p-lang thread: (now (have you (problems two)))

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > p-lang thread: (now (have you (problems two)))

«‹›1784 »

Zlodo: Nov 25, 2006

Share Bear posted:

Maybe I should stick to the terrible programmer thread

python chat belongs in both

# ? Jul 11, 2019 18:47

Adbot: ADBOT LOVES YOU

# ? May 17, 2024 07:06

VikingofRock: Aug 24, 2008

Share Bear posted:

OK so what is the preferred way to transform another class or dict or other list into a new list? I thought this was pretty snappy.

My read is that Xarn is saying that adding a comment explaining the list comprehension is terrible, not that the list comprehension itself is terrible.

Personally, I like list comprehensions so long as they don't get too complicated. I think a lot of times they are more readable and less error prone than the equivalent for loops. For example I'd much rather read

Python code:


valid_foos = [foo for foo in foos if valid(foo)]

than

Python code:


valid_foos = []
for foo in foos:
  if valid(foo):
    valid_foos.append(foo)

because to me the latter seems needlessly verbose and stateful

# ? Jul 11, 2019 19:03

ynohtna: Feb 16, 2007; backwoods compatible; Illegal Hen

did you just tell me to go foo myself?

# ? Jul 11, 2019 19:15

Cybernetic Vermin: Apr 18, 2005

i kind of like list comprehensions, but tbqh i don't think they improve on map/filter/fold-chains enough to be a real improvement. like, the natural language'y look they have is sort of pleasing and gives the *appearance* of being easier, but teaching python they didn't seem any less of a stumbling block, and it does amount to additional magical syntax.

# ? Jul 11, 2019 19:30

Notorious b.s.d.: Jan 25, 2003; by Reene

VikingofRock posted:

My read is that Xarn is saying that adding a comment explaining the list comprehension is terrible, not that the list comprehension itself is terrible.

Personally, I like list comprehensions so long as they don't get too complicated. I think a lot of times they are more readable and less error prone than the equivalent for loops. For example I'd much rather read
Python code:
valid_foos = [foo for foo in foos if valid(foo)]
than
Python code:
valid_foos = []
for foo in foos:
  if valid(foo):
    valid_foos.append(foo)
because to me the latter seems needlessly verbose and stateful

i'd rather see

code:

valid_foos = foos.select ...

you know, like a sane language with higher-order functions

bonus points if it's typesafe, which python never is

# ? Jul 11, 2019 19:41

brap: Aug 23, 2004; Grimey Drawer

I really enjoy the way python throws if you concatenate a number to a string as if this behavior is type safety

# ? Jul 11, 2019 19:46

Zlodo: Nov 25, 2006

sometimes i visit amiga forums out of morbid curiosity and they have decided to reboot that video game making basic that existed back then but now it transpiles to javascript in case anyone wanted that

also apparently to do anything useful you have to inline some javascript anyway

"The simplicity of AMOS, the speed of Javascript!"

http://amos2.tech/Code-Showcase/Animated-Julia

http://amos2.tech/Code-Showcase/AMOS-Ball

im the Boob Set Hotspot

# ? Jul 11, 2019 19:55

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

lol nice. I hosed around with a used* copy of Amos pro 4 back in the day

* yes, in the original box w manuals and all

# ? Jul 11, 2019 21:09

Pie Colony: Dec 8, 2006; I AM SUCH A FUCKUP THAT I CAN'T EVEN POST IN AN E/N THREAD I STARTED

brap posted:

I really enjoy the way python throws if you concatenate a number to a string as if this behavior is type safety

smh that is type safety, if you want build-time type safety use python�s type checker

# ? Jul 11, 2019 22:09

Pie Colony: Dec 8, 2006; I AM SUCH A FUCKUP THAT I CAN'T EVEN POST IN AN E/N THREAD I STARTED

Notorious b.s.d. posted:

i'd rather see
code:
valid_foos = foos.select ...
you know, like a sane language with higher-order functions

bonus points if it's typesafe, which python never is

lol, zero programming languages are sane, there�s just the set you know how to use and the set you don�t

