Pytek

@AquaTart Design choices have no bearing on speed. It's implementation choices that matter. As long as Pytek is designed well, the implementation can be optimized behind the scenes without changing the design.

Nathan Merrill

even if the array is small

Aqua Tart

2:32 AM

@Mego That's what I meant but I've gone semi-insane trying to get PyPy working so my brain isn't operating at 100%

Anyways, thoughts:

Cython

The Cython programming language is a superset of Python with a foreign function interface for invoking C/C++ routines and the ability to declare the static type of subroutine parameters and results, local variables, and class attributes. == Overview == Cython is a compiled language that generates CPython extension modules. These extension modules can then be loaded and used by regular Python code using the import statement. Cython is written in Python and works on Windows, Linux, and Mac OS X, producing source files compatible with CPython 2.4 through 3.4. It works by producing a standard Python...

As I understand it this could be used to compile the parser to a Python C module

We'd still need a Python wrapper though.

El'endia Starman

I'll let you guys worry about that. :P

Anonymous

Cython is neat but I don't know how much it will improve things

Anonymous

Regardless; optimization comes after full, correct implementation

El'endia Starman

*full-enough :P

Anonymous

Nope

Anonymous

2:34 AM

Optimization is the last step

El'endia Starman

Would you say that Python 3 has been fully implemented?

Anonymous

If you're doing something else after optimizing, 90% of the time you spent on optimization is wasted

Nathan Merrill

while optimization comes last, design decisions must take possible implementations into account

Aqua Tart

@Mego That's true of manual optimization, not "picking an implementation of Python that will optimize our code for us"

:P

Anonymous

@AquaTart We can deal with that after we get a working interpreter :P

Aqua Tart

2:36 AM

OK, OK.

By the way, \print(\factorial(\factorial(3))) totally works and ouputs 720. That's nice.

Anonymous

For example, I still haven't bothered with optimizations in Seriously 1 (aside from the minor ones I've done so I can do rigorous tests that will complete before the heat death of the universe, like basic memoization)

Aqua Tart

I'm going to clean up this code a bit then I think we should do variables next

El'endia Starman

I was thinking operators, actually.

Those are harder, in my mind.

Anonymous

I'm looking at the c9 ide now :P

Anonymous

If you want to do operators, you need a full precedence table

Aqua Tart

2:38 AM

yep

El'endia Starman

yup

TBD :P

Aqua Tart

That's why I was thinking variables first :P

Python handles scope with a list of hashes, so I figured we'd do something similar

Anonymous

Yeah variables first makes more sense to me

El'endia Starman

...do you guys mind if I put my foot down and say we are not designing/implementing this by committee? :P

Aqua Tart

Go ahead

Anonymous

2:40 AM

Sure, you can be BDFL. But we're going to harass you and make fun of you if you override our obviously smart ideas :P

El'endia Starman

lol

You can be assured that I will take every idea into consideration.

Anonymous

I may have missed it somewhere - do you have a design doc anywhere? The transcript doesn't count.

El'endia Starman

I posted a syntax spec earlier.

Bolded message.

Anonymous

Anonymous

y u no pin

El'endia Starman

2:41 AM

37 mins ago, by El'endia Starman

Draft of Pytek syntax spec: http://pastebin.com/pqwRUN5K

Well, 'cause it's only a draft. :P

I typed that up while at youth group.

Anonymous

Still, it's better than the transcript

Nathan Merrill

@El'endiaStarman I wasn't trying to act as a committee, sorry

El'endia Starman

@NathanMerrill I wasn't referring to you at all there, actually.

Nathan Merrill

oh, I'm just very opinionated

El'endia Starman

I was talking about both quartata and Mego saying "variables next!" and me saying "operators next!". :P

Nathan Merrill

2:43 AM

and I like pushing them

Aqua Tart

@El'endiaStarman What's your plan for parsing the operators?

Anonymous

If you want to work on operators, then we need a full precedence table

El'endia Starman

@NathanMerrill To be honest, I don't mind that very much. If I/we make a poor design choice, it's good to have someone call us out on it.

Anonymous

And while you're working on said table, we can sneakily work on variables :P

El'endia Starman

@Mego I'm going to implement + and * first, and that's it.

Anonymous

2:44 AM

Easy enough

Anonymous

Shunting yard algo

El'endia Starman

Oh gosh no, let's not work on the interpreter at the same time.

Yeah, basically.

I just have to integrate it into the current setup.

Anonymous

Modularity means that simultaneous work can be done :P

Aqua Tart

@Mego Please don't.

El'endia Starman

Ostensibly, but I will be distracted by trying to keep an eye on your work.

