Bug 115905 – Do not copy the HashMap template for use with RefPtr keys

• Component: Web Template Framework
• Status: RESOLVED DUPLICATE

[Mikhail Pozdnyakov]

SSIA. At the moment specialization of HashMap for use with RefPtr keys contains the whole copy of generic HashMap class, which is bad.

[Mikhail Pozdnyakov]

Created attachment 201343 [details]
WIP

[Mikhail Pozdnyakov]

Comment on attachment 201343 [details]
WIP

View in context: https://bugs.webkit.org/attachment.cgi?id=201343&action=review

> Source/WTF/wtf/RefPtrHashMap.h:151
> +    inline void HashMapBase<RefPtr<T>, U, V, W, X>::remove(RawKeyType key)

argh! it will be probably hidden with HashMap::remove(iterator it)..

[Darin Adler]

(In reply to comment #2)
> > Source/WTF/wtf/RefPtrHashMap.h:151
> > +    inline void HashMapBase<RefPtr<T>, U, V, W, X>::remove(RawKeyType key)
> 
> argh! it will be probably hidden with HashMap::remove(iterator it)..

You should be able to fix that with "using".

[Darin Adler]

(In reply to comment #3)
> (In reply to comment #2)
> > > Source/WTF/wtf/RefPtrHashMap.h:151
> > > +    inline void HashMapBase<RefPtr<T>, U, V, W, X>::remove(RawKeyType key)
> > 
> > argh! it will be probably hidden with HashMap::remove(iterator it)..
> 
> You should be able to fix that with "using".

And I think there’s even a way to delete it in C++11 compilers.

[Mikhail Pozdnyakov]

Created attachment 201483 [details]
WIP 2

Another approach, the specialization is inherited from generic HashMap, providing original traits templates arguments.

[Darin Adler]

I like the idea of eliminating this silly duplication. I think I liked your first attempt better than the second, not sure. I’m hoping we can use this for other types, such as RetainPtr, once we make it more rational.

[Mikhail Pozdnyakov]

Created attachment 201560 [details]
WIP 3

Combination of 1st and 2nd approaches, also made it re-usable for other key types, like RetainPtr.

[WebKit Commit Bot]

Attachment 201560 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/HashTraits.h', u'Source/WTF/wtf/RefPtrHashMap.h']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:29:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Total errors found: 1 in 3 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Build Bot]

Comment on attachment 201560 [details]
WIP 3

Attachment 201560 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/346957

New failing tests:
http/tests/security/same-origin-document-domain-storage-allowed.html
fast/loader/stateobjects/replacestate-in-iframe.html
storage/domstorage/sessionstorage/window-open.html

[Build Bot]

Created attachment 201655 [details]
Archive of layout-test-results from webkit-ews-15 for mac-mountainlion-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: webkit-ews-15  Port: mac-mountainlion-wk2  Platform: Mac OS X 10.8.2

[Mikhail Pozdnyakov]

(In reply to comment #9)
> (From update of attachment 201560 [details])
> Attachment 201560 [details] did not pass mac-wk2-ews (mac-wk2):
> Output: http://webkit-queues.appspot.com/results/346957
> 
> New failing tests:
> http/tests/security/same-origin-document-domain-storage-allowed.html
> fast/loader/stateobjects/replacestate-in-iframe.html
> storage/domstorage/sessionstorage/window-open.html

that looks strange..

[Mikhail Pozdnyakov]

Created attachment 201684 [details]
WIP 3 run EWS

[WebKit Commit Bot]

Attachment 201684 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/HashTraits.h', u'Source/WTF/wtf/RefPtrHashMap.h']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:29:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Total errors found: 1 in 3 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Build Bot]

Comment on attachment 201684 [details]
WIP 3 run EWS

Attachment 201684 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/477020

New failing tests:
http/tests/security/same-origin-document-domain-storage-allowed.html
fast/loader/stateobjects/replacestate-in-iframe.html
storage/domstorage/sessionstorage/window-open.html

[Build Bot]

Created attachment 201692 [details]
Archive of layout-test-results from webkit-ews-09 for mac-mountainlion-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: webkit-ews-09  Port: mac-mountainlion-wk2  Platform: Mac OS X 10.8.2

[Mikhail Pozdnyakov]

(In reply to comment #15)
> Created an attachment (id=201692) [details]
> Archive of layout-test-results from webkit-ews-09 for mac-mountainlion-wk2
> 
> The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
> Bot: webkit-ews-09  Port: mac-mountainlion-wk2  Platform: Mac OS X 10.8.2

ok, ok.. found a stupid mistake in my patch: 
AddResult add(const KeyType& key, MappedPassInType mapped) { return BaseType::set(key, mapped); } -> AddResult add(const KeyType& key, MappedPassInType mapped) { return BaseType::add(key, mapped); }

[Mikhail Pozdnyakov]

Created attachment 201702 [details]
patch

Fixed couple of stupid mistakes (see comment above), re-arranged methods definitions accordingly to their declaration order, added change log.

