When you use the block syntax with Hash.new to define a default value, you lock yourself to only using the square bracket syntax for value lookup. The fetch method ignores the block, default, and default_proc.
This bit me hard this week, and my pair, @gd, got to watch me flounder on why fetch wasn’t getting the default value.
By the way, the Hash.new with a block is great for memoization of calculated or network lookup data.
If you wish to memoize (cache / calculate only once) a value, a Hash with a block in ruby is a good way to do it.
For example:
h = Hash.new do |hash, key|
hash[key] = "Default value for #{key} "
end
h.fetch(:no_such, "not defined")
#=> "not defined"
h[:no_such]
# => "Default value for no_such"
This is means the value for h[:no_such] gets set the first time it is accessed and no longer needs to be calculated with the block, but fetch does not use that default proc.
Note, you don't need to store the default value in the proc, only if you wish to no longer calculate it on next access.
If you don't store it, it will be called every time until the key is set on the hash.
@jnanney it bothers me that the arity for the default_proc and the fetch block is different.
After reading this I wanted to see if the default_proc could be used:
h = Hash.new do |hash, key|
hash[key] = "Default value for #{key}"
end
puts h.fetch(:nope, &h.default_proc.curry(2)[h])
It works... but I would only use it if it was buried in some object that wraps the hash or something.
Jim Nanney
Jim Gay
Jochen Lillich
Evan Light