Project

General

Profile

Actions

Feature #8992

closed

Use String#freeze and compiler tricks to replace "str"f suffix

Added by headius (Charles Nutter) over 10 years ago. Updated almost 5 years ago.

Status:
Closed
Target version:
[ruby-core:57705]

Description

BACKGROUND:

In https://bugs.ruby-lang.org/issues/8579 @charliesome introduced the "f" suffix for creating already-frozen strings. A string like "str"f would have the following characteristics:

  • It would be frozen before the expression returned
  • It would be the same object everywhere, pulling from a global "fstring" table

To avoid memory leaks, these pooled strings would remove themselves from the "fstring" table on GC.

However, there are problems with this new syntax:

  • It will never parse in Ruby 2.0 and earlier.
  • It's not particularly attractive, though this is a subjective matter.
  • It does not lend itself well to use in other scenarios, such as for arrays and hashes (http://bugs.ruby-lang.org/issues/8909 )

PROPOSAL:

I propose that we eliminate the new "f" suffix and just make the compiler smart enough to see literal strings with .frozen the same way.

So this code:

str = "mystring".freeze

Would be equivalent in the compiler to this code:

str = "mystring"f

And the fstring table would still be used to return pooled instances.

IMPLEMENTATION NOTES:

The fstring table already exists on master and would be used for these pooled strings. An open question is whether the compiler should forever optimize "str".frozen to return the pooled version or whether it should check (inline-cache style) whether String#freeze has been replaced. I am ok with either, but the best potential comes from ignoring String#freeze redefinitions...or making it impossible to redefine String#freeze.

BONUS BIKESHEDDING:

If we do not want to overload the existing .freeze method in this way, we could follow suggestions in http://bugs.ruby-lang.org/issues/8977 to add a new "frozen" method (or some other name) that the compiler would understand.

If it were "frozen", the following two lines would be equivalent:

str = "mystring".frozen
str = "mystring"f

In addition, using .frozen on any string would put it in the fstring table and return that pooled version.

I also propose one alternative method name: the unary ~ operator.

There is no ~ on String right now, and it has no meaning for strings that we'd be overriding. So the following two lines would be equivalent:

str = ~"mystring"
str = "mystring"f

JUSTIFICATION:

Making the compiler aware of normal method-based String freezing has the following advantages:

  • It will parse in all versions of Ruby.
  • It will be equivalent in all versions of Ruby other than the fstring pooling.
  • It extends neatly to Array and Hash; the compiler can see Array or Hash with literal elements and return the same object.
  • It does not require a pragma (http://bugs.ruby-lang.org/issues/8976 )
  • It looks like Ruby.

Related issues 2 (0 open2 closed)

Related to Ruby master - Feature #8579: Frozen string syntaxClosed06/29/2013Actions
Related to Ruby master - Misc #20222: Dedup-ing clarificationClosedActions

Updated by ko1 (Koichi Sasada) over 10 years ago

  • Category set to core
  • Assignee set to matz (Yukihiro Matsumoto)

If we do not want to overload the existing .freeze method in this way, we could follow suggestions in http://bugs.ruby-lang.org/issues/8977 to add a new "frozen" method (or some other name) that the compiler would understand.

I like this idea. No compatibility issue.

(except naming issue)

Updated by calebthompson (Caleb Thompson) over 10 years ago

Optimizing the #freeze or #frozen methods to do this make a lot more sense to me
than the ""f suffix to me.

I'm +1 on #frozen and preventing redefinition of that method, but I further
propose that we raise an error if redefining frozen rather than silently
ignoring it, which would be very surprising.

Updated by Anonymous over 10 years ago

+1 for optimized String#freeze as it will work on previous versions of Ruby:

I've written a patch implementing this feature: https://gist.github.com/charliesome/6836600

Updated by headius (Charles Nutter) over 10 years ago

=begin
I am leaning toward #frozen if we want a new name... as in "give me the frozen version of this string". I know that there was some concern that "frozen" was too similar to "freeze" in http://bugs.ruby-lang.org/issues/8977 but it still feels like the best name.

If we can't do a new name that refers to freezing, I'd rather just stick with .freeze.

