Project

General

Profile

Feature #15815

Add option to raise NoMethodError for OpenStruct

Added by mtsmfm (Fumiaki Matsushima) 3 months ago. Updated 3 months ago.

Status:
Open
Priority:
Normal
Assignee:
-
Target version:
-
[ruby-core:92513]

Description

GitHub PR: https://github.com/ruby/ruby/pull/2164

Currently, OpenStruct#method_missing returns nil even if the key isn't registered.

require 'ostruct'
os = OpenStruct.new({a: 1})
os.a #=> 1
os.b #=> nil

I'd like to add exception option to raise NoMethodError in such case.

require 'ostruct'
os = OpenStruct.new({a: 1}, exception: true)
os.a #=> 1
os.b #=> NoMethodError

Use case

I sometimes use OpenStruct as a JSON API response wrapper.
It's useful to use method call instead of key access (obj[:key]) because we can use Symbol#to_proc if it's a method (for example users.map(&:id))

But I want to prevent typo for a key name. Currently users.map(&:idd) just returns [nil,...]

Even if we have this exception option, we can't enable this option for JSON parser easily though:

JSON.parse(response, object_class: Class.new(OpenStruct) { def initialize(hash); super(hash, exception: true); end })

What do you think?


I've searched with "openstruct nomethoderror" on bugs.ruby-lang.org though, please let me know if it's duplicated.
https://bugs.ruby-lang.org/search?utf8=%E2%9C%93&scope=&q=nomethoderror+openstruct

History

Updated by marcandre (Marc-Andre Lafortune) 3 months ago

It's hard to decide on features of OpenStruct as it's somewhat an antipattern.

Note that you might be able to write users.map{@[:id]} in the next version and avoid using an OpenStruct altogether.

That being said, I am wondering if a frozen OpenStruct could possibly raise an error upon reading an undefined key.

Updated by shevegen (Robert A. Heiler) 3 months ago

Note that you might be able to write users.map{@[:id]} in the next version and
avoid using an OpenStruct altogether.

I think that the idiom, or "mental mode", e. g. how users use ruby, is a bit different
between the two examples. At the least to me, "OpenStruct" is instantly recognizable
and "exception: true" is also no longer a rare idiom, after some core methods accepted
it; the map {@[:id] variant, at the least to me, is slightly harder to understand for
my brain. But anyway, that's just my opinion - to the suggestion itself, one described
use case:

But I want to prevent typo for a key name.

Although I personally do not use struct and openstruct a lot, even though I think it is
a cool idea (prototypical objects all the way, at all times), to me the suggestion appears
useful. So I am in light support of the suggestion; I don't feel particularly strong
either way, though. (May help for people to comment who actually use struct and openstruct
a lot.)

Also available in: Atom PDF