I am glad to see that ruby1.9.0 preserves hash insertion order.
It will finally allow me to generate human-readable yaml dumps
without ugly workarounds.
Some questions:
(1) Can I rely upon hash insertion order preservation as an approved
new feature of Ruby, without any risk that it may get lost again in
future releases ?
(2) ri1.9 says:
"The order in which you traverse a hash by either key or value
may seem arbitrary, and will generally not be in the insertion order."
Do I understand correctly that this text is outdated ?
If the answer is two times yes, then read on, please:
(3) Would you agree that the almost undocumented yaml/omap
class is obsolete ? Maybe, one should clearly marked it as such,
or even exclude it from 1.9 ??
(4) Once hashes are ordered, new use cases will be found, and
new methods will become useful. I would like to propose the following
new instance methods
Hash#insert_before( existing_key, other_hash )
Hash#insert_after( existing_key, other_hash )
and possibly also
Hash#insert_before( existing_key, key, value )
Hash#insert_after( existing_key, key, value )
to insert new items at well defined locations. My use case, as I
said before, is a hash which is dumped into human-readable yaml.
In message "Re: Hash insertion order preservation in ruby1.9" on Mon, 31 Mar 2008 19:55:12 +0900, Joachim (München) <wuttke1@web.de> writes:
I am glad to see that ruby1.9.0 preserves hash insertion order.
It will finally allow me to generate human-readable yaml dumps
without ugly workarounds.
Some questions:
(1) Can I rely upon hash insertion order preservation as an approved
new feature of Ruby, without any risk that it may get lost again in
future releases ?
For CRuby, yes. For Ruby the language, we have to make up consensus
with other implementations.
At Mon, 31 Mar 2008 19:55:12 +0900,
Joachim (München) wrote in [ruby-talk:296442]:
(4) Once hashes are ordered, new use cases will be found, and
new methods will become useful. I would like to propose the following
new instance methods
Hash#insert_before( existing_key, other_hash )
Hash#insert_after( existing_key, other_hash )
and possibly also
Hash#insert_before( existing_key, key, value )
Hash#insert_after( existing_key, key, value )
to insert new items at well defined locations. My use case, as I
said before, is a hash which is dumped into human-readable yaml.
I think it is safe to say it will remain in JRuby as well. The new
behavior has allowed us to optimize hash iteration since it is now
entirely deterministic.
>(1) Can I rely upon hash insertion order preservation as an approved
>new feature of Ruby, without any risk that it may get lost again in
>future releases ?
For CRuby, yes. For Ruby the language, we have to make up consensus
with other implementations.
I see two possible conflicts:
- existing _key does not exist
- key already exists
Right?
I would raise an exception in both cases.
Rationale:
One could imagine several other ways to resolve conflicts.
I do not think one of them will be widely accepted as the obvious
right thing to do.
Therefore not attempting to reconcile conflicts will cause the least
surprise.
Regards, Joachim
···
On 2 Apr., 03:39, Nobuyoshi Nakada <n...@ruby-lang.org> wrote:
Hi,
At Mon, 31 Mar 2008 19:55:12 +0900,
Joachim (München) wrote in [ruby-talk:296442]:
> (4) Once hashes are ordered, new use cases will be found, and
> new methods will become useful. I would like to propose the following
> new instance methods
> Hash#insert_before( existing_key, other_hash )
> Hash#insert_after( existing_key, other_hash )
> and possibly also
> Hash#insert_before( existing_key, key, value )
> Hash#insert_after( existing_key, key, value )
> to insert new items at well defined locations. My use case, as I
> said before, is a hash which is dumped into human-readable yaml.
Cool! I've seen folks trip themselves up on this using 1.8.6 before because of an assumption in their mental model of the language. Nice to see this addressed.
···
Sent from my iPhone
On Mar 31, 2008, at 5:30 PM, MenTaLguY <mental@rydia.net> wrote:
>(1) Can I rely upon hash insertion order preservation as an approved
>new feature of Ruby, without any risk that it may get lost again in
>future releases ?
For CRuby, yes. For Ruby the language, we have to make up consensus
with other implementations.
I think it is safe to say it will remain in JRuby as well. The new
behavior has allowed us to optimize hash iteration since it is now
entirely deterministic.