# ? Jul 11, 2019 22:11

VikingofRock: Aug 24, 2008

Notorious b.s.d. posted:

i'd rather see
code:
valid_foos = foos.select ...
you know, like a sane language with higher-order functions

bonus points if it's typesafe, which python never is

Yeah, agreed on that point. There is

Python code:


valid_foos = list(filter(valid, foos))

and frequently you can drop the list() call if you only need to iterate it once. I understand the motivation for making filter() a built-in instead of a member function for the different iterables, but I think I'd still prefer it as a member function in the end when possible.

Regardless, I was mostly responding to the idea that list comprehensions are bad python style or particularly hard to read, since I disagree with both of those notions.

# ? Jul 11, 2019 22:43

animist: Aug 28, 2018

Pie Colony posted:

smh that is type safety, if you want build-time type safety use python’s type checker

which has no typings for 90% of the ecosystem :argh:

# ? Jul 12, 2019 00:21

ErIog: Jul 11, 2001

VikingofRock posted:

Regardless, I was mostly responding to the idea that list comprehensions are bad python style or particularly hard to read, since I disagree with both of those notions.

I think the minimal example cases that get used to say, "See! List comprehensions are so easy!" are a bad framing of the problem because that's not really what people are complaining about when they say they're complicated. In isolation, a simple list comprehension is fine. However, when you extend it even slightly with multiple for statements, complicated anonymous functions, or functions with non-obvious purpose it degenerates very quickly. For the simplest cases list comprehensions are easier to understand than their equivalent expanded for loop versions. However, for the more complicated cases the list comprehensions just become a kind of terse obfuscation.

The other thing which also maybe happens is that the list comprehensions get out of hand over time rather than all at once. Somebody thinks what they're doing is fundamentally simple and very obvious so they use a simple list comprehension, but then the list comprehension gets more and more complicated as the developer modifies it to meet the more complicated needs as they realize they have them. The developer, at that point, probably should just replace it with some for loops to spread the complexity out over more lines with more opportunities for helpful comments. However, when working on code people often just go with inertia, making modifications to code rather than rewriting or restructuring it.

In a world where developers are very good and mindful about writing readable code list comprehensions would be fine, but in practice it seems like they're yet another case of a Python feature that makes it easier to write bad code than good code.

# ? Jul 12, 2019 00:48

Progressive JPEG: Feb 19, 2003

Pie Colony posted:

smh that is type safety, if you want build-time type safety use python�s type checker

but I thought the whole point of PLs was to shift as many errors as possible into run time?

# ? Jul 12, 2019 01:13

Progressive JPEG: Feb 19, 2003

I mean why use a PL if you aren't going to leave little typing time bombs strewn throughout your code?

# ? Jul 12, 2019 01:14

theodop: Dec 30, 2005; rock solid, heart touching

ErIog posted:

I think the minimal example cases that get used to say, "See! List comprehensions are so easy!" are a bad framing of the problem because that's not really what people are complaining about when they say they're complicated. In isolation, a simple list comprehension is fine. However, when you extend it even slightly with multiple for statements, complicated anonymous functions, or functions with non-obvious purpose it degenerates very quickly. For the simplest cases list comprehensions are easier to understand than their equivalent expanded for loop versions. However, for the more complicated cases the list comprehensions just become a kind of terse obfuscation.

The other thing which also maybe happens is that the list comprehensions get out of hand over time rather than all at once. Somebody thinks what they're doing is fundamentally simple and very obvious so they use a simple list comprehension, but then the list comprehension gets more and more complicated as the developer modifies it to meet the more complicated needs as they realize they have them. The developer, at that point, probably should just replace it with some for loops to spread the complexity out over more lines with more opportunities for helpful comments. However, when working on code people often just go with inertia, making modifications to code rather than rewriting or restructuring it.

