Confusion Over Keyword Arguments


#1

Ruby 2.0 will include new syntax for hash literals: {a:3, b:4}. However,
one can leave off the {}s to create a hash. Current software uses this
for “faked” keyword arguments.

def my_meth(options={})
end

(1) my_meth(:keyword => 3) # Ruby 1.8
(2) my_meth(keyword:3, another_option:10) # Ruby 2.0

Won’t this create confusion? Why must keyword arguments use the same
syntax as new hash literals? How about “=” for keyword arguments instead
(such as in python)?


#2

On Mar 1, 2006, at 3:32 PM, Mr. Big wrote:

Won’t this create confusion? Why must keyword arguments use the same
syntax as new hash literals? How about “=” for keyword arguments
instead
(such as in python)?


Posted via http://www.ruby-forum.com/.

Ruby doesn’t have keyword arguments at all. When you call a method
with “keyword arguments” you are really just passing in a single
hash. There’s no confusion because the syntaxes do the exact same thing
example_hash = { a: 1, b: 2 }
my_meth(example_hash) #pass a hash as the single argument to my_meth
my_meth(a: 1, b: 2) # pass a hash as the single argument to my_meth

Neither of these will actually set local variables, the signature for
a method like this is as follows:

def my_meth(argument_hash)

end

e.g.:

def my_meth(arg_hash)
arg_hash[:a] + arg_hash[:b]
end

So one would say my_meth(:a => 1, :b => 2) and get back 3
or one would say my_meth(a: 1, b: 2) and get back 3
or one would say
example_hash = { :a => 1, :b => 2 }
my_meth(example_hash) # returns 3 also


#3

When you call a method with “keyword arguments” you are really just passing in a single hash.

Keyword arguments aren’t hashes.

def my_meth(foo:, bar:)
foo + bar
end
my_meth(foo:3, bar:4)

later we see this, we do not know how the method is defined.

my_meth2(arg:100, arg2:10)

There is no way to tell if my_meth2 is using a hash or keyword arguments
without looking it up. However, if keyword arguments used “=” instead,
there would be no double checking.
my_meth2(arg=100, arg2=10)

A keyword argument that takes a symbol ends up looking very ugly (I
guess “=” doesn’t work well here, either):
foo(key::symbol) or
foo(key: :symbol)


#4

Mr. Big wrote:

Ruby 2.0 will include new syntax for hash literals: {a:3, b:4}. However,
one can leave off the {}s to create a hash.

Won’t this create confusion?

I am personally not in favor of allowing the : character instead of =>.
(The last time I checked, this new syntax was marked as highly
experimental and not a done deal yet).

With colons already in use to denote the start of a symbol name, using a
colon as a separator between a key-value pair is going to confuse many
people, especially since everyone seems to like their own whitespace
conventions.

I hope Matz reconsiders and decides that the “experiment” is not worth
it.

As an aside, I wonder why => was chosen instead of a simple = sign.
Seems like it’s un-ruby-like to make us type the extra character :slight_smile:

Jeff
www.softiesonrails.com


#5

Quoting Jeff C. removed_email_address@domain.invalid:

As an aside, I wonder why => was chosen instead of a simple =
sign.
Seems like it’s un-ruby-like to make us type the extra character
:slight_smile:

Assignments are legal in most places that => is; would you want to
trade one for the other?

-mental


#6

Hi –

On Thu, 2 Mar 2006, Jeff C. wrote:

With colons already in use to denote the start of a symbol name, using a
colon as a separator between a key-value pair is going to confuse many
people, especially since everyone seems to like their own whitespace
conventions.

I hope Matz reconsiders and decides that the “experiment” is not worth
it.

As an aside, I wonder why => was chosen instead of a simple = sign.
Seems like it’s un-ruby-like to make us type the extra character :slight_smile:

I don’t think it’s extra. I would hate to have to parse – visually
– things like:

hash = { a = 1, b = 2, 4 = 5 }

etc.

David


David A. Black (removed_email_address@domain.invalid)
Ruby Power and Light (http://www.rubypowerandlight.com)

“Ruby for Rails” chapters now available
from Manning Early Access Program! http://www.manning.com/books/black


#7

unknown wrote:

Quoting Jeff C. removed_email_address@domain.invalid:

Assignments are legal in most places that => is; would you want to
trade one for the other?

-mental

Really? This works for me:

h = { ‘a’ => 5, ‘b’ => 6 }

but this gives my syntax errors:

h = { ‘a’ = 5, ‘b’ = 6 }

test.rb:1: odd number list for Hash
h = { ‘a’ = 5, ‘b’ = 6 }
^
test.rb:1: syntax error
h = { ‘a’ = 5, ‘b’ = 6 }
^

Maybe I’m doing something wrong here?


#8

Hi –

On Thu, 2 Mar 2006, Jeff C. wrote:

h = { ‘a’ => 5, ‘b’ => 6 }

but this gives my syntax errors:

h = { ‘a’ = 5, ‘b’ = 6 }

test.rb:1: odd number list for Hash
h = { ‘a’ = 5, ‘b’ = 6 }
^
test.rb:1: syntax error
h = { ‘a’ = 5, ‘b’ = 6 }

No one said illegal assignments were allowed :slight_smile:

But consider this:

h = { a = 5, b = 6 } # {5 => 6}

David


David A. Black (removed_email_address@domain.invalid)
Ruby Power and Light (http://www.rubypowerandlight.com)

“Ruby for Rails” chapters now available
from Manning Early Access Program! http://www.manning.com/books/black


#9

David Black wrote:

I don’t think it’s extra. I would hate to have to parse – visually
– things like:

hash = { a = 1, b = 2, 4 = 5 }

That’s true. I guess I was thinking of method calls, like:

start_hockey_game(:home_team = “Chicago”, :away_team = “Detroit”)

In other words, in reality most of the hashes I create are implicit
hashes created when passing method arguments.

