Complement

Ginger

13:20

update: it looks like the code required for function blocks to work won't be released for a few months

so I'm going to remove them from the toolbox until then

and for that matter:

Feeds

Ginger has made a change to the feeds posted into this room

Ginger

now we'll get notified when it happens

Ginger

14:25

→ 5 messages moved to Trash

thanks feeds

Ginger

15:53

I wish it were possible to specify more verbose messages in dropdown menus for blocks

ideally I'd have the symbol by itself in the block and the symbol with a description in the menu

guess I get to roll my own!

Ginger

16:27

problem: even though I have overridden a method of a JS subclass, the parent's un-overridden constructor is calling its method instead of the subclass's one

caird coinheringaahing

Just override the parent's constructor then :P

Radvylf Programs

Parent.prototype.method = function() { if (this instanceof Child) { ... } }

^^

Ginger

that'd involve duplicating a lot of code, which I really do not want to do

Radvylf Programs

Duplicating?

Just store the old one in a variable

Ginger

I'd have to copy the constructor and the fifty-line utility method it calls

Radvylf Programs

16:29

.apply(this, ...arguments)

@Ginger Why

Ginger

because I can't call the method myself due to it being set to private

pxeger

You shouldn't override it then

Ginger

what?

the constructor calls a utility method which isn't exported in the built version of the library

Radvylf Programs

@Ginger source?

pxeger

@Ginger oh is the utility method a third method, not the one you're trying to override?

Ginger

16:31

@RadvylfPrograms blockly, I'm overriding the trimOptions_ method of the FieldDropdown class

@pxeger yes

@RadvylfPrograms see this line

Radvylf Programs

Weird

pxeger

2 mins ago, by Radvylf Programs

.apply(this, ...arguments)

doesn't that^ work on the constructor?

Ginger

eh?

I am not familiar with this "feature"

Radvylf Programs

this.trimOptions_() should refer to the one in your class, not the parent

Not sure how extends works in detail though

@pxeger What do you mean

Like, create a new constructor that .applys the super constructor?

pxeger

yes

Ginger

16:33

I have confirmed that if I copy the constructor myself (commenting out the utility method) it calls my overriden one

Radvylf Programs

I'd assume that's done my default

But maybe not

pxeger

(not knowing much about JS's OOP features) I was inferring that the superclass's constructor wasn't doing that

even though it feels like it should

shall we go to another room so caird and catja can talk about actual code golf?

Ginger

@pxeger JS doesn't have OOP features, it just has bugs that resemble OOP :b

@pxeger sure, we can go to the Complement room

pxeger

← 36 messages moved from The Nineteenth Byte

Ginger

that works

pxeger

16:35

← 1 message moved from The Nineteenth Byte

Ginger

so, what is this apply thing?

Radvylf Programs

It calls a function but overrides its this

Ginger

so I'd do super.apply(this, ...)?

Radvylf Programs

No, super.apply(this, ...)

Ginger

get ninja'd on

Radvylf Programs

16:37

You'd supply ... as an array of arguments, or you could use .call which takes them splatted

Ginger

Nope:

Uncaught ReferenceError: Must call super constructor in derived class before accessing 'this' or returning from derived constructor

Radvylf Programs

I figured that'd happen

Ginger

JS is the Ohio of programming languages

Radvylf Programs

I still don't understand what makes your situation different

Ginger

maybe that the class I'm overriding is a TypeScript one?

that could change something

Radvylf Programs

16:39

TypeScript compiles down to JS, doesn't matter

Show me your override

Ginger

class VerboseDropdown extends Blockly.FieldDropdown {
    // Have to redefine this so it uses our custom applyTrim_
    trimOptions_() {

Radvylf Programs

How about the constructor

Ginger

I'm overriding this function so it'll call another overridden static function which is referenced by name in the base one (as in it calls FieldDropdown.applyTrim_(...)

@RadvylfPrograms don't have one

Radvylf Programs

There's your issue

You're using the parent's constructor

Add a constructor that just calls super()

Ginger

._.

pxeger

16:41

that's so stupid

Why does it work like that

Ginger

but it didn't

@RadvylfPrograms it's still not working

Radvylf Programs

Wait yeah:

Type VerboseDropdown into the console and hit enter

Ginger

class VerboseDropdown extends Blockly.FieldDropdown {
    constructor(gen, f) {
        super(gen, f)
    }
    // Have to redefine this so it uses our custom applyTrim_
    trimOptions_() {
        consol…

that's the result

Radvylf Programs

Oops do its .prototype

Ginger

FieldDropdown$$module$build$src$core$field_dropdown {constructor: ƒ, trimOptions_: ƒ, dropdownCreate_: ƒ}
constructor
:
class VerboseDropdown
dropdownCreate_
:
ƒ dropdownCreate_()
trimOptions_
:
ƒ trimOptions_()
[[Prototype]]
:
Field$$module$build$src$core$field

weird

Radvylf Programs

16:45

Click trimOptions_() and make sure it's the new one

Ginger

brb

Radvylf Programs

Y'know what I bet it is

TypeScript probably transpiled this.trimOptions_(); to some weird thing

Do FieldDropdown.prototype.constructor and see what it transpiled down to

WAIT

I bet it shortened the name of trimOptions_

If TypeScript was told it was a private method, and the transpilation output is minified (which I'd bet it is), no way it's not called x1z or something now

Ginger

back (temporarily)

@RadvylfPrograms edge's being a jerk and compressing the result but it doesn't look minified at all

I can see all the method names

and g2g again

Ginger

18:40

@RadvylfPrograms any ideas for fixing the bug?

