Make GDScript allow some keywords as identifiers

[bojidar-bg]

Fixes #8085

Here is a test script:

extends SceneTree

func log(a): # Call with self.log()
	print(a)

func _initialize():
	var dict = {
		"true": 1,
		"false": 1,
		"null": 1,
		"sync": 1,
		"remote": 1,
		"slave": 1,
		"if": 1,
		"match": 1,
	}

	print(dict.true)
	print(dict.false)
	print(dict.null)
	print(dict.sync)
	print(dict.remote)
	print(dict.slave)
	print(dict.if)
	print(dict.match)
	print(log(1))
	print(self.log(1))
	quit()

Here are some things that won't work with this changes:

var var = 1
var const = 1
var func = 1

func and(a, b): # Technical issues for this one, since it is equivalent to func &&(a, b):..

And here are some strange ones that would work

func var(): pass # Call with self.var()
func func(): pass # self.func() or just func() if lambda isn't merged
class class:
  static func static(): return true
func _ready():
  self.var()
  self.func()
  print(class.static()) # true

[karroffel]

Looks good so far, I'd keep the old error messages though.

[karroffel]

literal? It's still an identifier. It's just that the identifier might have the same name as a keyword, it's an identifier nonetheless.

Usually "constant values" (like a string or a number) are called literals. I'd keep the old error message.

[bojidar-bg]

Ookay, would revert that part.

[karroffel]

Old relict? 😄 This is more descriptive, cool. But I'd still change it to back to identifier.

[bojidar-bg]

@karroffel Fixed

[karroffel]

👍 Looks good!

[akien-mga]

Since we now prevent naming variables like properties of the base class (e.g. if you name a member variable position or rotation in a Node2D, the debugger will raise an error), I think it's unnecessary to allow some half-shadowing of keywords and global scope methods. Because to be consistent we'd have to allow the same for member variables, and it would just become very confusing.

I think Godot users will learn to live with this limitation, it might be frustrating a couple times, but then you make do and benefit from the consistency.

[bojidar-bg]

@akien-mga Still, I think we should allow calling methods like sync or log, even if we disallow creating variables or methods with such names.

[akien-mga]

Well if we disallow creating them, it makes no sense to allow calling them, since the situation would never arise.

[bojidar-bg]

@akien-mga That's wrong, since we can just have a GDNative or C# script which would be free to register them...

[akien-mga]

So what should we do here (apart from a needed rebase)?

[vnen]

I'm OK with this change as it fixes some current problems, like String.match() that can't be called. Creating functions with keywords as names should be documented to explain the need of the self. prefixing.

[akien-mga]

@bojidar-bg You've got the go from @reduz, albeit you need to add some more documentation (and rebase):

<reduz> it's ok, but i would add a bit more doc on the source code to make sure what everything is doing and why, since it's clearly an usability hack
<reduz> as in, anyone who looks at the source eould expect to find TK_IDENTIFIER in the switch, for example
<reduz> so explaining those hacks would be a good idea

[bojidar-bg]

Ok, rebased, now I would have to think how to include the doc. 😃 Edit: done

[reduz]

we really should update all docs..

…

On Sun, Jul 23, 2017 at 5:36 PM, Bojidar Marinov ***@***.***> wrote: (Bump, done, updating for email notifications) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8217 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF-Z26NxyYnWoqKhWrR5YRMG8wlUkBsyks5sQ66ygaJpZM4Mv8uM> .

[Geequlim]

After this commit the GDScript is totally broken!

[akien-mga]

Let's give @bojidar-bg some time to try a fix :)

[akien-mga]

@Geequlim Fixed by #9846.