1
0
Fork 0
mirror of https://github.com/ruby/ruby.git synced 2022-11-09 12:17:21 -05:00
ruby--ruby/NEWS.md

600 lines
18 KiB
Markdown
Raw Normal View History

# NEWS for Ruby 3.1.0
2021-12-15 11:24:31 -05:00
This document is a list of user-visible feature changes
since the **3.0.0** release, except for bug fixes.
Note that each entry is kept to a minimum, see links for details.
## Language changes
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
2021-12-15 11:24:31 -05:00
* The block argument can now be anonymous if the block will
only be passed to another method. [[Feature #11256]]
```ruby
def foo(&)
bar(&)
end
```
* Pin operator now takes an expression. [[Feature #17411]]
```ruby
Prime.each_cons(2).lazy.find_all{_1 in [n, ^(n + 2)]}.take(3).to_a
#=> [[3, 5], [5, 7], [11, 13]]
```
* Pin operator now supports instance, class, and global variables.
[[Feature #17724]]
```ruby
@n = 5
Prime.each_cons(2).lazy.find{_1 in [n, ^@n]}
#=> [3, 5]
```
* One-line pattern matching is no longer experimental.
* Parentheses can be omitted in one-line pattern matching.
[[Feature #16182]]
```ruby
[0, 1] => _, x
{y: 2} => y:
x #=> 1
y #=> 2
```
* Multiple assignment evaluation order has been made consistent with
single assignment evaluation order. With single assignment, Ruby
uses a left-to-right evaluation order. With this code:
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
```ruby
foo[0] = bar
```
The following evaluation order is used:
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
1. `foo`
2. `bar`
3. `[]=` called on the result of `foo`
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
In Ruby before 3.1.0, multiple assignment did not follow this
evaluation order. With this code:
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
```ruby
foo[0], bar.baz = a, b
```
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
Versions of Ruby before 3.1.0 would evaluate in the following
order
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
1. `a`
2. `b`
3. `foo`
4. `[]=` called on the result of `foo`
5. `bar`
6. `baz=` called on the result of `bar`
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
2021-12-15 11:24:31 -05:00
Starting in Ruby 3.1.0, the evaluation order is now consistent with
single assignment, with the left-hand side being evaluated before
the right-hand side:
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
1. `foo`
2. `bar`
3. `a`
4. `b`
5. `[]=` called on the result of `foo`
6. `baz=` called on the result of `bar`
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
[[Bug #4443]]
Evaluate multiple assignment left hand side before right hand side In regular assignment, Ruby evaluates the left hand side before the right hand side. For example: ```ruby foo[0] = bar ``` Calls `foo`, then `bar`, then `[]=` on the result of `foo`. Previously, multiple assignment didn't work this way. If you did: ```ruby abc.def, foo[0] = bar, baz ``` Ruby would previously call `bar`, then `baz`, then `abc`, then `def=` on the result of `abc`, then `foo`, then `[]=` on the result of `foo`. This change makes multiple assignment similar to single assignment, changing the evaluation order of the above multiple assignment code to calling `abc`, then `foo`, then `bar`, then `baz`, then `def=` on the result of `abc`, then `[]=` on the result of `foo`. Implementing this is challenging with the stack-based virtual machine. We need to keep track of all of the left hand side attribute setter receivers and setter arguments, and then keep track of the stack level while handling the assignment processing, so we can issue the appropriate topn instructions to get the receiver. Here's an example of how the multiple assignment is executed, showing the stack and instructions: ``` self # putself abc # send abc, self # putself abc, foo # send abc, foo, 0 # putobject 0 abc, foo, 0, [bar, baz] # evaluate RHS abc, foo, 0, [bar, baz], baz, bar # expandarray abc, foo, 0, [bar, baz], baz, bar, abc # topn 5 abc, foo, 0, [bar, baz], baz, abc, bar # swap abc, foo, 0, [bar, baz], baz, def= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz], baz, foo # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0 # topn 3 abc, foo, 0, [bar, baz], baz, foo, 0, baz # topn 2 abc, foo, 0, [bar, baz], baz, []= # send abc, foo, 0, [bar, baz], baz # pop abc, foo, 0, [bar, baz] # pop [bar, baz], foo, 0, [bar, baz] # setn 3 [bar, baz], foo, 0 # pop [bar, baz], foo # pop [bar, baz] # pop ``` As multiple assignment must deal with splats, post args, and any level of nesting, it gets quite a bit more complex than this in non-trivial cases. To handle this, struct masgn_state is added to keep track of the overall state of the mass assignment, which stores a linked list of struct masgn_attrasgn, one for each assigned attribute. This adds a new optimization that replaces a topn 1/pop instruction combination with a single swap instruction for multiple assignment to non-aref attributes. This new approach isn't compatible with one of the optimizations previously used, in the case where the multiple assignment return value was not needed, there was no lhs splat, and one of the left hand side used an attribute setter. This removes that optimization. Removing the optimization allowed for removing the POP_ELEMENT and adjust_stack functions. This adds a benchmark to measure how much slower multiple assignment is with the correct evaluation order. This benchmark shows: * 4-9% decrease for attribute sets * 14-23% decrease for array member sets * Basically same speed for local variable sets Importantly, it shows no significant difference between the popped (where return value of the multiple assignment is not needed) and !popped (where return value of the multiple assignment is needed) cases for attribute and array member sets. This indicates the previous optimization, which was dropped in the evaluation order fix and only affected the popped case, is not important to performance. Fixes [Bug #4443]
2021-04-21 13:49:19 -04:00
* Values in Hash literals and keyword arguments can be omitted.
[[Feature #14579]]
2021-09-16 04:31:39 -04:00
For example,
* `{x:, y:}` is a syntax sugar of `{x: x, y: y}`.
* `foo(x:, y:)` is a syntax sugar of `foo(x: x, y: y)`.
Constant names, local variable names, and method names are allowed as
key names. Note that a reserved word is considered as a local
variable or method name even if it's a pseudo variable name such as
`self`.
2021-12-15 11:24:31 -05:00
* Non main-Ractors can get instance variables (ivars) of classes/modules
if ivars refer to shareable objects.
[[Feature #17592]]
2021-10-22 12:38:05 -04:00
2021-12-15 11:24:31 -05:00
* A command syntax is allowed in endless method definitions, i.e.,
you can now write `def foo = puts "Hello"`.
Note that `private def foo = puts "Hello"` does not parse.
[[Feature #17398]]
## Command line options
* `--disable-gems` is now explicitly declared as "just for debugging".
2021-12-15 11:24:31 -05:00
Never use it in any real-world codebase.
[[Feature #17684]]
## Core classes updates
2021-12-15 11:24:31 -05:00
Note: We're only listing outstanding class updates.
2021-04-16 03:12:10 -04:00
* Array
* Array#intersect? is added. [[Feature #15198]]
* Class
* Class#subclasses, which returns an array of classes
directly inheriting from the receiver, not
including singleton classes.
[[Feature #18273]]
```ruby
class A; end
class B < A; end
class C < B; end
class D < A; end
A.subclasses #=> [D, B]
B.subclasses #=> [C]
C.subclasses #=> []
```
2021-01-02 02:42:45 -05:00
* Enumerable
* Enumerable#compact is added. [[Feature #17312]]
2021-01-02 02:42:45 -05:00
* Enumerable#tally now accepts an optional hash to count. [[Feature #17744]]
* Enumerable#each_cons and each_slice to return a receiver. [[GH-1509]]
```ruby
[1, 2, 3].each_cons(2){}
# 3.0 => nil
# 3.1 => [1, 2, 3]
[1, 2, 3].each_slice(2){}
# 3.0 => nil
# 3.1 => [1, 2, 3]
```
2021-01-02 02:42:45 -05:00
* Enumerator::Lazy
* Enumerator::Lazy#compact is added. [[Feature #17312]]
2021-01-02 02:42:45 -05:00
2021-03-15 07:05:03 -04:00
* File
* File.dirname now accepts an optional argument for the level to
strip path components. [[Feature #12194]]
2021-03-15 07:05:03 -04:00
2021-12-06 03:36:01 -05:00
* GC
* "GC.measure_total_time = true" enables the measurement of GC.