In a world where developers are very good and mindful about writing readable code list comprehensions would be fine, but in practice it seems like they're yet another case of a Python feature that makes it easier to write bad code than good code.

if I had a code review and it had LINQ which was more complicated than butts.Where(butt -> butt.HasFart).OrderBy(butt -> butt.Size).ToArray() I would fail it and ask for every chained function to be split out to its own variable. Forcing people to name the variables means 1) they have to think about what each step is actually achieving and 2) it implicitly documents it. it's harder to do the same thing with their bizarre list comprehension syntax but fortunately nobody uses that in .NET.

I've seen too many monstrosities that chain together 5-10 filters and maps and just because they've used line breaks doesn't make it less of an assache to troubleshoot

# ? Jul 12, 2019 01:17

Share Bear: Apr 27, 2004

theodop posted:

if I had a code review and it had LINQ which was more complicated than butts.Where(butt -> butt.HasFart).OrderBy(butt -> butt.Size).ToArray() I would fail it and ask for every chained function to be split out to its own variable. Forcing people to name the variables means 1) they have to think about what each step is actually achieving and 2) it implicitly documents it. it's harder to do the same thing with their bizarre list comprehension syntax but fortunately nobody uses that in .NET.

even though ive been championing list comps, 100% this

�clever� or �clean� code is usually worse than either explicit poo poo or �clever� poo poo with comments

# ? Jul 12, 2019 01:57

luchadornado: Oct 7, 2004; A boombox is not a toy!

adding comments to poo poo code just means its poo poo code with comments

# ? Jul 12, 2019 02:14

Notorious b.s.d.: Jan 25, 2003; by Reene

VikingofRock posted:

and frequently you can drop the list() call if you only need to iterate it once. I understand the motivation for making filter() a built-in instead of a member function for the different iterables, but I think I'd still prefer it as a member function in the end when possible.

focusing on built-in vs member function misses the forest for the trees

the problem is that python treats higher-order functions as "un-pythonic" and doing super basic poo poo that is trivial in c# or java requires you to jump through dumb hoops

even ruby and perl do this better. (not that it matters, because ruby and perl are terrible.)

# ? Jul 12, 2019 02:20

Share Bear: Apr 27, 2004

Helicity posted:

adding comments to poo poo code just means its poo poo code with comments

explicit poo poo is better than poo poo with nothing going for it, because not ever org has time to refactor

comments are usually good

pragmatism not idealism, will not cry in public

# ? Jul 12, 2019 02:23

Share Bear: Apr 27, 2004

everyone's pretending they're not constantly turning out poo poo that sucks, because their poo poo that sucks is better than other people's poo poo that sucks

addressing that all this poo poo sucks is worse than being quiet

# ? Jul 12, 2019 02:26

Pie Colony: Dec 8, 2006; I AM SUCH A FUCKUP THAT I CAN'T EVEN POST IN AN E/N THREAD I STARTED

Notorious b.s.d. posted:

focusing on built-in vs member function misses the forest for the trees

the problem is that python treats higher-order functions as "un-pythonic" and doing super basic poo poo that is trivial in c# or java requires you to jump through dumb hoops

even ruby and perl do this better. (not that it matters, because ruby and perl are terrible.)

what exactly are you complaining about? what is trivial in java that requires dumb hoops in python, or are you saying list comprehensions are dumb hoops?

no one is saying you have to write your python code any certain way. you can still use the built-in functions or the functools module. but like all functional code, it'll be slower than the alternative

# ? Jul 12, 2019 02:32

Pie Colony: Dec 8, 2006; I AM SUCH A FUCKUP THAT I CAN'T EVEN POST IN AN E/N THREAD I STARTED

the python defender has logged on

# ? Jul 12, 2019 02:32

Notorious b.s.d.: Jan 25, 2003; by Reene

Helicity posted:

adding comments to poo poo code just means its poo poo code with comments

poo poo code with detailed comments is better than poo poo code without detailed comments

but...

if you know you are doing something tricky and hard to understand, maybe you should try and make it simpler?