Aqua Tart

2:45 AM

If we fuck up the parser @El'endiaStarman is the only one that can fix it

Anonymous

I was joking :P

Nathan Merrill

this is why we need git

Aqua Tart

So we don't want to get on the bad side of Lord of the Parsers

Anonymous

^^ is there not a github repo?

Aqua Tart

There will be when we have something good.

El'endia Starman

2:46 AM

There will be one.

Anonymous

We should make one and put the spec in it for now

El'endia Starman

@AquaTart I won't wait until "something good". I'll have something on GitHub by next week for sure.

Anonymous

GitHub's markdown editor beats the pants off of pastebin :P

El'endia Starman

It just so happens that the current C9 setup is really convenient for collaborative work.

Nathan Merrill

@El'endiaStarman Right, but if he's working locally, then there's no "keeping an eye" :)

Anonymous

2:47 AM

Besides, we could be able to understand the parser

Anonymous

It might take some effort and alcohol, but it's certainly not impossible to understand

Sp3000

@El'endiaStarman Have you taken a look at Python's grammar? (It's up somewhere). Might be good for operator precedence

El'endia Starman

Ehhhh....still iffy. I have a fairly good idea of how to modify the parser for infix functions, but I might have to change other stuff to make it all work.

@Sp3000 Yeah, I'll take a look at their stuff when it comes time to implement all the operators.

Sp3000

Well, I mean, even for + and * ...

shrugs

El'endia Starman

I'm fairly sure 1 + 2 * 3 is 7 and not 9. :P

Sp3000

2:50 AM

Are you sure you don't want a parser generator for this? :P

El'endia Starman

Do you think a parser generator would properly handle [1,2,3] \pairwise(+) [4,5,6]?

(And besides, I want the experience.)

Aqua Tart

@El'endiaStarman Are you sure you can't just isPossible [\+\*]. isComplete true?

It seems like the way to go.

oh boy I can't wait until we get to the part where we parse chains

El'endia Starman

Chains are essentially operators, in my mind.

Anonymous

Man looking at the parser is making me glad I chose to make Seriously a stack-based imperative language

El'endia Starman

@AquaTart That won't work because the parser will treat those like a child node, and grab the next item as its child, but have no way to grab the previous element to make it a child of the operator node.

Sp3000

2:53 AM

Parser generators can handle whatever - but I think your example would be done differently to what you have in mind

El'endia Starman

In the meantime, @Mego and @AquaTart, we will need unit tests.

And you guys can add functions to pytek_functions.

Aqua Tart

@El'endiaStarman I have some test cases already.

I'll add a couple more.

Nathan Merrill

hey, what's a common use case for a queue

and are we going to support them?

El'endia Starman

@NathanMerrill Plastic utensils in restaurants.

@NathanMerrill Probably.

Nathan Merrill

are they going to get their own syntax, or simply be a class?

El'endia Starman

3:00 AM

@AquaTart The test#.pytek files, right? Any way we could set up a single file that has all of them, with inputs and expected outputs, and have a script that goes through and checks each input and output with the expected output, and notifies us when a test fails?

Probably just a type. queue:.

Nathan Merrill

ok. I'd recommend making your queue a dequeue (you can push/pop on both ends efficiently)

El'endia Starman

I probably would've done that, but I'll be sure to definitely do that.

Aqua Tart

@El'endiaStarman Sure

El'endia Starman

I really need to write a function that prints the AST tree.

Nathan Merrill

oooh, what if collections could define an "ordering"

and if you did, it would automagically keep the elements in order behind the scenes

(basically, turn it into priority queue)

El'endia Starman

3:10 AM

That could be useful. ordered:x = [5,3,6,1] would result in x being [1,3,5,6].

Nathan Merrill

it wouldn't just be a function either

if you appended to the array, it would maintain its order

El'endia Starman

yeah

Nathan Merrill

ensure that you allow for custom orderings in your design though

El'endia Starman

I think I'd use the ordering method of the type of the elements in the list.

If one is defined, that is.

Nathan Merrill

right, but I've often wanted to order a list of objects by something other than its natural ordering

El'endia Starman

3:13 AM

Oh right, a key.

hmm, yeah

I'll think on that.

Aqua Tart

@El'endiaStarman Hmm, I don't really want to invoke another subprocess for the unit tester.

Would there be a way to add a function that terminates execution and makes .run() return the argument given to it?

El'endia Starman

Not sure what you mean?

Aqua Tart

OK.

