Accessing Nested Hashes Directly

A

atraver

I've been having a problem for a while with Ruby where accessing a hash
within a hash will fail if the first hash is nil. As an example, Rails
accepts XML input to an action by passing it through REXML. On an XML
document like this:

<body>
<user>
<name>
<given>Tony</given>
<family>Robbins</family>
</name>
</user>
</body>

You will receive a hash that looks like this (within the params hash):

"body" => { "user" => { "name" => { "given" => "Tony", "family" =>
"Robbins" } } }

Thus, you can access the elements like this:

params["body"]["user"]["name"]["given"]

However, sometimes the "name" element might not come over the wire
(through no fault of Ruby's nor Rails' -- it's just a problem we have
to deal with in our system). In this case, Ruby will throw a
NoMethodError on nil.[] because we tried to access the value at "given"
on an element that does not exist. Furthermore, sometimes the "user"
element does not exist, either! This led me to code that looks like
this:

if params["body"]["user"]
if params["body"]["user"]["name"]
first_name = params["body"]["user"]["name"]["given"]
end
end

That's not very elegant when you're doing it with many (15 - 30)
fields. I also tried doing it with exceptions:

begin
first_name = params["body"]["user"]["name"]["given"]
rescue NoMethodError
end

Also, not very elegant.

Finally, I wrote a method to work around this problem:

def get_value(object, *path)
new_object = object

for item in path
item = item.to_s

if new_object
if new_object[item]
new_object = new_object[item]
else
new_object = nil
end
else
break
end
end

if new_object.kind_of?(String)
return new_object.strip
else
return new_object
end
end

This method can be called like this:

first_name = get_value(params, :body, :user, :name, :given)

It will traverse the hash and kick back a nil at the first problem it
finds instead of raising an exception, or will return the value if it
actually exists.

Here's my question, though: is this code efficient? Is there a better
way? Am I missing something fundamental in Ruby that would solve this
without the need for the new method?
 
R

Ross Bamford

I've been having a problem for a while with Ruby where accessing a hash
within a hash will fail if the first hash is nil.
[...snipped...]
Finally, I wrote a method to work around this problem:

def get_value(object, *path)
new_object = object

for item in path
item = item.to_s

if new_object
if new_object[item]
new_object = new_object[item]
else
new_object = nil
end
else
break
end
end

if new_object.kind_of?(String)
return new_object.strip
else
return new_object
end
end

This method can be called like this:

first_name = get_value(params, :body, :user, :name, :given)

It will traverse the hash and kick back a nil at the first problem it
finds instead of raising an exception, or will return the value if it
actually exists.

Here's my question, though: is this code efficient? Is there a better
way? Am I missing something fundamental in Ruby that would solve this
without the need for the new method?

I think you can rewrite that method like this:

def get_value(hash, *path)
path.inject(hash) { |obj, item| obj[item] || break }
end

If you're up for some core-modifying mayhem, You could put it on Hash,
and extend [] with 'path' capabilities:

class Hash
alias :__fetch :[]

def traverse(*path)
path.inject(self) { |obj, item| obj.__fetch(item) || break }
end

def [](*args)
(args.length > 1) ? traverse(*args) : __fetch(*args)
end
end

This way works like this:

h = {
:name => {
:first => 'Ross',
},
:contact => {
:phone => {
:eek:ffice => '345345'
}
}
}

p h.traverse:)name, :first)
# => "Ross"

p h.traverse:)name, :middle)
# => nil

p h[:contact]
# => {:phone=>{:eek:ffice=>"345345"}}

p h[:contact, :phone, :eek:ffice]
# => "345345"

p h[:contact, :phone, :fax]
# => nil

Of course, there may be better ways to solve the underlying problem...
 
V

Verno Miller

Posted by Adam Traver
on 26.09.2006 19:57
I've been having a problem for a while with Ruby where accessing a hash
within a hash will fail if the first hash is nil.

...

You will receive a hash that looks like this (within the params hash):

"body" => { "user" => { "name" => { "given" => "Tony", "family" => "Robbins" } } }

Thus, you can access the elements like this:

params["body"]["user"]["name"]["given"]

...

Here's my question, though: is this code efficient? Is there a better
way? Am I missing something fundamental in Ruby that would solve this
without the need for the new method?

Yet another (though indirect) approach could be to first extract all
keys from the nested hash!

http://bigbold.com/snippets/posts/show/1908
 
A

atraver

Thanks for all the responses.

Here is the method I ended up with after looking over everything, just
in case this topic is searched for in the future:

def get_value(object, *path)
for item in path
break unless object
object = object ? object[item.to_s] : nil
end

return object.kind_of?(String) ? object.strip : object
end

This might be a little more verbose than some of the solutions (also,
Ross, your "inject" solution didn't seem to work out although I'm sure
it would with some tweaking). However, it ends up doing exactly what I
need it to do: get the value or return nil, in a quick-and-dirty way.