[WebKit Commit Bot]

Attachment 201702 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/ChangeLog', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/HashTraits.h', u'Source/WTF/wtf/RefPtrHashMap.h']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:29:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Total errors found: 1 in 4 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Darin Adler]

Comment on attachment 201702 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=201702&action=review

I worry that these make it a harder to see what set of functions is supported by HashMap.

> Source/WTF/ChangeLog:61
> +        * wtf/HashMap.h:
> +        (WTF):
> +        (HashMap):
> +        (WTF::HashMap::keys):
> +        (WTF::HashMap::values):
> +        (WTF::HashMap::impl):
> +        (WTF::::find):
> +        (WTF::::contains):
> +        (WTF::::get):
> +        (WTF::::set):
> +        (WTF::::add):
> +        (WTF::::remove):
> +        (WTF::::take):
> +        (WTF::::inlineAdd):
> +        (WTF::::swap):
> +        (WTF::::size):
> +        (WTF::::capacity):
> +        (WTF::::isEmpty):
> +        (WTF::::begin):
> +        (WTF::::end):
> +        (WTF::::clear):
> +        (WTF::::checkConsistency):
> +        * wtf/HashTraits.h:
> +        (GenericHashTraits):
> +        * wtf/RefPtrHashMap.h:
> +        (WTF):
> +        (WTF::::find):
> +        (WTF::::contains):
> +        (WTF::::inlineAdd):
> +        (WTF::::set):
> +        (WTF::::add):
> +        (WTF::::inlineGet):
> +        (WTF::::get):
> +        (WTF::::remove):
> +        (WTF::::take):

The prepare-ChangeLog script completely screwed this up. Please edit by hand so that it’s a useful list of functions, or delete the functions entirely. Either would be better than this near-gibberish the script produced.

> Source/WTF/wtf/HashMap.h:128
> +    private:
> +        typedef HashMapBase<KeyArg, MappedArg, HashArg, KeyTraitsArg, MappedTraitsArg, KeyTraitsArg::traitTypesCount> BaseType;
> +        typedef typename BaseType::KeyTraits KeyTraits;
> +        typedef typename BaseType::MappedTraits MappedTraits;
> +        typedef typename BaseType::ValueTraits ValueTraits;

Are all these typedefs needed? We should omit any that are not needed.

> Source/WTF/wtf/HashMap.h:133
> +        typedef typename BaseType::KeyType KeyType;
> +        typedef typename BaseType::MappedType MappedType;
> +        typedef typename BaseType::ValueType ValueType;

Are these typedefs needed? If these types are public and in the base class, why do we need to re-typedef them? Could we omit them or make use of “using” instead?

> Source/WTF/wtf/HashMap.h:145
> +    private:
> +        typedef typename BaseType::MappedPassInType MappedPassInType;
> +        typedef typename BaseType::MappedPassOutType MappedPassOutType;
> +        typedef typename BaseType::MappedPeekType MappedPeekType;
> +        typedef typename BaseType::MappedPassInReferenceType MappedPassInReferenceType;
> +
> +        typedef typename BaseType::HashFunctions HashFunctions;
> +        typedef typename BaseType::HashTableType HashTableType;
> +
> +        class HashMapKeysProxy;
> +        class HashMapValuesProxy;

Are all these typedefs and declarations needed? We should omit any that are not needed.

> Source/WTF/wtf/HashMap.h:150
> +        typedef typename BaseType::iterator iterator;
> +        typedef typename BaseType::const_iterator const_iterator;
> +        typedef typename BaseType::AddResult AddResult;

Are these typedefs needed? If these types are public and in the base class, why do we need to re-typedef them? Could we omit them or make use of “using” instead?

> Source/WTF/wtf/HashMap.h:169
> +        HashMapKeysProxy& keys() { return static_cast<HashMapKeysProxy&>(*this); }
> +        const HashMapKeysProxy& keys() const { return static_cast<const HashMapKeysProxy&>(*this); }
> +
> +        HashMapValuesProxy& values() { return static_cast<HashMapValuesProxy&>(*this); }
> +        const HashMapValuesProxy& values() const { return static_cast<const HashMapValuesProxy&>(*this); }

Why is a cast needed? Won’t these all work without an explicit static_cast?