And I still kinda like ~@ :-)

class String
alias ~@ freeze
end

...although there's operator precedence problems (puts ~"str".frozen? does not work).
=end

Updated by Eregon (Benoit Daloze) over 10 years ago

I am happy to see another discussion on this, I feel "str"f is just a hack.

I strongly agree and think #freeze is the right name.

On the aesthetics side, I personally dislike prefix/suffix forms, they feel like u'str' in python which just makes me think it does not support the right strings by default. I kind of like %f{ ... } but #freeze fits even more in Ruby I think, even if a bit long but at least its semantics are clear.

Updated by normalperson (Eric Wong) over 10 years ago

"headius (Charles Nutter)" wrote:

I propose that we eliminate the new "f" suffix and just make the compiler
smart enough to see literal strings with .frozen the same way.

So this code:

str = "mystring".freeze

Would be equivalent in the compiler to this code:

str = "mystring"f

And the fstring table would still be used to return pooled instances.

This is a great idea IMHO. The backwards compatibility is a huge win
and I think this is the best idea so far regarding frozen strings.

IMPLEMENTATION NOTES:

The fstring table already exists on master and would be used for these
pooled strings. An open question is whether the compiler should
forever optimize "str".frozen to return the pooled version or whether
it should check (inline-cache style) whether String#freeze has been
replaced. I am ok with either, but the best potential comes from
ignoring String#freeze redefinitions...or making it impossible to
redefine String#freeze.

Initially (a few minutes ago), I thought it'd be better to inline-cache
to minimize surprise/keep compatibility. And maybe spew a loud warning on
String#freeze redefinition.

But thinking about this more, string literals are already special.
String#initialize is already ignored for string literals, so perhaps
#freeze may be unredefinedable, as well.

If we do not want to overload the existing .freeze method in this way,
we could follow suggestions in http://bugs.ruby-lang.org/issues/8977
to add a new "frozen" method (or some other name) that the compiler
would understand.

I think having only .freeze is better (especially for compatibility) and
a new .frozen method would be of minimal benefit.
(But you know Ruby far better than I do)

I also propose one alternative method name: the unary ~ operator.

There is no ~ on String right now, and it has no meaning for strings that we'd be overriding. So the following two lines would be equivalent:

str = ~"mystring"
str = "mystring"f

I hate this alternative "name". It's too ambiguous/confusing
(consider: ~"username" or /regexp/ =~ ~"foo"),
not backwards compatible, and hard to search for documentation on.

I think a beginner new to the language would be very confused by this.
You have my strong support of #freeze, but my strongest disapproval of ~

JUSTIFICATION:

Making the compiler aware of normal method-based String freezing has the following advantages:

  • It will parse in all versions of Ruby.
  • It will be equivalent in all versions of Ruby other than the fstring pooling.
  • It extends neatly to Array and Hash; the compiler can see Array or Hash with literal elements and return the same object.
  • It does not require a pragma (http://bugs.ruby-lang.org/issues/8976 )
  • It looks like Ruby.

Amen!

Updated by cibernox (Miguel Camba) over 10 years ago

I have another idea What about wrap strings in double backquotes or accents?

"this is an interpolable string"
'this is an uninterpolable string'
this is a frozen string => double backquoute
´this is also a frozen string´ => simple accent

Accents looks elegant, but I don't know if they are cumbersome in some keyboard distributions and I am not sure if they al limited to UTF8 code, but double backquote seems easy to add to the parser and feels pretty natural.

Whatever character(s) you like the most, (string, ´sting´, ^string^, ~string~, \string) I like the idea to be used as a wrapper, not adding a special symbol before or after the string definition.

Updated by headius (Charles Nutter) over 10 years ago

cibernox (Miguel Camba) wrote:

I have another idea What about wrap strings in double backquotes or accents?

"this is an interpolable string"
'this is an uninterpolable string'
this is a frozen string => double backquoute

I don't think the parser can distinguish this from normal backquotes that are empty...but there's no value to supporting empty shelling-out backquotes, so maybe it's not hard to special-case it.

if backquotes_contains_content
parse_as_normal_backquote
else
parse_as_frozen_string
end

Or something :-D

´this is also a frozen string´ => simple accent

Not on standard US keyboards, so this will never fly.