So you know how once you have the AST you execute it with .run()

El'endia Starman

uh-huh

Aqua Tart

Can we add a Pytek function that makes .run() immediately return the argument given to the function

El'endia Starman

3:22 AM

Like an identity function?

Aqua Tart

Right but currently .run() doesn't return anything

Nathan Merrill

oh, another thing: I'd strongly recommend static typing, but implied static typing

El'endia Starman

Oh right, hmm.

Aqua Tart

I want to be able to do \test(<code>) for the unit test codes and just be able to do assert(x.run() == foo)

El'endia Starman

@NathanMerrill Typing will be Python-esque if no type identifier is given.

Aqua Tart

3:24 AM

This way I don't have to use Popen.

Nathan Merrill

@El'endiaStarman I'm saying to go one step further than python. Variables should always be of the same type

Aqua Tart

I can literally just take every file in tests/ and do pytek_parser.AST_parse(blah).run()

Nathan Merrill

aka, a=1;a="var"; throws an error

Aqua Tart

@NathanMerrill Why?

El'endia Starman

No, I like the ability to change a variable's type.

@AquaTart It's not that bad... :P

Aqua Tart

3:25 AM

@El'endiaStarman Yes it is

Nathan Merrill

it prevents a host of errors, and allows better optimization (I know we aren't working on that know, but you have to consider it for the future)

El'endia Starman

That's why I allow static type enforcement through the use of type identifiers.

Nathan Merrill

I can't think of a use case for wanting to change a type of a variable

unless you're trying to conserve the number of variables you are using (which is a terrible idea)

Aqua Tart

@NathanMerrill I can but probably not for the reasons you're thinking

Static typing = collections have to be of one type

nasty

El'endia Starman

> When languages are designed for other people, it's always a specific group of other people: people not as smart as the language designer. So you get a language that talks down to you.

> Many languages (especially the ones designed for other people) have the attitude of a governess: they try to prevent you from doing things that they think aren't good for you. I like the opposite approach: give the programmer as much control as you can.

Nathan Merrill

3:30 AM

@El'endiaStarman that's the entire reason C is a terrible language

they let you do anything you want, and will only throw an error when it has to

El'endia Starman

And it's also one of the most powerful.

What's your favorite language(s)?

Nathan Merrill

only because its fast

Python

El'endia Starman

Huh, wouldn't have expected that, actually.

quintopia

what is pytek m8

Aqua Tart

@NathanMerrill But C is statically typed.

El'endia Starman

3:32 AM

@AquaTart: On Linux, how do I search for a specific phrase in files?

quintopia

aqua tart... nice anagram

Aqua Tart

@El'endiaStarman grep

Nathan Merrill

right, but C also lets you access an array out of bounds

El'endia Starman

@NathanMerrill That has nothing to do with static/dynamic typing.

Nathan Merrill

@El'endiaStarman right, but the mantra of "give the programmer as much control as you can" is the reason they don't do bound checking

(that and efficieny)

El'endia Starman

3:33 AM

Well, C is also a pretty old language.

They didn't terminate strings automatically either...

Nathan Merrill

I actually think that C is a great language, because it has a very good focus on what it wants

but from a design standpoint, its pretty out of date :)

quintopia

strong typing is good

so what does this pytek thing do that haskell can't?

El'endia Starman

Hmm. Let's put it this way: if a design detail is intended to prevent bad things, it's probably a good idea. If it's intended to prevent "Why would you do that?" decisions that aren't intrinsically good or bad, then maybe it's not needed.

Aqua Tart

@quintopia What does Haskell do that assembly can't?

:P

quintopia

@AquaTart loads :)

Aqua Tart

3:36 AM

But to answer your question, you can look at our spec.

And the transcript

quintopia

i just did

Nathan Merrill

@El'endiaStarman I can think of plenty of potential of bad things that could happen from allowing dynamic typing.

quintopia

well obvs not the whole transcript...that's too many things. I'm begging for a gistier gist

El'endia Starman

@NathanMerrill Python developers seem to do just fine with it.

Nathan Merrill

I've run into bugs where I've assumed the wrong type in an array

and I imagine many other python developers have as well

El'endia Starman

3:37 AM

@quintopia The first few pages are probably the most useful with regards to getting an idea of what I was thinking when I started designing the language.

Aqua Tart

@NathanMerrill This is degraded into a dynamic vs static typed argument, which is one that people have been arguing about since the dawn of time when dinosaurs roamed the land

There's no easy answer

I'm favoring dynamic typing if only because it's easier to write short code for

Nathan Merrill

@AquaTart fair point

quintopia

@El'endiaStarman what does it mean: - \funcName= will set the first argument in (arguments) to the return value of the function.