Measurement can introduce overhead. It is enabled by default.
GC.measure_total_time returns the current setting.
GC.stat[:time] or GC.stat(:time) returns measured time
2021-12-21 12:49:19 -05:00
in milli-seconds. [[[Feature #10917]]]
2021-12-06 03:36:01 -05:00
2021-12-21 12:49:19 -05:00
* GC.total_time returns measured time in nano-seconds. [[[Feature #10917]]]
2021-12-06 03:36:01 -05:00
* Integer
* Integer.try_convert is added. [[Feature #15211]]
* Kernel
* Kernel#load now accepts a module as the second argument,
2021-12-15 11:24:31 -05:00
and will load the file using the given module as the
top-level module. [[Feature #6210]]
* Marshal
* Marshal.load now accepts a `freeze: true` option.
All returned objects are frozen except for `Class` and
`Module` instances. Strings are deduplicated. [[Feature #18148]]
2021-09-16 06:55:31 -04:00
* MatchData
* MatchData#match is added [[Feature #18172]]
2021-09-16 06:55:31 -04:00
* MatchData#match_length is added [[Feature #18172]]
2021-09-16 06:55:31 -04:00
* Method/UnboundMethod
* Method#public?, Method#private?, Method#protected?,
UnboundMethod#public?, UnboundMethod#private?,
UnboundMethod#protected? have been added. [[Feature #11689]]
* Module
* Module#prepend now modifies the ancestor chain if the receiver
already includes the argument. Module#prepend still does not
modify the ancestor chain if the receiver has already prepended
the argument. [[Bug #17423]]
* Module#private, #public, #protected, and #module_function will
now return their arguments. If a single argument is given, it
is returned. If no arguments are given, nil is returned. If
multiple arguments are given, they are returned as an array.
[[Feature #12495]]
* Process
* Process.\_fork is added. This is a core method for fork(2).
Do not call this method directly; it is called by existing
fork methods: Kernel.#fork, Process.fork, and IO.popen("-").
2021-12-15 11:24:31 -05:00
Application monitoring libraries can overwrite this method to
hook fork events. [[Feature #17795]]
* Struct
* Passing only keyword arguments to Struct#initialize is warned.
You need to use a Hash literal to set a Hash to a first member.
[[Feature #16806]]
* StructClass#keyword_init? is added [[Feature #18008]]
2021-07-15 05:21:49 -04:00
* String
* Update Unicode version to 13.0.0 [[Feature #17750]]
and Emoji version to 13.0 [[Feature #18029]]
2021-12-15 11:24:31 -05:00
* String#unpack and String#unpack1 now accept an `offset:` keyword
argument to start the unpacking after an arbitrary number of bytes
have been skipped. If `offset` is outside of the string bounds
`ArgumentError` is raised. [[Feature #18254]]
* Queue
* Queue#initialize now accepts an Enumerable of initial values.
[[Feature #17327]]
* Thread
* Thread#native_thread_id is added. [[Feature #17853]]
* Thread::Backtrace
* Thread::Backtrace.limit, which returns the value to limit backtrace
length set by `--backtrace-limit` command line option, is added.
[[Feature #17479]]
* Time
* Time.new now accepts optional `in:` keyword argument for the
timezone, as well as `Time.at` and `Time.now`, so that is now
you can omit minor arguments to `Time.new`. [[Feature #17485]]
* Time#strftime supports RFC 3339 UTC for unknown offset local
time, `-0000`, as `%-z`. [[Feature #17544]]
* TracePoint
2021-12-21 03:44:48 -05:00
* TracePoint.allow_reentry is added to allow reenter while TracePoint
callback.