Whatever character(s) you like the most, (string, ´sting´, ^string^, ~string~, \string) I like the idea to be used as a wrapper, not adding a special symbol before or after the string definition.

Agreed. I still like just adding smarts for .freeze over any of the other options.

Updated by cibernox (Miguel Camba) over 10 years ago

Yep, we can discard the accent.

The double pipe can be tricky is you want to set a frozen string as default value of a block argument.
The double ^ has confict with the xor operator.
The double ~ has conflict with the complement operator
The double backslash seems fancy, like an opposite of regex, but since can also be used to break lines and escape characters. And probably is more difficult to parse . You tell me.

I know nothing about the parser internals, but the double backquote feels like a string and seems reasonable easy to implement. Seems the best option.

This idea can just coexist with the .freeze method. Is a bit of syntax sugar.

Updated by Anonymous over 10 years ago

I feel like we're getting a bit off topic in this thread.

The main benefit of optimizing String#freeze is maintaining backwards compatibility with older Rubies. Think of it as a kind of 'progressive enhancement'.

Introducing new syntax or a new method completely removes any backwards compatibility benefits and will mean that most Ruby code will not be able to use this syntax for quite some time.

Updated by sam.saffron (Sam Saffron) over 10 years ago

I am actually very concerned about compiler tricks with freeze cause it leads to non-obvious code.

x = "hello".freeze
y = "hello".freeze
x.object_id

10
x.object_id == y.object_id

a = "hello"
a.object_id

100
a.freeze
a.object_id
100 # must be 100


So the way #freeze operates then depends on where it is being executed, I dislike that.

Much prefer just adding #frozen, we can implement it sort of cleanly in 2.0 (except for GC hooking) and simply alias #freeze in 1.9 and earlier.

Updated by headius (Charles Nutter) over 10 years ago

sam.saffron (Sam Saffron) wrote:

I am actually very concerned about compiler tricks with freeze cause it leads to non-obvious code.

x = "hello".freeze
y = "hello".freeze
x.object_id

10
x.object_id == y.object_id

I don't think you should ever rely on this to be true, since it won't be on older Ruby impls or impls that don't yet have #freeze optimizations. Even Java, with its interned Strings, strongly discourages ever using object identity to compare strings. IDEs even flag it as a warning.

a = "hello"
a.object_id

100
a.freeze
a.object_id
100 # must be 100
...
So the way #freeze operates then depends on where it is being executed, I dislike that.

Much prefer just adding #frozen, we can implement it sort of cleanly in 2.0 (except for GC hooking) and simply alias #freeze in 1.9 and earlier.

So here's the same question I asked in the #frozen feature: why can't #freeze just use the fstring table?

  • fstrings will GC and clear themselves from that table
  • large strings put into the table will take up no more space than if they were not frozen

So #freeze could do what you suggest here and always use the fstring table. In the "literal".freeze case, the compiler could do additional magic to go to the table immediately.

Updated by normalperson (Eric Wong) over 10 years ago

"headius (Charles Nutter)" wrote:

So here's the same question I asked in the #frozen feature: why can't
#freeze just use the fstring table?

That would be an interesting experiment. After all, it is #freeze and
not #freeze!, so maybe we have some leverage there.

  • fstrings will GC and clear themselves from that table

I think this needs some work for the non-parser case, there seems to
be a bad interaction with lazy sweep. My analysis of my failed
patch for Feature #8998:
http://mid.gmane.org/20131009021547.GA1839@dcvr.yhbt.net

I also get (identical?) segfaults with the following:

diff --git a/object.c b/object.c
--- a/object.c
+++ b/object.c
@@ -1029,6 +1029,8 @@ VALUE
rb_obj_freeze(VALUE obj)
{
if (!OBJ_FROZEN(obj)) {

  • if (TYPE(obj) == T_STRING)
  •  return rb_fstring(obj);
    
    OBJ_FREEZE(obj);
    if (SPECIAL_CONST_P(obj)) {
    if (!immediate_frozen_tbl) {

Updated by sam.saffron (Sam Saffron) over 10 years ago

@hedius

What happens when a string pointer leaks out to a c extension?