[Mikhail Pozdnyakov]

(In reply to comment #19)
> (From update of attachment 201702 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=201702&action=review
> 
> I worry that these make it a harder to see what set of functions is supported by HashMap.
yes, but we gain some flexibility in the class interface..
we have already examples in WTF when the User needs to look at both base and child class in order to see the whole picture (for instance RefCounted, RefCountedBase). 
Moreover the current situation with having HashMap<RefPtr<>, > specialization
as a separate class does not make life any better.

> 
> > Source/WTF/ChangeLog:61
> > +        * wtf/HashMap.h:
> > +        (WTF):
> > +        (HashMap):
> > +        (WTF::HashMap::keys):
> > +        (WTF::HashMap::values):
> > +        (WTF::HashMap::impl):
> > +        (WTF::::find):
> > +        (WTF::::contains):
> > +        (WTF::::get):
> > +        (WTF::::set):
> > +        (WTF::::add):
> > +        (WTF::::remove):
> > +        (WTF::::take):
> > +        (WTF::::inlineAdd):
> > +        (WTF::::swap):
> > +        (WTF::::size):
> > +        (WTF::::capacity):
> > +        (WTF::::isEmpty):
> > +        (WTF::::begin):
> > +        (WTF::::end):
> > +        (WTF::::clear):
> > +        (WTF::::checkConsistency):
> > +        * wtf/HashTraits.h:
> > +        (GenericHashTraits):
> > +        * wtf/RefPtrHashMap.h:
> > +        (WTF):
> > +        (WTF::::find):
> > +        (WTF::::contains):
> > +        (WTF::::inlineAdd):
> > +        (WTF::::set):
> > +        (WTF::::add):
> > +        (WTF::::inlineGet):
> > +        (WTF::::get):
> > +        (WTF::::remove):
> > +        (WTF::::take):
> 
> The prepare-ChangeLog script completely screwed this up. Please edit by hand so that it’s a useful list of functions, or delete the functions entirely. Either would be better than this near-gibberish the script produced.
> 
Sure thing.
> > Source/WTF/wtf/HashMap.h:128
> > +    private:
> > +        typedef HashMapBase<KeyArg, MappedArg, HashArg, KeyTraitsArg, MappedTraitsArg, KeyTraitsArg::traitTypesCount> BaseType;
> > +        typedef typename BaseType::KeyTraits KeyTraits;
> > +        typedef typename BaseType::MappedTraits MappedTraits;
> > +        typedef typename BaseType::ValueTraits ValueTraits;
> 
> Are all these typedefs needed? We should omit any that are not needed.
I guess some of them can be omitted indeed.
> 
> > Source/WTF/wtf/HashMap.h:133
> > +        typedef typename BaseType::KeyType KeyType;
> > +        typedef typename BaseType::MappedType MappedType;
> > +        typedef typename BaseType::ValueType ValueType;
> 
> Are these typedefs needed? If these types are public and in the base class, why do we need to re-typedef them? 
For convenience, otherwise we'd need something like BaseType::KeyType in every place where they are used.

Could we omit them or make use of “using” instead?
mm.. would "using" give a more benefit here comparing to "typedef"? 
As far as I know "using" has the same semantics as if it were introduced by the typedef specifier, and the benefit of "using" is that it makes possible to "typedef" template classes which however is not the case here.
> 
> > Source/WTF/wtf/HashMap.h:145
> > +    private:
> > +        typedef typename BaseType::MappedPassInType MappedPassInType;
> > +        typedef typename BaseType::MappedPassOutType MappedPassOutType;
> > +        typedef typename BaseType::MappedPeekType MappedPeekType;
> > +        typedef typename BaseType::MappedPassInReferenceType MappedPassInReferenceType;
> > +
> > +        typedef typename BaseType::HashFunctions HashFunctions;
> > +        typedef typename BaseType::HashTableType HashTableType;
> > +
> > +        class HashMapKeysProxy;
> > +        class HashMapValuesProxy;
> 
> Are all these typedefs and declarations needed? We should omit any that are not needed.
> 
> > Source/WTF/wtf/HashMap.h:150
> > +        typedef typename BaseType::iterator iterator;
> > +        typedef typename BaseType::const_iterator const_iterator;
> > +        typedef typename BaseType::AddResult AddResult;
> 
> Are these typedefs needed? If these types are public and in the base class, why do we need to re-typedef them? Could we omit them or make use of “using” instead?
> 
> > Source/WTF/wtf/HashMap.h:169
> > +        HashMapKeysProxy& keys() { return static_cast<HashMapKeysProxy&>(*this); }
> > +        const HashMapKeysProxy& keys() const { return static_cast<const HashMapKeysProxy&>(*this); }
> > +
> > +        HashMapValuesProxy& values() { return static_cast<HashMapValuesProxy&>(*this); }
> > +        const HashMapValuesProxy& values() const { return static_cast<const HashMapValuesProxy&>(*this); }
> 
> Why is a cast needed? Won’t these all work without an explicit static_cast?
That had been already there, and the cast is needed I guess as those classes are just inherited from HashMap, so the pointer to base class needs cast to be treated as a pointer to child class.

