Coding Horrors: You can gather all your technical debt into one easy framework!

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Cavern of COBOL > Coding Horrors: You can gather all your technical debt into one easy framework!

«‹›1503 »

Soricidus: Oct 21, 2010; freedom-hating statist shill

Volguus posted:

Hahaha. This reminds me of another JSP page that I had the pleasure of seeing a few years back: It was holding the currently logged in users in a static java.util.Hashtable. 2 people could not log in/browse at the exact same time or the application would crash.

Why did it crash? Hashtable is supposedly thread-safe so it can�t be that.

# ? Jun 8, 2019 11:37

Adbot: ADBOT LOVES YOU

# ? May 17, 2024 15:06

Volguus: Mar 3, 2009

Soricidus posted:

Why did it crash? Hashtable is supposedly thread-safe so it can�t be that.

Hashtable is not thread safe. Methods in it are synchronized, that's true, so you can call "put" from 2 different threads and it works fine. But you can't iterate over it while trying to call put, for example. In general, in java, if you need a thread safe map one should use ConcurrentHashMap.

# ? Jun 8, 2019 15:23

Nth Doctor: Sep 7, 2010; Darkrai used Dream Eater!
It's super effective!

Volguus posted:

Hahaha. This reminds me of another JSP page that I had the pleasure of seeing a few years back: It was holding the currently logged in users in a static java.util.Hashtable. 2 people could not log in/browse at the exact same time or the application would crash. That was in addition to its many other issues.

"Can you fix it?" was the question.
"No" was the only possible answer that i could give.

We had something similarly horrific in our DBManager class through which all db connectivity in the platform flowed. As a manager it was a Singleton, but one of the class level static fields of the thing was WHO IS CONNECTING. Normally this wasn't a problem because we would at least set the thing at the last possible moment before connections opened, until we happened to have two requests come in at the exact same millisecond. Ugh that was a bad night.

Thankfully my only involvement in it was in the aftermath because I was one of the few devs who knew how to read packet traces.

# ? Jun 9, 2019 00:57

Jabor: Jul 16, 2010; #1 Loser at SpaceChem

Volguus posted:

Hashtable is not thread safe. Methods in it are synchronized, that's true, so you can call "put" from 2 different threads and it works fine. But you can't iterate over it while trying to call put, for example. In general, in java, if you need a thread safe map one should use ConcurrentHashMap.

Hashtable has the worst kind of thread-safety, in that you pay the performance cost of locking but don't actually get any beneficial semantics in exchange.

# ? Jun 9, 2019 09:52

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

I think you can read and write elements atomically; the problem is that that�s not really an adequate API for a concurrent dictionary. But iterating isn�t a good example of why, both because IIRC you can safely iterate it (by synchronizing on it yourself) and because iteration is not an operation you should generally be using on concurrent data structures.

# ? Jun 9, 2019 21:52

Munkeymon: Aug 14, 2003; Motherfucker's got an
armor-piercing crowbar! Rigoddamndicu𝜆ous.

Java code:

@DontIterateMeBro
public class HashTable

# ? Jun 10, 2019 17:35

Cuntpunch: Oct 3, 2003; A monkey in a long line of kings

code:

public class HttpClientFactory : HttpClient
{
	public HttpClientFactory(string uri)
	{
		BaseAddress = new Uri(uri);
	}
}

As seen in such classic usage cases such as

code:

var httpClient = new HttpClientFactory("http://some.address")
{
	Timeout = TimeSpan.FromMinutes(10)
};

# ? Jun 12, 2019 07:32

No Safe Word: Feb 26, 2005

Cuntpunch posted:

code:

public class HttpClientFactory : HttpClient
{
	public HttpClientFactory(string uri)
	{
		BaseAddress = new Uri(uri);
	}
}

As seen in such classic usage cases such as

code:

var httpClient = new HttpClientFactory("http://some.address")
{
	Timeout = TimeSpan.FromMinutes(10)
};

smh if you don't build a new factory to produce every single widget

# ? Jun 12, 2019 18:15

Kuule hain nussivan: Nov 27, 2008

Just found out that our software provider who have trouble providing fixes for relatively simple bugs, have decided to not use on open source map solution with the new GIS module, but will instead roll out their own implementation. Fun times ahead.