Aqua Tart

We are providing static typing so people who don't like bugs can use that if they need to

Nathan Merrill

but I'd argue actually for what facebook does

El'endia Starman

3:39 AM

@quintopia Intended to be an analogue of += and the like.

quintopia

oh. makes sense.

Nathan Merrill

everything you do implies a type

so, if you have a function(a){return a.someVal}, then the type of a would be has(someVal)

or something along those lines

Aqua Tart

4:14 AM

@El'endiaStarman Welp, I'll leave you to that "distraction-free coding" you were talking about ;)

quintopia

@AquaTart how bout some distraction-free gaming

El'endia Starman

@AquaTart Haha, thanks. I'm gonna be Skyping with my fiancée soon though, just as soon as I finish this argparse thing.

PhiNotPi

If there's people with nothing to do right now (since Pytek development is suspended due to Skype), I am welcome to hear your suggestions for my own language. Just be warned: it doesn't exist yet.

El'endia Starman

4:37 AM

You can now run Pytek files like python3 pytek_main.py ../tests/test5.pytek or like python3 pytek_main.py -c "\print(\add(2,2))", and -c can be --code.

@AquaTart: ^ yay

I think unit tests should be run with the -c option so we don't have to make separate files for all of them.

El'endia Starman

10:20 AM

Glorious!

Coming soon: everything.

Aqua Tart

4:28 PM

@El'endiaStarman Done except Pytek is throwing a strange decoding error with -c:

Traceback (most recent call last):
  File "/usr/local/lib/python3.4/dist-packages/cx_Freeze/initscripts/Console.py", line 27, in <module>
    exec(code, m.__dict__)
  File "../src/pytek_main.py", line 19, in <module>
  File "/usr/lib/python3.4/codecs.py", line 319, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xfc in position 24: invalid start byte

Aqua Tart

6:12 PM

@El'endiaStarman -1 for including jQuery

Aqua Tart

9:16 PM

I was thinking it would be nice if we made a preprocessor that ran before the code was parsed.

Have stuff like @define @include and what not

El'endia Starman

@AquaTart Did you edit unit_tester or just run it from the terminal?

Nevermind, it's not working for me either.

I hate it when things break overnight.

Aqua Tart

9:55 PM

It's something wacky with Unicode

I see you changed it back for now

El'endia Starman

I dunno what happened with unit_tester if you changed it.

Aqua Tart

?

Aren't you the person who changed it back to not use -c?

Or do we have poltergeists?

El'endia Starman

I did nothing with unit_tester.py.

Aqua Tart

wtf

El'endia Starman

Last night or today.

Maybe @Mego is to blame?

Aqua Tart

10:05 PM

Nope looking at it right now and it has orange highlighting

Anyways.

El'endia Starman

unit_tester.py? All purple, which makes sense, since you were the one that made it.

But yeah, I'm gonna go ahead and make sure the -c thing absolutely works.

Aqua Tart

El'endia Starman

.........poltergeists.

Aqua Tart

poltergeists

El'endia Starman

Alright, fixed.

Well, in the terminal, at least.

Put quotes around code if you use -c.

Aqua Tart

10:10 PM

OK.

Anonymous

@El'endiaStarman I didn't do nothin

Anonymous

I just looked around the code for a few minutes last night

El'endia Starman

Definitely poltergeists then.

Anonymous

blame-doorknob

Aqua Tart

11:02 PM

@El'endiaStarman Should change the room tags to pytek language-design awsum poltergeists

:P

El'endia Starman

room topic changed to Pytek: Discussion for this in-development language [awsum] [language-design] [poltergeists] [pytek]

Aqua Tart

perfect

What an accurate summary.

El'endia Starman

New potential name: Polty. Or Poltek.

Aqua Tart

Hmm something is going wrong with unit tester

El'endia Starman

Taking a look...

Aqua Tart

11:10 PM

Pytek is just not outputting anything: b''

El'endia Starman

Okay. I think that what happened was that when you do -c in the terminal, you have to put the code in quotes because otherwise bash tries to do stuff with \p or whatever. With this method, using Python to do it, the code has already been "chunked" into its own argument.

Aqua Tart

Oh, OK.

Yep, all good.

El'endia Starman

Another thing to consider: also toss in code that should fail, and the expected outputs would be the expected errors.

For instance, \ print is a syntax error.

Aqua Tart

Made the output a little prettier

El'endia Starman

Nice, though it's a little harder to see which ones succeeded and which ones failed at first glance.

I'd prefer to make successes be invisible. As they say, no news is good news.

Aqua Tart