Handling null and undefined

[richardboehme]

Currently it's not possible to determine if a JS::Object wrapping null or undefined is null/undefined. The method for checking other types (JS#is_a?) does not work on those types.

null_value = JS::eval('return null')
p JS.is_a?(null_value, JS.global[:null])
# returns false because `JS.global[:null]` is undefined so this cannot work

undefined_value = JS::eval('return undefined')
p JS.is_a?(undefined_value, JS.global[:undefined])
# returns false because `typeof undefined` is 'undefined' and not 'function'

I think there are two options for handling this situation:

1. Convert JS null to Ruby nil and fix JS#is_a? for undefined.
2. Fix JS#is_a? for null and undefined and keep both wrapped in a JS::Object.

I personally think the advantage of 1. is that you can write code like:

if js_value
  # do something
else
  # do something else
end

However undefined is also used by a lot of JS code and you would still have to write code like this:

if JS.is_a?(js_object, JS.global[:undefined])
  # do something
else
  # do something else
end

The advantage of the second approach is that both cases are handled in the same way. However JS.global[:null] is not defined so we'd have to define what value to pass into JS#is_a?.

What are your opinions on this?

[kateinoigakukun]

Option 1. Convert JS null to Ruby nil and fix JS#is_a? for undefined.
Option 2. Fix JS#is_a? for null and undefined and keep both wrapped in a JS::Object.

JS.is_a? method is not designed to test whether a value is equal to another value, and it's just an alias of instanceof operator in JS. So I don't want to change its behavior.

As a third option, I've added == / eql? / strictly_eql? methods to JS::Object to be able to test a JS object is null or undefined. #28

It would allow you to write null/undefined check by:

JS_NULL = JS.eval("return null")
JS_UNDEFINED = JS.eval("return undefined")

if js_value == JS_NULL
  puts "null"
elsif js_value == JS_UNDEFINED
  puts "undefined"
end

Does it make sense to you?

[kateinoigakukun]

Thank you for your try! Hmm, it's weird. Could you give me a stack trace where the OOB access happens?

[richardboehme]

This is the stack trace I got... The code I used is puts JS::NULL. I'm new to Ruby C-Extension stuff so maybe I messed something up there? I tried just setting a string as a constant and this worked though.

[kateinoigakukun]

Sorry for my slow reply. I've added null and undefined constants by #30
Could you try it?

[richardboehme]

Hey! You do not have to apologize for slow replies! This is open source and I'm so grateful that you even put your time into this ❤️

I tried your patch and sadly it does not seem to work for me. Comparing like in your tests works, but when trying to print the constant (p JS::NULL) it throws the following error:

Uncaught Error: eval:103:in `p': method `inspect' called on hidden T_DATA object (0x02369eb0 flags=0xc) (NotImplementedError)
eval:103:in `<main>'
    at checkStatusTag (browser.umd.js:1449:21)
    at evalRbCode (browser.umd.js:1462:7)
    at RubyVM.eval (browser.umd.js:1258:18)
    at HTMLDocument.<anonymous> (application.rb?t=1657540653298&import:1:69)
    at main (main.js?t=1657540653298:23:14)

I'm a bit confused by the error message.... I tried to lookup what it could be and stumbled upon a ruby/openssl issue (see here). I'm not sure if it's really related but they fixed it by marking the VALUE. I tried adding a rb_gc_mark to _rb_js_eval_js_cstr but this just leads to another error:

0a2dd336:0x2836b5 Uncaught RuntimeError: table index is out of bounds
    at gc_mark_ptr (0a2dd336:0x2836b5)
    at rb_gc_mark (0a2dd336:0x28464d)
    at _rb_js_eval_js_cstr (0a2dd336:0x64a6)
    at Init_js (0a2dd336:0x5a88)
    at require_internal (0a2dd336:0x3879a6)
    at rb_f_require (0a2dd336:0x38739f)
    at ractor_safe_call_cfunc_1 (0a2dd336:0x5f4569)
    at vm_call_cfunc_with_frame (0a2dd336:0x5e640d)
    at vm_call_method_each_type (0a2dd336:0x5e4724)
    at vm_call_method (0a2dd336:0x5ae24b)

At this point I'm out of ideas. Do you have an idea how we can fix this issue?

[kateinoigakukun]

Ah, I see. puts was the error cause. Let me investigate further later.