Updated by headius (Charles Nutter) over 10 years ago

sam.saffron (Sam Saffron) wrote:

@hedius

What happens when a string pointer leaks out to a c extension?

This question applies equally to "str"f logic. I'm not sure what the answer is, because I don't know how frozen strings in @charliesome's patch interact with C extensions.

I don't think making freeze act like Java's String#intern has any better or worse interaction with C extensions than normal "str"f fstrings interact with C extensions.

Updated by headius (Charles Nutter) over 10 years ago

headius (Charles Nutter) wrote:

This question applies equally to "str"f logic. I'm not sure what the answer is, because I don't know how frozen strings in @charliesome's patch interact with C extensions.

Actually, it occurred to me that the interaction with C extensions is actually even simpler; if a string leaks out to C exts, it's no worse than any string leaking out. The only difference is that the C ext would still have a reference while the fstring table does not. So I think it's no worse than current interaction with strings.

Updated by headius (Charles Nutter) over 10 years ago

normalperson (Eric Wong) wrote:

"headius (Charles Nutter)" wrote:

So here's the same question I asked in the #frozen feature: why can't
#freeze just use the fstring table?

That would be an interesting experiment. After all, it is #freeze and
not #freeze!, so maybe we have some leverage there.

I think we do. The worst case scenario is that while referenced we have more entries in the table, which may include strings that become "shady" and pass out to C exts. But those strings would stay alive under the current definition of "shady" and even under older Ruby versions with a purely conservative GC the effects are no worse.

So basically:

  • If the string is long lived normally, it will take up X bytes for its lifetime.
  • If the string gets stored in the fstring table, it will last no longer than it would without the fstring table.
  • If the string is short-lived, it will have a bit more overhead for dealing with fstring table, but very little; hash calculation and table management at most.

It seems acceptable to have #freeze basically be Java's #intern.

  • fstrings will GC and clear themselves from that table

I think this needs some work for the non-parser case, there seems to
be a bad interaction with lazy sweep. My analysis of my failed
patch for Feature #8998:
http://mid.gmane.org/20131009021547.GA1839@dcvr.yhbt.net

I don't doubt your analysis, but I don't think it's any worse with #freeze using fstring table. It's just multiplied by the number of strings that get frozen. Critical failure * N is still a critical failure.

I also get (identical?) segfaults with the following:

diff --git a/object.c b/object.c
--- a/object.c
+++ b/object.c
@@ -1029,6 +1029,8 @@ VALUE
rb_obj_freeze(VALUE obj)
{
if (!OBJ_FROZEN(obj)) {

  • if (TYPE(obj) == T_STRING)
  •    return rb_fstring(obj);
    
    OBJ_FREEZE(obj);
    if (SPECIAL_CONST_P(obj)) {
    if (!immediate_frozen_tbl) {

Same argument here, I think.

Updated by sam.saffron (Sam Saffron) over 10 years ago

@hedius

There are 3 things being discussed here, I think it is fairly important we split them out.

  1. Parser optimisation for "string".freeze
  2. Unconditionally have #freeze return a pooled string
  3. Change the semantics of #freeze so it amends the current object and operates like .NET / Java intern does.
  1. is completely doable with little side-effects. My caveat is that if #1 is the only thing done, the semantics for #freeze depend on the invocation. That said, this is minor. I totally accept that and prefer "string".freeze to "string"f.

  2. without 3) really scares me.

Imagine the odd semantics:

a = "hello"
a.freeze # freezes one RVALUE in memory and returns a different RVALUE

As to 3) I don't think it can be implemented in MRI. If an RVALUE is moved in memory, MRI is going to have to crawl the heap and rewrite all the RVALUE that hold a ref to it, it does not keep track of this internally.

@charliesome thoughts?

Updated by nobu (Nobuyoshi Nakada) over 10 years ago

(13/10/14 8:40), sam.saffron (Sam Saffron) wrote:

Issue #8992 has been updated by sam.saffron (Sam Saffron).

Thank you for summarizing.