-Adam
 
H

hadley wickham

I've been having a problem for a while with Ruby where accessing a hash
within a hash will fail if the first hash is nil. As an example, Rails
accepts XML input to an action by passing it through REXML. On an XML
document like this:

<body>
<user>
<name>
<given>Tony</given>
<family>Robbins</family>
</name>
</user>
</body>

You will receive a hash that looks like this (within the params hash):

"body" => { "user" => { "name" => { "given" => "Tony", "family" =>
"Robbins" } } }

Thus, you can access the elements like this:

params["body"]["user"]["name"]["given"]

However, sometimes the "name" element might not come over the wire
(through no fault of Ruby's nor Rails' -- it's just a problem we have
to deal with in our system). In this case, Ruby will throw a
NoMethodError on nil.[] because we tried to access the value at "given"
on an element that does not exist. Furthermore, sometimes the "user"
element does not exist, either! This led me to code that looks like
this:

if params["body"]["user"]
if params["body"]["user"]["name"]
first_name = params["body"]["user"]["name"]["given"]
end
end

That's not very elegant when you're doing it with many (15 - 30)
fields. I also tried doing it with exceptions:

begin
first_name = params["body"]["user"]["name"]["given"]
rescue NoMethodError
end

What about:
first_name = params["body"]["user"]["name"]["given"] rescue nil

Hadley
 
R

Ross Bamford

Thanks for all the responses.

Here is the method I ended up with after looking over everything, just
in case this topic is searched for in the future:

def get_value(object, *path)
for item in path
break unless object
object = object ? object[item.to_s] : nil
end

return object.kind_of?(String) ? object.strip : object
end

This might be a little more verbose than some of the solutions (also,
Ross, your "inject" solution didn't seem to work out although I'm sure
it would with some tweaking). However, it ends up doing exactly what I
need it to do: get the value or return nil, in a quick-and-dirty way.

Hmm, out of interest, what didn't work out about it? Maybe I
misunderstand what you're doing, but (if you ignore the Hash-extending
silliness) it seems to work the same.

require 'test/unit'

def get_value(object, *path)
for item in path
break unless object
object = object ? object[item.to_s] : nil
end

return object.kind_of?(String) ? object.strip : object
end

def my_get_value(hsh, *path)
path.inject(hsh) { |obj, item| obj[item.to_s] || break }
end

class TestGetValue < Test::Unit::TestCase
def setup
@h = {
'name' => {
'first' => 'Ross'
},
'contact' => {
'phone' => {
'office' => '345345'
}
}
}
end

def test_it
assert_equal get_value(@h, :name, :first),
my_get_value(@h, :name, :first)
assert_equal get_value(@h, :name, :middle),
my_get_value(@h, :name, :middle)
assert_equal get_value(@h, :contact),
my_get_value(@h, :contact)
assert_equal get_value(@h, :contact, :phone, :eek:ffice),
my_get_value(@h, :contact, :phone, :eek:ffice)
end
end

# -> 1 tests, 4 assertions, 0 failures, 0 errors


The only thing I changed here is adding to_s on the item in my version,
so I could use the same hash to test both while keeping symbol keys.
 
J

John Turner

This method can be called like this:

first_name = get_value(params, :body, :user, :name, :given)

It will traverse the hash and kick back a nil at the first problem it
finds instead of raising an exception, or will return the value if it
actually exists.

Here's my question, though: is this code efficient? Is there a better
way? Am I missing something fundamental in Ruby that would solve this
without the need for the new method?

I didn't really like the way that method is called, doesn't seem very
"ruby" exactly. Came up with the following which is really dirty at the
moment but I prefer how it's called. The method_missing in NilClass is
kind of unforgivably dirty. Would like to see improvements I guess, but
not sure there's really a way to do this nicely:

So, yeah, don't use this, but possibly be amused by it anyway:

require 'ostruct'
class NestStruct < OpenStruct
def self.fromNestedHash(h)
struct = NestStruct.new(h)
struct.marshal_dump.keys.each do |key|
if struct.send(key).is_a? Hash
struct.send(key.to_s + '=', fromNestedHash(struct.send(key)))
end
end
return struct
end

def method_missing(m, *args)
super(m, args)
rescue NoMethodError
nil
end
end

class NilClass
def method_missing(m, *args)
nil
end
end

h = {
'name' => {
'first' => 'Ross'
},
'contact' => {
'phone' => {
'office' => '345345'
}
}
}

s = NestStruct.fromNestedHash(h)
puts s.contact.phone # => #<NestStruct office="345345">
puts s.contact.phone.office # => 345345
puts s.contact.phone.home # => nil
puts s.foo.bar # => nil
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

No members online now.

Forum statistics

Threads
473,768
Messages
2,569,575
Members
45,053
Latest member
billing-software

Latest Threads

Top