[Mikhail Pozdnyakov]

Created attachment 201752 [details]
patch v2

Fixed change log, removed "typedefs" of types which are not used frequently within their classes.

[WebKit Commit Bot]

Attachment 201752 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/ChangeLog', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/HashTraits.h', u'Source/WTF/wtf/RefPtrHashMap.h']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:29:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Total errors found: 1 in 4 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Mikhail Pozdnyakov]

(In reply to comment #20)
> (In reply to comment #19)
> > (From update of attachment 201702 [details] [details])
> > View in context: https://bugs.webkit.org/attachment.cgi?id=201702&action=review
> > 
> > I worry that these make it a harder to see what set of functions is supported by HashMap.
> yes, but we gain some flexibility in the class interface..
> we have already examples in WTF when the User needs to look at both base and child class in order to see the whole picture (for instance RefCounted, RefCountedBase). 
> Moreover the current situation with having HashMap<RefPtr<>, > specialization
> as a separate class does not make life any better.
> 
Think it's worth adding that the proposed approach can be quite easily applied to various cases when a supplementary key type would make sense, HashMap<RetainPtr, > was already mentioned, but this also can be used with HashMap<String, ..> (having const char* as another key type) and probably others.

[Mikhail Pozdnyakov]

Created attachment 201947 [details]
patch v3

Now HashMap is not split to HashMapBase and HashMap, so that it is easy to see all the existing methods.

[WebKit Commit Bot]

Attachment 201947 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/ChangeLog', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/RefPtrHashMap.h']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:26:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Source/WTF/wtf/HashMap.h:47:  When wrapping a line, only indent 4 spaces.  [whitespace/indent] [3]
Total errors found: 2 in 3 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Build Bot]

Comment on attachment 201947 [details]
patch v3

Attachment 201947 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/470927

[Mikhail Pozdnyakov]

Created attachment 201950 [details]
patch v4

Arranged HashMap method definitions in a correct order.

[WebKit Commit Bot]

Attachment 201950 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/ChangeLog', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/RefPtrHashMap.h']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:26:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Source/WTF/wtf/HashMap.h:47:  When wrapping a line, only indent 4 spaces.  [whitespace/indent] [3]
Total errors found: 2 in 3 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Build Bot]

Comment on attachment 201950 [details]
patch v4

Attachment 201950 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/487006

[Mikhail Pozdnyakov]

(In reply to comment #29)
> (From update of attachment 201950 [details])
> Attachment 201950 [details] did not pass mac-wk2-ews (mac-wk2):
> Output: http://webkit-queues.appspot.com/results/487006

Should tweak exported symbols for MAC apparently.. but still it would be nice to get reviewers feedback on the approach.

[Build Bot]

Comment on attachment 201950 [details]
patch v4

Attachment 201950 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.appspot.com/results/480644

[Mikhail Pozdnyakov]

Created attachment 202050 [details]
try to fix MAC build

[WebKit Commit Bot]

Attachment 202050 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/ChangeLog', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/RefPtrHashMap.h', u'Source/WebCore/WebCore.exp.in']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:26:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Source/WTF/wtf/HashMap.h:47:  When wrapping a line, only indent 4 spaces.  [whitespace/indent] [3]
Total errors found: 2 in 3 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Build Bot]

Comment on attachment 202050 [details]
try to fix MAC build

Attachment 202050 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/489232

[Build Bot]

Comment on attachment 202050 [details]
try to fix MAC build

Attachment 202050 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.appspot.com/results/492233

[Build Bot]

Comment on attachment 202050 [details]
try to fix MAC build

Attachment 202050 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.appspot.com/results/488284

[Mikhail Pozdnyakov]

Darin, could you please give your opinion on the latest approach?