There are 3 things being discussed here, I think it is fairly important we split them out.

  1. Parser optimisation for "string".freeze
  1. is completely doable with little side-effects. My caveat is that if #1 is the only thing done, the semantics for #freeze depend on the invocation. That said, this is minor. I totally accept that and prefer "string".freeze to "string"f.

It's a part of byte-code optimization, not parser. Since we have done
it already for several methods, no problem there.

  1. Unconditionally have #freeze return a pooled string
  2. Change the semantics of #freeze so it amends the current object and operates like .NET / Java intern does.
  1. without 3) really scares me.

Imagine the odd semantics:

a = "hello"
a.freeze # freezes one RVALUE in memory and returns a different RVALUE

As to 3) I don't think it can be implemented in MRI. If an RVALUE is moved in memory, MRI is going to have to crawl the heap and rewrite all the RVALUE that hold a ref to it, it does not keep track of this internally.

Totally agree.

2+3 is wrong idea, I think.

Updated by headius (Charles Nutter) over 10 years ago

Ok, let's just focus on #1 for now...

It seems like everyone agrees that "string".freeze is a better choice than adding incompatible syntax now. That was the original proposal in this issue.

Should we remove "string"f on master and replace it with charliesome's patch for "string".freeze? Or do we want to bikeshed a shorter name?

It occurred to me the there's already "string".b which returns a binary string. Should we consider "string".f which is similar to "string"f syntax but is just a normal method?

I think we're in agreement that we want the method format rather than the "f" suffix, so it's just a matter of deciding if we want a different method name for the new compiler-aware method.

Updated by phluid61 (Matthew Kerwin) over 10 years ago

On Oct 14, 2013 5:00 PM, "headius (Charles Nutter)"
wrote:

Ok, let's just focus on #1 for now...

It seems like everyone agrees that "string".freeze is a better choice
than adding incompatible syntax now. That was the original proposal in this
issue.

Should we remove "string"f on master and replace it with charliesome's
patch for "string".freeze? Or do we want to bikeshed a shorter name?

It occurred to me the there's already "string".b which returns a binary
string. Should we consider "string".f which is similar to "string"f syntax
but is just a normal method?

I think we're in agreement that we want the method format rather than the
"f" suffix, so it's just a matter of deciding if we want a different method
name for the new compiler-aware method.

Yes. I feel like regexen have suffixes because of decades of perl
precedence, but they (suffixes) don't belong anywhere else.

For a method, I feel like #freeze is the better name, my only question is:
is anyone monkeypatching it (and therefore will be bitten by this
optimisation)? I doubt it, but we should still ask. The same question would
have to be asked of the new method; I think there's more chance of #f being
used in the wild than an overridden #freeze

Also in the favour of #freeze, it gives existing code a boost without any
modification.

Sent from my phone, so excuse the typos.

Updated by normalperson (Eric Wong) over 10 years ago

Nobuyoshi Nakada wrote:

(13/10/14 8:40), sam.saffron (Sam Saffron) wrote:

Issue #8992 has been updated by sam.saffron (Sam Saffron).

Thank you for summarizing.

There are 3 things being discussed here, I think it is fairly important we split them out.

  1. Parser optimisation for "string".freeze
  1. is completely doable with little side-effects. My caveat is that if #1 is the only thing done, the semantics for #freeze depend on the invocation. That said, this is minor. I totally accept that and prefer "string".freeze to "string"f.

It's a part of byte-code optimization, not parser. Since we have done
it already for several methods, no problem there.

So can we move this optimization to the parser instead? I think Sam means:

# optimized literal by parser.  This may use pooled string since
# the string never existed in ObjectSpace before this line of code:
"string".freeze

However, if .freeze operates on a variable (and not literal):

# leave this optimized by parser to avoid incompatibility
a = "string"
a.freeze => preserve existing behavior
  1. Unconditionally have #freeze return a pooled string
  2. Change the semantics of #freeze so it amends the current object and operates like .NET / Java intern does.
  1. without 3) really scares me.