[[Feature #15912]]
2021-02-16 04:12:47 -05:00
* $LOAD_PATH
* $LOAD_PATH.resolve_feature_path does not raise. [[Feature #16043]]
2021-02-16 04:12:47 -05:00
2021-06-22 07:14:45 -04:00
* Fiber Scheduler
* Add support for `Addrinfo.getaddrinfo` using `address_resolve` hook.
[[Feature #17370]]
2021-06-22 07:14:45 -04:00
* Introduce non-blocking `Timeout.timeout` using `timeout_after` hook.
[[Feature #17470]]
2021-06-22 07:14:45 -04:00
* Introduce new scheduler hooks `io_read` and `io_write` along with a
low level `IO::Buffer` for zero-copy read/write. [[Feature #18020]]
2021-07-02 06:41:16 -04:00
* IO hooks `io_wait`, `io_read`, `io_write`, receive the original IO object
where possible. [[Bug #18003]]
2021-06-22 07:14:45 -04:00
* Make `Monitor` fiber-safe. [[Bug #17827]]
2021-06-22 07:14:45 -04:00
* Replace copy coroutine with pthread implementation. [[Feature #18015]]
* Refinement
* New class which represents a module created by Module#refine.
`include` and `prepend` are deprecated, and `import_methods` is added
instead. [[Bug #17429]]
## Stdlib updates
* The following default gem are updated.
* RubyGems 3.3.2
* base64 0.1.1
* benchmark 0.2.0
* bigdecimal 3.1.1
* bundler 2.3.2
* cgi 0.3.1
* csv 3.2.2
* date 3.2.2
* did_you_mean 1.6.0.pre.alpha
* digest 3.1.0.pre3
* drb 2.1.0
* erb 2.2.3
* error_highlight 0.3.0
* etc 1.3.0
* fcntl 1.0.1
* fiddle 1.1.0
* fileutils 1.6.0
* find 0.1.1
* io-console 0.5.9
* io-wait 0.2.1
* ipaddr 1.2.3
* irb 1.3.8.pre.11
* json 2.6.1
2021-12-20 04:02:19 -05:00
* logger 1.5.0
* net-http 0.2.0
* net-protocol 0.1.2
* nkf 0.1.1
* open-uri 0.2.0
* openssl 3.0.0.pre
* optparse 0.2.0
* ostruct 0.5.2
* pathname 0.2.0
2021-12-19 22:46:11 -05:00
* pp 0.3.0
* prettyprint 0.1.1
* psych 4.0.3
* racc 1.6.0
* rdoc 6.4.0
* readline-ext 0.1.3
* reline 0.2.8.pre.11
* resolv 0.2.1
* rinda 0.1.1
* ruby2_keywords 0.0.5
* securerandom 0.1.1
* set 1.0.2
* stringio 3.0.1
* strscan 3.0.1
* tempfile 0.1.2
* time 0.2.0
* timeout 0.2.0
* tmpdir 0.1.2
* un 0.2.0
* uri 0.11.0
* yaml 0.2.0
* zlib 2.1.1
* The following bundled gems are updated.
2021-12-15 04:53:16 -05:00
* minitest 5.15.0
* power_assert 2.0.1
* rake 13.0.6
2021-12-20 02:05:44 -05:00
* test-unit 3.5.3
2021-12-17 11:18:25 -05:00
* rexml 3.2.5
2021-12-14 02:06:50 -05:00
* rbs 1.8.1
* typeprof 0.21.1
* The following default gems are now bundled gems.
2021-12-17 11:18:25 -05:00
* net-ftp 0.1.3
* net-imap 0.2.2
* net-pop 0.1.1
* net-smtp 0.3.1
* matrix 0.4.2
* prime 0.1.2
2021-12-17 14:21:56 -05:00
* debug 1.4.0
2021-12-17 06:34:14 -05:00
* Coverage measurement now supports suspension. You can use `Coverage.suspend`
to stop the measurement temporarily, and `Coverage.resume` to restart it.
See [[Feature #18176]] in detail.
* Random::Formatter is moved to random/formatter.rb, so that you can
2021-12-15 11:24:31 -05:00
use `Random#hex`, `Random#base64`, and so on without SecureRandom.