# ? Jun 13, 2019 18:28

Happy Thread: Jul 10, 2005; by Fluffdaddy; Plaster Town Cop

Do they have access to ships, cartographers, or constellation-tracking equipment? Or are they gonna drive around in their car taking GPS readings

# ? Jun 13, 2019 18:48

Kuule hain nussivan: Nov 27, 2008

Dumb Lowtax posted:

Do they have access to ships, cartographers, or constellation-tracking equipment? Or are they gonna drive around in their car taking GPS readings

They did complain about not having enough time to answer the issues we report, so they must be busy with something else.

# ? Jun 13, 2019 19:13

Volguus: Mar 3, 2009

code:

if (s1 BETTER_THAN WORST_SCORE)

with

code:

/**
 * Is one score better than another?
 */
#define BETTER_THAN >

# ? Jun 13, 2019 21:14

Nolgthorn: Jan 30, 2001; The pendulum of the mind alternates between sense and nonsense

Do they also define IS_THE_SAME because if so I'm behind them.

# ? Jun 13, 2019 21:30

Volguus: Mar 3, 2009

Nolgthorn posted:

Do they also define IS_THE_SAME because if so I'm behind them.

No, only

code:

#define WORSE_THAN <

because of course.

# ? Jun 13, 2019 21:36

CPColin: Sep 9, 2003; Big ol' smile.

At least it's not:

code:

#define BETTER_THAN >=
#define WORSE_THAN <=

# ? Jun 13, 2019 21:44

redleader: Aug 18, 2005; Engage according to operational parameters

Y'all are unnecessarily afraid of some literate programming.

# ? Jun 13, 2019 21:45

Scaramouche: Mar 26, 2001; SPACE FACE! SPACE FACE!

I'm being triggered by these worth assignments, who's to say one number is better than another??

# ? Jun 13, 2019 21:48

Dylan16807: May 12, 2010

CPColin posted:

At least it's not:

code:

#define BETTER_THAN >
#define WORSE_THAN <=

[...200 lines...]

#define BETTER_THEN >=
#define WORSE_THEN <=

# ? Jun 13, 2019 21:49

Doom Mathematic: Sep 2, 2008

Scaramouche posted:

I'm being triggered by these worth assignments, who's to say one number is better than another??

You're having a panic attack because of bad code?

# ? Jun 13, 2019 22:17

Scaramouche: Mar 26, 2001; SPACE FACE! SPACE FACE!

Doom Mathematic posted:

You're having a panic attack because of bad code?

More of a Guru Meditation

# ? Jun 13, 2019 22:33

Phobeste: Apr 9, 2006; never, like, count out Touchdown Tom, man

Dylan16807 posted:

If you did

code:

#define BETTER_THEN(x,y) if (x>=y) {

Then imagine the pseudocode you could write

# ? Jun 13, 2019 23:43

pokeyman: Nov 26, 2006; That elephant ate my entire platoon.

Subtly grammatically incorrect literate programming, now there�s a trend I could jump on!

# ? Jun 14, 2019 00:00

iospace: Jan 19, 2038

# ? Jun 14, 2019 00:02

The MUMPSorceress: Jan 6, 2012; ^SHTPSTS; Gary’s Answer

iospace posted:

So I assume there's potential for very slightly negative floats to compare as not less than 0 and they solved this issue in the most ridiculous way possible?

# ? Jun 14, 2019 00:21

MrMoo: Sep 14, 2000

I've seen code like that in production, namely the lack of a free alongside stupid things like calling strdup doing one test and abandoning the memory.

# ? Jun 14, 2019 00:23

iospace: Jan 19, 2038

jit bull transpile posted:

So I assume there's potential for very slightly negative floats to compare as not less than 0 and they solved this issue in the most ridiculous way possible?

You tell me (yes that's python and not a C-like, sue me)

# ? Jun 14, 2019 00:24

iospace: Jan 19, 2038

MrMoo posted:

I've seen code like that in production, namely the lack of a free alongside stupid things like calling strdup doing one test and abandoning the memory.

Could be Java :v:

# ? Jun 14, 2019 00:32

Jeb Bush 2012: Apr 4, 2007; A mathematician, like a painter or poet, is a maker of patterns. If his patterns are more permanent than theirs, it is because they are made with ideas.

jit bull transpile posted:

So I assume there's potential for very slightly negative floats to compare as not less than 0 and they solved this issue in the most ridiculous way possible?