Radvylf Programs

Look at FieldDropdown.prototype

Or try to make an MCVE

Ginger

Field$$module$build$src$core$field {constructor: ƒ, fromXml: ƒ, loadState: ƒ, initView: ƒ, shouldAddBorderRect_: ƒ, …}
applyColour
:
ƒ applyColour()
constructor
:
class extends
createSVGArrow_
:
ƒ createSVGArrow_()
createTextArrow_
:
ƒ createTextArrow_()
doClassValidation_
:
doClassValidation_(a){return this.getOptions(!0).some(b=> {…}
doValueUpdate_
:
ƒ doValueUpdate_(a)
dropdownCreate_
:
ƒ dropdownCreate_()
dropdownDispose_
:
ƒ dropdownDispose_()
fromXml
:
ƒ fromXml(a)
getOptions
:
ƒ getOptions(a)

Radvylf Programs

I don't see trimOptions_ there

Does it inherit it from what it extends?

Ginger

uhhhhhhh

that's a good point

where the hell is it

maybe it gets shoehorned into the constructor during minimization?

Radvylf Programs

That's why I'm saying, look at the actual constructor

Ginger

18:44

that'd answer the question, at least: it's not being overrided because it's not actually there

Radvylf Programs

Who knows what the transpilation/minification is doing

Ginger

19:03

@RadvylfPrograms got it working!

Radvylf Programs

Nice!

Ginger

...by copying the entire constructor, but at least it works

now to go add details and tooltips to every block

Ginger

19:22

@RadvylfPrograms so just so I'm sure, the syntax for set-builder is {function | variables}?

also for that matter will we support what I'm going to call "set-builder ellipsis inference notation"?

where a set like {0, 1, 2, ...} is equal to the set of all numbers

Radvylf Programs

Hmm....maybe a good ides

a

Ginger

yes, but more work requiref

d

Radvylf Programs

Not really, basically just a range operator

Maybe we could have it detect geometric series too

Ginger

hm

Radvylf Programs

E.g., {1, 3, ..., 27}

Ginger

19:34

I'm guessing {1, ..., ...} is bad syntax?

Radvylf Programs

Yeah

Ginger

oh boy, I get to write a validator! ;-;

Radvylf Programs

... would only be allowed when it's bordered on both sides by either a number of the beginning/end of the set literal

In fact, for simplicity, only the following would be allowed:

{...}, {n, ...}, {..., n}, {n, ..., n}, {n, n, ...}, {n, n, ..., n}, {n, ..., n, n}, {..., n, n}

Ginger

what's {...}?

Radvylf Programs

The latter 4 being for geometric series

@Ginger All integers

Maybe that one would be disallowed

Ginger

19:39

that should be disallowed

it's ambiguous

Radvylf Programs

In what way

Ginger

it could be the set of all sets

Radvylf Programs

Oh true

Ginger

or a set of mixed ints and sets

or anything else really

also, since we aren't going to support non-integer numbers I guess I need to remove any instance where that could be possible

Radvylf Programs

Such as?

Ginger

19:41

the set builtins for complex numbers, etc

Radvylf Programs

Natural numbers are fine

Ginger

also I'll say it here: I think not supporting floats is a mistake and limits what this language will be able to do, especially in actual mathematical contexts

Radvylf Programs

Those are just positive integers

Ginger

but anyway

@RadvylfPrograms typo :b

Radvylf Programs

@Ginger If you think you can meaningfully implement constraints on floats you go right ahead :p

Ginger

19:42

what'd the issue be?

...oh

oh, right

yeah I see your point :p

but it still bugs me

maybe we could limit precision?

Radvylf Programs

Maybe to single-precision floats, but that's still a massive pain

And it'd make brute forcing many many times slower

(About 4 billion to be precise)

Which is a lotta manys

Ginger

how come Desmos can handle (limited) constraints easily?

is it doing something different?

Radvylf Programs

I would guess symbolic math

Which we'll have to a limited extent, but Desmos never has to worry about brute forcing and we do

Ginger

why?

Radvylf Programs

Because it doesn't do that

Ginger

19:49

wait, what exactly is the interpreter going to do? I thought that, like, the "print all numbers" program would just output <Set [1, 2, 3, ...]> or something

Radvylf Programs

Dec 6 at 23:40, by Radvylf Programs

1. We convert the AST into a set of contraints on variables
2. We repeatedly perform optimizations, merging constraints where possible and searching for contradictions
3. We brute force the unsolved variables

Desmos likely works using a variety of tricks like newton-raphson to approximate answers

Ginger

mfw I don't even know what the language I invented is supposed to do

@RadvylfPrograms ok so how can I figure out if a given position for an ellipsis is valid?

Radvylf Programs

You check if the set literal matches one of the seven allowed patterns

Ginger

hmm, ok

Radvylf Programs

How you do that is up to you, I don't know what tools you have available to implement that behavior

Ginger

19:55

~~me neither~~

Ginger

20:12

I'm not really sure how to implement this :p

like, what happens if you try to make an invalid setup? does the ellipsis block get ejected?

(@RadvylfPrograms)

what'd be the most user-friendly way of doing it?

Radvylf Programs

Don't allow invalid state

Have a separate ellipsis block that lets you pick each of the two bounds

With -inf and inf as options

Ginger

oh, I shoehorned it into the existing set-builder block

Radvylf Programs

Yeah I wouldn't do that

Maybe even have a whole set literals section

Ginger

hmmmmmmmmmmmmmmmm

I'll do that later™

Transcript for