[[Feature #18190]]
## Compatibility issues
2021-12-15 11:24:31 -05:00
Note: Excluding feature bug fixes.
2021-06-22 07:14:45 -04:00
* `rb_io_wait_readable`, `rb_io_wait_writable` and `rb_wait_for_single_fd` are
deprecated in favour of `rb_io_maybe_wait_readable`,
`rb_io_maybe_wait_writable` and `rb_io_maybe_wait` respectively.
`rb_thread_wait_fd` and `rb_thread_fd_writable` are deprecated. [[Bug #18003]]
## Stdlib compatibility issues
2020-06-18 08:11:19 -04:00
2021-01-21 01:02:24 -05:00
* `ERB#initialize` warns `safe_level` and later arguments even without -w.
[[Feature #14256]]
2021-10-19 21:18:54 -04:00
* `lib/debug.rb` is replaced with `debug.gem`
* `Kernel#pp` in `lib/pp.rb` uses the width of `IO#winsize` by default.
This means that the output width is automatically changed depending on
your terminal size. [[Feature #12913]]
* Psych 4.0 changes `Psych.load` as `safe_load` by the default.
You may need to use Psych 3.3.2 for migrating to this behavior.
[[Bug #17866]]
## C API updates
* Documented. [[GH-4815]]
* `rb_gc_force_recycle` is deprecated and has been changed to a no-op.
[[Feature #18290]]
## Implementation improvements
* `instance_eval` and `instance_exec` now only allocate a singleton class when
required, avoiding extra objects and improving performance. [[GH-5146]]
* The performance of `Struct` accessors is improved. [[GH-5131]]
2021-12-06 12:03:06 -05:00
* `mandatory_only?` builtin special form to improve performance on
builtin methods. [[GH-5112]]
* Experimental feature Variable Width Allocation in the garbage collector.
This feature is turned off by default and can be enabled by compiling Ruby
with flag `USE_RVARGC=1` set. [[Feature #18045]] [[Feature #18239]]
## JIT
* Rename Ruby 3.0's `--jit` to `--mjit`, and alias `--jit` to `--yjit`
on non-Windows x86-64 platforms and to `--mjit` on others.
### MJIT
* The default `--mjit-max-cache` is changed from 100 to 10000.
* JIT-ed code is no longer cancelled when a TracePoint for class events
is enabled.
* The JIT compiler no longer skips compilation of methods longer than
1000 instructions.
* `--mjit-verbose` and `--mjit-warning` output "JIT cancel" when JIT-ed
code is disabled because TracePoint or GC.compact is used.
2021-10-20 15:51:07 -04:00
### YJIT: New experimental in-process JIT compiler
New JIT compiler available as an experimental feature. [[Feature #18229]]
See [this blog post](https://shopify.engineering/yjit-just-in-time-compiler-cruby
) introducing the project.
2021-10-20 15:51:07 -04:00
* Disabled by default, use `--yjit` command-line option to enable YJIT.
* Performance improvements on most real-world software, up to 22% on railsbench, 39% on liquid-render.
* Fast warm-up times.
* Limited to macOS & Linux on x86-64 platforms for now.
2020-11-10 12:15:57 -05:00
## Static analysis
2020-10-20 00:16:19 -04:00
### RBS
2020-09-25 04:08:39 -04:00
* [rbs collection](https://github.com/ruby/rbs/blob/master/docs/collection.md) has been introduced to manage gems RBSs.
2021-12-16 23:46:09 -05:00
* Many signatures for built-in and standard libraries have been added/updated.
* It includes many bug fixes and performance improvements too.
See the [CHANGELOG.md](https://github.com/ruby/rbs/blob/master/CHANGELOG.md) for more information.
2020-10-20 00:16:19 -04:00
### TypeProf
2021-10-19 22:05:05 -04:00
* [Experimental IDE support](https://github.com/ruby/typeprof/blob/master/doc/ide.md) has been implemented.
* Many bug fixes and performance improvements since Ruby 3.0.0.
