[0.9.0] Added limited table support to Channel:push?

[Azhukar]

What does limited table support mean in regards to capabilities?

[slime]

Simple tables can be copied (not referenced) between threads.
Keys and values which are numbers, strings, booleans, or references to love objects will be copied. Functions and tables inside tables won't.

[Azhukar]

Simple tables can be copied (not referenced) between threads.
Keys and values which are numbers, strings, booleans, or references to love objects will be copied. Functions and tables inside tables won't.

Why not use recursion on tables inside tables and assume the user will treat it gently, i.e. not use table loops or other shenanigans?

Other than that it sounds awesome

[Kingdaro]

In the event that some asshole decides to do something like this:

Code: Select all

tab = {}
tab[tab] = tab

That'll be a nice loop you can sort yourself out of there.

EDIT: I'm joking, for the record.

[raidho36]

Why not just let those assholes deal with them being assholes to themselves? Regular people don't do this kind of crap and therefore won't encounter looped references in tables. I mean, who cares about those retards who do that. Besides, you can make some degree of foolproofing (though that goes against Lua's philosophy), but idiotproofing is meaningless: idiot can always find a way to break it.

Like I said, why bothering over getting invalid code to run if you don't expect it to run in the first place?

[T-Bone]

Why not just let those assholes deal with them being assholes to themselves? Regular people don't do this kind of crap and therefore won't encounter looped references in tables. I mean, who cares about those retards who do that. Besides, you can make some degree of foolproofing (though that goes against Lua's philosophy), but idiotproofing is meaningless: idiot can always find a way to break it.

Like I said, why bothering over getting invalid code to run if you don't expect it to run in the first place?

But having storing a table within itself is not invalid code in any other LÖVE context, so why should it be one here?

I'm actually sort of against any table support that isn't complete, mostly because the rules it places on those tables has to be a bit arbitrary, which can be confusing. It doesn't feel very "clean".

[raidho36]

But having storing a table within itself is not invalid code in any other LÖVE context, so why should it be one here?

Because that's how it works? Really though, tables should be referenced rather than copied.

It doesn't feel very "clean".

Then think of them as arrays of data rather than actually a tables.

[Robin]

It's trivial to write a deepcopy that deals with cycles in Lua. It might not be that simple to write an implementation that does the same trick across threads.

Who knows, maybe full table support to Channel:push for 0.9.1? I might implement it myself*, and finally contribute something substantial to el engine.

* Hah, as if!

[Xgoff]

Why not just let those assholes deal with them being assholes to themselves? Regular people don't do this kind of crap and therefore won't encounter looped references in tables. I mean, who cares about those retards who do that.

Code: Select all

_G._G

[raidho36]

Good one, Xgoff. Not that anyone would pass root table to the thread, which would mean copying entire thing minus locals.