# ? Jul 12, 2019 02:35

Pie Colony: Dec 8, 2006; I AM SUCH A FUCKUP THAT I CAN'T EVEN POST IN AN E/N THREAD I STARTED

drat i wrote some dumb poo poo in python when i was in undergrad

code:

def curry(x, c=None):
    if c is None:
        c = x.func_code.co_argcount
    x.__curried = True
 
    def p(*a):
        if len(a) > c:
            f = x(*a[:c])
            if getattr(f, '__curried', False):
                f = curry(f)
            return f(*a[c:])
        elif len(a) == c:
            return x(*a)
 
        def q(*b):
            return x(*(a + b))
 
        q.__name__ = p.__name__[:-3] + ' -> '.join([''] + [str(b) for b in a])
        return curry(q, c - len(a))
   
    p.__name__ = '%s...' % x.__name__
    return p
 
>>> id = curry(lambda x: x)
>>> const = curry(lambda x, y: x)
>>> const(3)
<function <lambda> -> 3... at 0x106e97a28>
>>> _(4) == const(3,4)
True
>>> comp = curry(lambda f, g: lambda x: f(g(x)))
>>> id(comp, lambda x: x + 2, lambda x: x * 4, 100)
402

is this python or haskell? you can't even tell the difference. i also implemented tail call optimization

# ? Jul 12, 2019 02:44

mystes: May 31, 2006

Python gets pretty unreadable when you try to do stuff like that. Also, isn't that just partial application rather than currying?

mystes fucked around with this message at 02:52 on Jul 12, 2019

# ? Jul 12, 2019 02:49

rjmccall: Sep 7, 2007; no worries friend; Fun Shoe

drat, imagine worrying about making your python code slow

# ? Jul 12, 2019 03:08

Pie Colony: Dec 8, 2006; I AM SUCH A FUCKUP THAT I CAN'T EVEN POST IN AN E/N THREAD I STARTED

no you see, modifying your functions' ASTs at runtime actually makes them better

code:

from ast import *
from inspect import getsource

class TrampolineTransform(NodeTransformer):
    def visit_Return(self, node):
        return_val = node.value

        if return_val.__class__ == Call:
            func = return_val.func
            func_args = return_val.args
            return copy_location(Return(
                value=Tuple(elts=[
                    Str(s='__trampoline'), 
                    func, 
                    List(elts=func_args, ctx=Load())
                ], ctx=Load()), ctx=Load()
            ), node)

        return node


def trampoline(start):
    ret = start()
    while isinstance(ret, tuple) and ret[0] == '__trampoline':
        _, func, args = ret
        ret = func(*args)
    return ret


def odd(n):
    if n == 1: return True
    return even(n-1)


def even(n):
    if n == 1: return False
    return odd(n-1)


def compile_tco(f):
    ast = parse(getsource(f))
    tco_ast = fix_missing_locations(TrampolineTransform().visit(ast))
    return compile(tco_ast, __name__, 'exec')

exec compile_tco(odd)
exec compile_tco(even)

>>> print trampoline(lambda: even(59392))
True

Pie Colony fucked around with this message at 03:34 on Jul 12, 2019

# ? Jul 12, 2019 03:32

Lutha Mahtin: Oct 10, 2010; Your brokebrain sin is absolved...go and shitpost no more!

Nomnom Cookie posted:

isnt that perl. p sure i even read a quote some years back from larry wall about how he tried to make writing perl similar to a natural language--highly contextual, sometimes ambiguous, but expressive

one of the big choices of larry wall's life was when he decided between (a) designing a computer language, or (b) moving to a remote tribal area of the world, learning the language, creating a script for it, and translating the Bible into it (he's Church of the Nazarene). he chose (a) and the rest of us get to deal with it lol

# ? Jul 12, 2019 04:42

brap: Aug 23, 2004; Grimey Drawer

Pie Colony posted:

smh that is type safety, if you want build-time type safety use python�s type checker

it's not type safety, it's just a limitation. look at how any other language handles the scenario of `someString + someNumber`. it's a matter of what combinations of operand types are supported. python just made a choice that results in crappy ergonomics in practice.