[Darin Adler]

I think the approach is OK. It hadn’t occurred to me that the main feature of the RefPtr hash map is built-in support for an additional key type, raw pointers.

I think the followup question is what other key types might lead us to want support for additional key types. We talked about RetainPtr, but another obvious example is String and const char*.

HashSet’s approach to this is to allow explicit use of a hash translator for the find, contains, and add functions. Not supporting a hash translator for the remove function is a strange omission in HashSet, but I suppose we haven’t needed it yet.

The approach in RefPtr HashMap to automatically overload and translate raw pointers is more convenient than the hash translator approach, but less flexible since it only works for a single type. It’s also strange that a RefPtr HashSet requires a hash translator to work with raw pointers.

I’d like to see more unity between HashSet and HashMap on this. It might be cleaner if we could make it possible to use a translator with HashMap, and to have automatic support for additional key types be a feature of HashSet.

On the other hand, bloating both these class templates with these new features might be a mistake.

Another area we need to work on is to cut down on code size when these class templates are used. I think some of the Blink folks are interested in this because they did not like the fix I did for bug 115807 and would prefer that the code size be smaller for typical HashMap use.

[Mikhail Pozdnyakov]

(In reply to comment #38)
> I think the approach is OK. It hadn’t occurred to me that the main feature of the RefPtr hash map is built-in support for an additional key type, raw pointers.
> 
> I think the followup question is what other key types might lead us to want support for additional key types. We talked about RetainPtr, but another obvious example is String and const char*.
> 
> HashSet’s approach to this is to allow explicit use of a hash translator for the find, contains, and add functions. Not supporting a hash translator for the remove function is a strange omission in HashSet, but I suppose we haven’t needed it yet.
> 
> The approach in RefPtr HashMap to automatically overload and translate raw pointers is more convenient than the hash translator approach, but less flexible since it only works for a single type. It’s also strange that a RefPtr HashSet requires a hash translator to work with raw pointers.
> 
> I’d like to see more unity between HashSet and HashMap on this. It might be cleaner if we could make it possible to use a translator with HashMap, and to have automatic support for additional key types be a feature of HashSet.
> 
> On the other hand, bloating both these class templates with these new features might be a mistake.
> 
> Another area we need to work on is to cut down on code size when these class templates are used. I think some of the Blink folks are interested in this because they did not like the fix I did for bug 115807 and would prefer that the code size be smaller for typical HashMap use.

I could spread the proposed approach to HashSet as well, I also think it is more convenient to use than putting translator on caller side. As for flexibility, it is possible to create specializations for ExtraKeysCountArg==2 or 3 if needed.

[Mikhail Pozdnyakov]

Created attachment 204810 [details]
fixed MAC build

[WebKit Commit Bot]

Attachment 204810 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WTF/ChangeLog', u'Source/WTF/wtf/HashMap.h', u'Source/WTF/wtf/RefPtrHashMap.h', u'Source/WebCore/WebCore.exp.in']" exit_code: 1
Source/WTF/wtf/RefPtrHashMap.h:26:  Code inside a namespace should not be indented.  [whitespace/indent] [4]
Source/WTF/wtf/HashMap.h:47:  When wrapping a line, only indent 4 spaces.  [whitespace/indent] [3]
Total errors found: 2 in 4 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Mikhail Pozdnyakov]

Created attachment 206550 [details]
patch v7 (alternate approach)

Added alternate versions of HasMap interface methods which do not enforce converting key arguments to Key type but instead let HasMap's own HashTranslator to deal with them. Using of these functions do not usually require any extra effort on client side (e.g., explicit passing of used HashTranslator). Using of these methods is benefitial if Key type has some relative types that could be handled by HasMap's HashFunctions - objects of these types can be used as the key arguments, for example smart pointers and raw pointers of the same type or various string classes.

This approach is also much simpler than all the previous ones.

[Build Bot]

Comment on attachment 206550 [details]
patch v7 (alternate approach)

Attachment 206550 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/1066062

[Build Bot]

Comment on attachment 206550 [details]
patch v7 (alternate approach)

Attachment 206550 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.appspot.com/results/1066065

[Anders Carlsson]

Comment on attachment 206550 [details]
patch v7 (alternate approach)

Sam did this some time ago.

[Darin Adler]

He used bug 137200 to do it.

*** This bug has been marked as a duplicate of bug 137200 ***