-enabled. OpenSSL.fips_mode=
allows turning on and off FIPS mode manually in order to adapt to situations where FIPS mode would be an explicit requirement.
Authenticated Encryption with Associated Data (AEAD) is supported via Cipher#auth_data= and Cipher#auth_tag/Cipher#auth_tag=. Currently (OpenSSL
1.0.1c), only GCM mode is supported.
ostruct
new methods:
OpenStruct#[]
, []=
OpenStruct#to_h
converts the struct to a hash.
extended method:
OpenStruct.new
also accepts an OpenStruct
/ Struct
.
pathname
extended method:
Pathname#find
returns an enumerator if no block is given.
rake
rake has been updated to version 0.9.5.
This version is backwards-compatible with previous rake versions and contains many bug fixes.
See rake.rubyforge.org/doc/release_notes/rake-0_9_5_rdoc.html for a list of changes in rake 0.9.3, 0.9.4 and 0.9.5.
RDoc
has been updated to version 4.0
This version is largely backwards-compatible with previous rdoc versions. The most notable change is an update to the ri data format (ri data must be regenerated for gems shared across rdoc versions). Further API changes are internal and won't affect most users.
Notable changes include:
Page support for ri. Try `ri ruby:` for a list of pages in ruby or `ri ruby:syntax/literals` for the syntax documentation for literals.
This also works for gems such as `ri rspec:README` for the rspec gem's README file.
Markdown support. See ri RDoc::Markdown
for details.
See github.com/rdoc/rdoc/blob/master/History.rdoc for a full list of changes in rdoc 4.0.
resolv
new methods:
Resolv::DNS::Config#timeouts=
rexml
REXML::Document#write supports Hash
arguments.
REXML::Document#write supports new :encoding option. It changes XML document encoding. Without :encoding option, encoding in XML declaration is used for XML document encoding.
RubyGems
Updated to 2.0.0
RubyGems 2.0.0 features the following improvements:
Improved support for default gems shipping with ruby 2.0.0+
A gem can have arbitrary metadata through Gem::Specification#metadata
`gem search` now defaults to –remote and is anchored like gem list.
Added –document to replace –rdoc and –ri. Use –no-document to disable documentation, –document=rdoc to only generate rdoc.
Only ri-format documentation is generated by default.
`gem server` uses RDoc::Servlet
from RDoc
4.0 to generate HTML documentation.
For an expanded list of updates and bug fixes see: github.com/rubygems/rubygems/blob/master/History.txt
shellwords
Shellwords#shellescape
now stringifies the given object using to_s.
Shellwords#shelljoin
accepts non-string objects in the given array, each of which is stringified using to_s.
stringio
deprecated methods:
StringIO#lines, bytes, chars and codepoints are deprecated.
syslog
Added Syslog::Logger
which provides a Logger
API atop Syslog
.
Syslog::Priority, Syslog::Level
, Syslog::Option
and Syslog::Macros
are introduced for easy detection of available constants on a running system.
tmpdir
incompatible changes:
Dir.mktmpdir
uses FileUtils.remove_entry
instead of FileUtils.remove_entry_secure
. This means that applications should not change the permission of the created temporary directory to make writable from other users.
yaml
Syck has been removed. YAML
now completely depends on libyaml being installed.
libyaml is now bundled with ruby, for cases where the library is not installed locally.
zlib
Added streaming support for Zlib::Inflate
and Zlib::Deflate
. This allows processing of a stream without the use of large amounts of memory.
Added support for the new deflate strategies Zlib::RLE and Zlib::FIXED.
Zlib
streams are now processed without the GVL. This allows gzip, zlib and deflate streams to be processed in parallel.
deprecated methods:
Zlib::GzipReader#lines and bytes are deprecated.
OpenStruct
new methods can conflict with custom attributes named “each_pair”, “eql?”, “hash” or “to_h”.
Dir.mktmpdir
in lib/tmpdir.rb
See above.
NUM2SHORT() and NUM2USHORT() added. They are similar to NUM2INT, but short.
rb_newobj_of() and NEWOBJ_OF() added. They create a new object of a given class.