# ? Jul 12, 2019 04:50

mystes: May 31, 2006

It's reasonable not to allow that for type safety but it makes less sense in a language where you can multiply strings by numbers.

# ? Jul 12, 2019 05:02

Kazinsal: Dec 13, 2011

mystes posted:

It's reasonable not to allow that for type safety but it makes less sense in a language where you can multiply strings by numbers.

imo if "yospos" * 3 = "yosposyosposyospos" then "yospos" + 4 should = "yosposyosp"

# ? Jul 12, 2019 05:28

Nomnom Cookie: Aug 30, 2009

mystes posted:

It's reasonable not to allow that for type safety but it makes less sense in a language where you can multiply strings by numbers.

you can also multiply lists by numbers

I�ve done this in production code

it made me feel dirty

# ? Jul 12, 2019 05:40

cinci zoo sniper: Mar 15, 2013

Pie Colony posted:

no you see, modifying your functions' ASTs at runtime actually makes them better

code:

from ast import *
from inspect import getsource

class TrampolineTransform(NodeTransformer):
    def visit_Return(self, node):
        return_val = node.value

        if return_val.__class__ == Call:
            func = return_val.func
            func_args = return_val.args
            return copy_location(Return(
                value=Tuple(elts=[
                    Str(s='__trampoline'), 
                    func, 
                    List(elts=func_args, ctx=Load())
                ], ctx=Load()), ctx=Load()
            ), node)

        return node


def trampoline(start):
    ret = start()
    while isinstance(ret, tuple) and ret[0] == '__trampoline':
        _, func, args = ret
        ret = func(*args)
    return ret


def odd(n):
    if n == 1: return True
    return even(n-1)


def even(n):
    if n == 1: return False
    return odd(n-1)


def compile_tco(f):
    ast = parse(getsource(f))
    tco_ast = fix_missing_locations(TrampolineTransform().visit(ast))
    return compile(tco_ast, __name__, 'exec')

exec compile_tco(odd)
exec compile_tco(even)

>>> print trampoline(lambda: even(59392))
True

you will face the court of code for your crimes

# ? Jul 12, 2019 05:43

Gazpacho: Jun 18, 2004; by Fluffdaddy; Slippery Tilde

e: wrang thad

Gazpacho fucked around with this message at 07:07 on Jul 12, 2019

# ? Jul 12, 2019 06:01

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

Kazinsal posted:

imo if "yospos" * 3 = "yosposyosposyospos" then "yospos" + 4 should = "yosposyosp"

checks out

# ? Jul 12, 2019 07:03

eschaton: Mar 7, 2007; Don't you just hate when you wind up in a store with people who are in a socioeconomic class that is pretty obviously about two levels lower than your own?

if only python had something simple and readable like the Common Lisp LOOP macro

# ? Jul 12, 2019 07:12

Xarn: Jun 26, 2015

Kazinsal posted:

imo if "yospos" * 3 = "yosposyosposyospos" then "yospos" + 4 should = "yosposyosp"

# ? Jul 12, 2019 07:12

Progressive JPEG: Feb 19, 2003

Nomnom Cookie posted:

you can also multiply lists by numbers

I�ve done this in production code

it made me feel dirty

python in production? oof

# ? Jul 12, 2019 08:19

Adbot: ADBOT LOVES YOU

# ? May 17, 2024 07:06

Progressive JPEG: Feb 19, 2003

Pie Colony posted:

the python defender has logged on

does python still have that thing where input written by a foreigner suddenly triggers a requirement to blindly add .decode("utf8") to everything until the errors go away?

it's so type safe that some strings are differently typed from other strings

# ? Jul 12, 2019 08:26

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > p-lang thread: (now (have you (problems two)))

«‹›1784 »