2021-10-19 21:18:54 -04:00
## Debugger
* A new debugger [debug.gem](https://github.com/ruby/debug) is bundled.
2021-12-15 11:24:31 -05:00
debug.gem is a fast debugger implementation, and it provides many features
like remote debugging, colorful REPL, IDE (VSCode) integration, and more.
2021-10-19 21:18:54 -04:00
It replaces `lib/debug.rb` standard library.
* `rdbg` command is also installed into `bin/` directory to start and control
debugging execution.
2021-10-19 22:05:05 -04:00
## error_highlight
A built-in gem called error_highlight has been introduced.
2021-12-15 11:24:31 -05:00
It shows fine-grained error locations in the backtrace.
Example: `title = json[:article][:title]`
If `json` is nil, it shows:
```
$ ruby test.rb
test.rb:2:in `<main>': undefined method `[]' for nil:NilClass (NoMethodError)
title = json[:article][:title]
^^^^^^^^^^
```
If `json[:article]` returns nil, it shows:
2021-10-19 22:05:05 -04:00
```
$ ruby test.rb
test.rb:2:in `<main>': undefined method `[]' for nil:NilClass (NoMethodError)
2021-10-19 22:05:05 -04:00
title = json[:article][:title]
^^^^^^^^
2021-10-19 22:05:05 -04:00
```
This feature is enabled by default.
2021-10-19 22:05:05 -04:00
You can disable it by using a command-line option `--disable-error_highlight`.
See [the repository](https://github.com/ruby/error_highlight) in detail.
## Miscellaneous changes
2021-01-02 02:42:45 -05:00
* lib/objspace/trace.rb is added, which is a tool for tracing the object
allocation. Just by requiring this file, tracing is started *immediately*.
Just by `Kernel#p`, you can investigate where an object was created.
Note that just requiring this file brings a large performance overhead.
2021-12-15 11:24:31 -05:00
This is only for debugging purposes. Do not use this in production.
2021-07-22 23:41:01 -04:00
[[Feature #17762]]
2021-01-02 02:42:45 -05:00
* Now exceptions raised in finalizers will be printed to `STDERR`, unless
`$VERBOSE` is `nil`. [[Feature #17798]]
* `ruby -run -e httpd` displays URLs to access. [[Feature #17847]]
* Add `ruby -run -e colorize` to colorize Ruby code using
`IRB::Color.colorize_code`.
2021-07-22 23:41:01 -04:00
[Bug #4443]: https://bugs.ruby-lang.org/issues/4443
[Feature #6210]: https://bugs.ruby-lang.org/issues/6210
2021-12-21 12:49:19 -05:00
[Feature #10917]: https://bugs.ruby-lang.org/issues/10917
[Feature #11256]: https://bugs.ruby-lang.org/issues/11256
[Feature #11689]: https://bugs.ruby-lang.org/issues/11689
2021-03-16 01:07:27 -04:00
[Feature #12194]: https://bugs.ruby-lang.org/issues/12194
[Feature #12495]: https://bugs.ruby-lang.org/issues/12495
2021-12-07 01:02:08 -05:00
[Feature #12913]: https://bugs.ruby-lang.org/issues/12913
2021-01-21 01:02:24 -05:00
[Feature #14256]: https://bugs.ruby-lang.org/issues/14256
[Feature #14579]: https://bugs.ruby-lang.org/issues/14579
2021-04-16 03:12:10 -04:00
[Feature #15198]: https://bugs.ruby-lang.org/issues/15198
[Feature #15211]: https://bugs.ruby-lang.org/issues/15211
[Feature #15912]: https://bugs.ruby-lang.org/issues/15912
2021-02-16 04:12:47 -05:00
[Feature #16043]: https://bugs.ruby-lang.org/issues/16043