But you’re right, for your canonical case, it would be bad; and that
alone probably justifies having a separate syntax for the key-value
pair.

Jeff


#10

David Black wrote:

No one said illegal assignments were allowed :slight_smile:

Phhhewwwwww - Man, has it been a long day. Sorry for the brainfreeze.
:slight_smile:

Jeff


#11

Hi,

In message “Re: Confusion Over Keyword Arguments”
on Thu, 2 Mar 2006 05:32:50 +0900, “Mr. Big” removed_email_address@domain.invalid writes:

|def my_meth(options={})
|end
|
|(1) my_meth(:keyword => 3) # Ruby 1.8
|(2) my_meth(keyword:3, another_option:10) # Ruby 2.0
|
|Won’t this create confusion?

No, on the virtual implementation (only inside of my brain), they are
same things. The magic is in receiving and interpreting arguments.

|Why must keyword arguments use the same
|syntax as new hash literals?

No, we don’t have to. The colon can be ugly with symbols, for
example,

db.find(order: :date)

is not good looking. Any other proposal (except for "=’)?

|How about “=” for keyword arguments instead
|(such as in python)?

Unfortunately, assignments are legal in argument list in Ruby.

						matz.

#12

Quoting Jeff C. removed_email_address@domain.invalid:

h = { ‘a’ => 5, ‘b’ => 6 }

but this gives my syntax errors:

h = { ‘a’ = 5, ‘b’ = 6 }

Legal, not equivalent.

For example, you can do this:

p { ‘a’ => k = 5, ‘b’ => 6 }
p k

which would print:

{“a”=>5, “b”=>6}
5

The arguments to => can be any legal Ruby expressions, including
assignments.

-mental


#13

Yukihiro M. wrote:

Hi,

In message “Re: Confusion Over Keyword Arguments”
on Thu, 2 Mar 2006 05:32:50 +0900, “Mr. Big” removed_email_address@domain.invalid writes:

|def my_meth(options={})
|end
|
|(1) my_meth(:keyword => 3) # Ruby 1.8
|(2) my_meth(keyword:3, another_option:10) # Ruby 2.0
|
|Won’t this create confusion?

No, on the virtual implementation (only inside of my brain), they are
same things. The magic is in receiving and interpreting arguments.

|Why must keyword arguments use the same
|syntax as new hash literals?

No, we don’t have to. The colon can be ugly with symbols, for
example,

db.find(order: :date)

is not good looking. Any other proposal (except for "=’)?

We are running out of characters, could we instead switch to
be the very first fully Unicode language? Have the actual
lambda sign for lambdas, the sigil for symbols and so on… :slight_smile:

|How about “=” for keyword arguments instead
|(such as in python)?

Unfortunately, assignments are legal in argument list in Ruby.

  					matz.

E


#14

On Thu, 2 Mar 2006 08:46:23 +0900, Yukihiro M. wrote:

No, we don’t have to. The colon can be ugly with symbols, for
example,

db.find(order: :date)

is not good looking. Any other proposal (except for "=’)?

If the keyword has to be unadorned, and this has to be done solely
through
an infix operator, how about:

db.find (order ~> :date)
db.find (order -> :date)

If it’s ok to adorn the keyword, and the parser can distinguish these,
that
opens up lots of options:

db.find ([order] = :date)
db.find (|order| = :date)
db.find (-order- = :date)
db.find (/order/ = :date)
db.find (#order# = :date)
db.find (#order = :date)

I’ve got no clue how the parser works, so these may be non-starters for
various reasons.

Jay L.


#15

Hi,

In message “Re: Confusion Over Keyword Arguments”
on Thu, 2 Mar 2006 09:32:41 +0900, “E. Saynatkari” removed_email_address@domain.invalid
writes:

|We are running out of characters, could we instead switch to
|be the very first fully Unicode language? Have the actual
|lambda sign for lambdas, the sigil for symbols and so on… :slight_smile:

Too late. Perl6 took that place.

						matz.

#16

I don’t think anything works like ‘:’ does --it has a divisor quality.
So as long as a symbol is denotated with a prefix ‘:’ there’s the
symptom of non-intuitiveness involved.

So short of changing the symbol prefix --I was just playing with a
space+period idea, in fact

db.find( order: .date )

How does allowing

db.find( order::date )

and deprecating ::'s previous meaning, work out?

T.


#17

On 1-Mar-06, at 7:40 PM, Yukihiro M. wrote:

Too late. Perl6 took that place.
Does Perl 6 exist yet? (Yes I know Pugs exists, but Perl 6 seems to
always be about two years away :wink:

Mike

Mike S. removed_email_address@domain.invalid
http://www.stok.ca/~mike/

The “`Stok’ disclaimers” apply.


#18

On 3/1/06, Jay L. removed_email_address@domain.invalid wrote:

an infix operator, how about:

db.find (order ~> :date)
db.find (order -> :date)

How about using Pascal for inspiration? Keep both the : folks and the
= crowd happy…

db.find (order := :date)

-Adam