Imagine the odd semantics:

a = "hello"
a.freeze # freezes one RVALUE in memory and returns a different RVALUE

Yes, this scares me if a.freeze made a different RVALUE

As to 3) I don't think it can be implemented in MRI. If an RVALUE is moved in memory, MRI is going to have to crawl the heap and rewrite all the RVALUE that hold a ref to it, it does not keep track of this internally.

Totally agree.

2+3 is wrong idea, I think.

Agreed.

Actions #23

Updated by Anonymous over 10 years ago

Totally agree with ssaffron on options 2 and 3 being scary. 3 is also impossible in MRI at the moment.

I'm happy to commit my patch optimizing #freeze on a literal string if it looks good.

I think we should move the discussion of removing f-suffix and perhaps adding String#f to another ticket.

Updated by sam.saffron (Sam Saffron) over 10 years ago

+1 for removing f suffix and amending the optimiser for #freeze

Updated by headius (Charles Nutter) over 10 years ago

I have added #9042 and #9043 for removing the "f" suffix and adding the #f method, respectively.

I'm starting to lean toward making #f be the only magic form, so nobody can complain that we're adding incompatible syntax ("f" suffix) or changing the semantics of an existing method (#freeze optimization).

Updated by Anonymous over 10 years ago

I'm starting to lean toward making #f be the only magic form, so nobody can complain that we're ... changing the semantics of an existing method (#freeze optimization).

I don't get this argument. Optimized String#freeze doesn't really change semantics in any real way. I'm happy to just ignore anyone that complains about optimizing #freeze on a string literal.

There's only one way this could possibly affect any Ruby code - and that's if the code inspects the object_id of literal strings that it immediately calls #freeze on. I'd say any code that breaks due to this was already fairly brittle anyway.

Updated by Eregon (Benoit Daloze) over 10 years ago

I agree having optimized #freeze is better than #f.

Updated by akr (Akira Tanaka) over 10 years ago

2013/10/8 headius (Charles Nutter) :

Feature #8992: Use String#freeze and compiler tricks to replace "str"f suffix
https://bugs.ruby-lang.org/issues/8992

Does anyone measure actual performance benefit?

Tanaka Akira

Updated by normalperson (Eric Wong) over 10 years ago

Tanaka Akira wrote:

2013/10/8 headius (Charles Nutter) :

Feature #8992: Use String#freeze and compiler tricks to replace "str"f suffix
https://bugs.ruby-lang.org/issues/8992

Does anyone measure actual performance benefit?

Not directly, but I wasn't able to come up with performance benefits
from my patch for https://bugs.ruby-lang.org/issues/8998
However, I don't have real apps which depend on hash/string-keys
performance.

Perhaps other improvements (RGenGC) make it less useful.

Updated by ko1 (Koichi Sasada) over 10 years ago

Matz, could you conclude this ticket?

I like this idea because:

  • No syntax change
  • Semantics was changed ("literal".freeze.object_id => anytime same),
    but I can't imagine the apps which rely on this behavior.
  • Except 2nd point, this is no compatibility issue.

(2013/10/08 2:56), headius (Charles Nutter) wrote:

Issue #8992 has been reported by headius (Charles Nutter).


Feature #8992: Use String#freeze and compiler tricks to replace "str"f suffix
https://bugs.ruby-lang.org/issues/8992

--
// SASADA Koichi at atdot dot net

Updated by matz (Yukihiro Matsumoto) over 10 years ago

  • I am OK with adding a new method (e.g. String#f) and compiler trick.
  • I hesitate a bit to change String#freeze semantics.
  • String#~ is very concise and in that sense attractive, but relation of '~' with patterns may cause confusion.

So, my conclusion is adding String#f with compiler trick. I still need some discussion for changing String#freeze.
Note that I don't have strong objection against String#freeze.

Matz.

Updated by mame (Yusuke Endoh) over 10 years ago

matz (Yukihiro Matsumoto) wrote:

  • I am OK with adding a new method (e.g. String#f) and compiler trick.

Somewhat oppose against "f" with compiler trick.