[Feature #16182]: https://bugs.ruby-lang.org/issues/16182
2021-12-19 09:43:39 -05:00
[Feature #16806]: https://bugs.ruby-lang.org/issues/16806
2021-01-02 02:42:45 -05:00
[Feature #17312]: https://bugs.ruby-lang.org/issues/17312
[Feature #17327]: https://bugs.ruby-lang.org/issues/17327
2021-10-27 03:55:15 -04:00
[Feature #17370]: https://bugs.ruby-lang.org/issues/17370
[Feature #17398]: https://bugs.ruby-lang.org/issues/17398
[Feature #17411]: https://bugs.ruby-lang.org/issues/17411
2021-07-22 23:41:01 -04:00
[Bug #17423]: https://bugs.ruby-lang.org/issues/17423
2021-10-27 03:55:15 -04:00
[Bug #17429]: https://bugs.ruby-lang.org/issues/17429
[Feature #17470]: https://bugs.ruby-lang.org/issues/17470
[Feature #17479]: https://bugs.ruby-lang.org/issues/17479
[Feature #17485]: https://bugs.ruby-lang.org/issues/17485
[Feature #17544]: https://bugs.ruby-lang.org/issues/17544
2021-10-22 12:38:05 -04:00
[Feature #17592]: https://bugs.ruby-lang.org/issues/17592
[Feature #17684]: https://bugs.ruby-lang.org/issues/17684
[Feature #17724]: https://bugs.ruby-lang.org/issues/17724
[Feature #17744]: https://bugs.ruby-lang.org/issues/17744
2021-10-27 03:55:15 -04:00
[Feature #17750]: https://bugs.ruby-lang.org/issues/17750
[Feature #17762]: https://bugs.ruby-lang.org/issues/17762
2021-11-08 21:30:03 -05:00
[Feature #17795]: https://bugs.ruby-lang.org/issues/17795
[Feature #17798]: https://bugs.ruby-lang.org/issues/17798
2021-07-22 23:41:01 -04:00
[Bug #17827]: https://bugs.ruby-lang.org/issues/17827
[Feature #17847]: https://bugs.ruby-lang.org/issues/17847
2021-10-27 03:55:15 -04:00
[Feature #17853]: https://bugs.ruby-lang.org/issues/17853
[Bug #17866]: https://bugs.ruby-lang.org/issues/17866
2021-10-27 03:55:15 -04:00
[Bug #18003]: https://bugs.ruby-lang.org/issues/18003
2021-08-05 04:25:17 -04:00
[Feature #18008]: https://bugs.ruby-lang.org/issues/18008
[Feature #18015]: https://bugs.ruby-lang.org/issues/18015
2021-07-02 06:41:16 -04:00
[Feature #18020]: https://bugs.ruby-lang.org/issues/18020
2021-08-22 09:27:29 -04:00
[Feature #18029]: https://bugs.ruby-lang.org/issues/18029
2021-12-13 10:03:37 -05:00
[Feature #18045]: https://bugs.ruby-lang.org/issues/18045
[Feature #18148]: https://bugs.ruby-lang.org/issues/18148
2021-09-16 06:55:31 -04:00
[Feature #18172]: https://bugs.ruby-lang.org/issues/18172
[Feature #18176]: https://bugs.ruby-lang.org/issues/18176
[Feature #18190]: https://bugs.ruby-lang.org/issues/18190
2021-10-24 19:38:44 -04:00
[Feature #18229]: https://bugs.ruby-lang.org/issues/18229
2021-12-13 10:03:37 -05:00
[Feature #18239]: https://bugs.ruby-lang.org/issues/18239
[Feature #18254]: https://bugs.ruby-lang.org/issues/18254
[Feature #18273]: https://bugs.ruby-lang.org/issues/18273
2021-12-05 06:31:58 -05:00
[Feature #18290]: https://bugs.ruby-lang.org/issues/18290
[GH-1509]: https://github.com/ruby/ruby/pull/1509
[GH-4815]: https://github.com/ruby/ruby/pull/4815
2021-12-06 12:03:06 -05:00
[GH-5112]: https://github.com/ruby/ruby/pull/5112
[GH-5131]: https://github.com/ruby/ruby/pull/5131
2021-12-03 15:58:09 -05:00
[GH-5146]: https://github.com/ruby/ruby/pull/5146