"f" is a typical meta-syntactic variable of functions.
I guess some programs have the following code.

class String
def f
...
end
end

In fact, I've found some instances in a few minutes:

http://blog.goo.ne.jp/ruby-index/e/91995e39fe1fe72d339d882cf5ded055
http://python.g.hatena.ne.jp/mhrs/20060530

--
Yusuke Endoh

Updated by Anonymous over 10 years ago

Matz, I believe adding a new method will significantly limit the usefulness of this feature.

The reason optimizing "".freeze is superior to a new syntax is that gems can start using this feature now without dropping support for versions < 2.1.

I think many Gem authors will be very hesitant to monkey patch String#f to String#freeze.

Updated by Anonymous over 10 years ago

@mame (Yusuke Endoh): Good catch. How about String#fz then?

Updated by headius (Charles Nutter) over 10 years ago

matz: Thank you for weighing in. I think we are on the same page.

There are a few different versions of this proposal getting tossed around, so I'll try to summarize the key points.

  1. We would like a compiler trick to replace the "str"f suffix.
  2. The compiler trick would apply to literal strings with immediate calls to a freeze method and behave as "str"f does today.
  3. The freeze method optimized by the compiler will be #freeze or a new shorter-named method #f or both.
  4. Calls to #freeze against non-literals would behave as now.
  5. Calls to #f, would use the fstring functionality to return a globally cached string (similar to interning in Java, from #8997)

It sounds like almost everyone is in favor of eliminating the "str"f suffix, so I think we can take that as written.

Decisions to make:

  1. Do we add a new method that can do #8977 behavior?
    1a. What do we call it?
  2. If we do not add a new method, should compiler tricks also optimize "string".freeze?

I think the following is a good plan:

  • If we all agree that "string"f should go away, then we should switch to "string".freeze compiler trick asap. @charliesome provided a patch.
  • If we want another method, we should decide on a name and decide whether "string".freeze should still have compiler trick.

Updated by Anonymous over 10 years ago

At RubyConf Matz approved optimizing #freeze on static string literals and removing f-suffix.

Updated by Eregon (Benoit Daloze) over 10 years ago

charliesome (Charlie Somerville) wrote:

At RubyConf Matz approved optimizing #freeze on static string literals and removing f-suffix.

Great!

Updated by headius (Charles Nutter) over 10 years ago

  • Status changed from Open to Closed

This has been completed as of revisions r43627 and r43634.

Thanks for the collaboration, everyone :-)

Updated by Anonymous over 10 years ago

sawa: %-strings are already covered by #freeze optimization:

λ ./miniruby --dump=insns -e '%{}.freeze'
== disasm: <RubyVM::InstructionSequence:<main>@-e>======================
0000 trace            1                                               (   1)
0002 opt_str_freeze   ""
0004 leave

Updated by sawa (Tsuyoshi Sawada) over 10 years ago

charliesome (Charlie Somerville) wrote:

sawa: %-strings are already covered by #freeze optimization:

λ ./miniruby --dump=insns -e '%{}.freeze'
== disasm: <RubyVM::InstructionSequence:<main>@-e>======================
0000 trace            1                                               (   1)
0002 opt_str_freeze   ""
0004 leave

Thanks. I should have checked before. I deleted my previous post.

Updated by bughit (bug hit) almost 5 years ago

I recently had to remind myself what kind of magic is hiding behind 'str'.freeze and would like to provide some feedback.

While tactically this may have been justified, strategically this is a wrong direction. This approach moves the language towards a collection of magical spells. You end up with code that can't be understood through decomposition into primitives.

'str'.method1 - means construct a string object with a value 'str' and call method 'method1' on it.

The notion that a method call should have time travel powers and be able to influence how its already constructed receiver is allocated is absurd. It makes mockery of such fundamental concepts as expression evaluation order and method invocation. Perhaps this is already being done elsewhere, not sure, but that's not a justification, doubling down on nonsense does not make it sense.

Whereas special syntax (whatever it is) does not have this flaw. You get to define what new syntax means without corrupting existing concepts, and in the long run this is more valuable than helping gem authors transition to immutable strings.

Actions #42

Updated by byroot (Jean Boussier) about 1 month ago

  • Related to Misc #20222: Dedup-ing clarification added
Actions

Also available in: Atom PDF

Like0
Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0Like0