I don't think so, but it is different from return (x < 0.0) in exactly one situation

-0.0 is a valid floating point value, and -0.0 < 0.0 evaluates to false

# ? Jun 14, 2019 00:36

iospace: Jan 19, 2038

Jeb Bush 2012 posted:

I don't think so, but it is different from return (x < 0.0) in exactly one situation

-0.0 is a valid floating point value, and -0.0 < 0.0 evaluates to false

-0 is still 0. So therefore, it's false because it's less than, not less than or equal to

:goonsay:

# ? Jun 14, 2019 00:46

Jeb Bush 2012: Apr 4, 2007; A mathematician, like a painter or poet, is a maker of patterns. If his patterns are more permanent than theirs, it is because they are made with ideas.

iospace posted:

-0 is still 0. So therefore, it's false because it's less than, not less than or equal to

oh sure, I'm not saying it's the wrong answer (and if the output of your code depends on whether -0.0 < 0.0, something has almost certainly gone very wrong), just that it's the only situation I can think of where the weird stringcasting code would give a different answer from returning x < 0.0

# ? Jun 14, 2019 00:55

JawnV6: Jul 4, 2004; So hot ...

jit bull transpile posted:

So I assume there's potential for very slightly negative floats to compare as not less than 0 and they solved this issue in the most ridiculous way possible?

No. There's a sign bit. That doesn't change for the subnormal numbers, they just needed to check it.

Floating point does a lot with the space it has. The biggest trick is a hidden '1', every binary floating point number would look like 1.010101 x 2¹⁰¹⁰¹, so the leading "1." is dropped. To represent those very small numbers the exponent of '0' indicates that there's no leading "1.".

I thought it might have something to do with not caring if the input was a 32bit float or 64bit? A 20-character string is properly sized for a 64bit float, which can expand to 17 significant decimal digits plus 3 for the the sign, decimal point, and terminating \0. But I think it's just a waste of 8 bytes, per call.

The most obvious optimization is using 'snprintf' with the size set to 1 to (hopefully) bail once the stringifier has sussed out that sign bit.

# ? Jun 14, 2019 01:05

Absurd Alhazred: Mar 27, 2010; by Athanatos

Kuule hain nussivan posted:

Just found out that our software provider who have trouble providing fixes for relatively simple bugs, have decided to not use on open source map solution with the new GIS module, but will instead roll out their own implementation. Fun times ahead.

Ahaha. AHAHAHAHA. AHAHAHAHAHAHAHAHAHA

Sure. Yes. Professional software outfits at the very least use established open-source libraries like GDAL if they're cheap, but sure, I bet this one podunk outfit is going to outdo all of them. :allears:

# ? Jun 14, 2019 02:11

The MUMPSorceress: Jan 6, 2012; ^SHTPSTS; Gary’s Answer

JawnV6 posted:

No. There's a sign bit. That doesn't change for the subnormal numbers, they just needed to check it.

Floating point does a lot with the space it has. The biggest trick is a hidden '1', every binary floating point number would look like 1.010101 x 2¹⁰¹⁰¹, so the leading "1." is dropped. To represent those very small numbers the exponent of '0' indicates that there's no leading "1.".

I thought it might have something to do with not caring if the input was a 32bit float or 64bit? A 20-character string is properly sized for a 64bit float, which can expand to 17 significant decimal digits plus 3 for the the sign, decimal point, and terminating \0. But I think it's just a waste of 8 bytes, per call.

The most obvious optimization is using 'snprintf' with the size set to 1 to (hopefully) bail once the stringifier has sussed out that sign bit.

Cool. I know just enough about float to understand precision issues but wasn't sure if that extended to comparisons between zero and a really small negative number.

# ? Jun 14, 2019 03:35

Xarn: Jun 26, 2015

no you don't. nobody does. :suicide:

(For real though, to actually understand precision issues, you have to be domain expert in floating point computation and in whatever you computation is doing, because all of exact comparison, absolute margin comparison, relative margin comparison, ULP comparison, can be the right or wrong choice, dependent on what you are actually doing)

# ? Jun 14, 2019 09:02

MononcQc: May 29, 2007

Volguus posted:

code:

if (s1 BETTER_THAN WORST_SCORE)

with

code:

/**
 * Is one score better than another?
 */
#define BETTER_THAN >

How else are you going to make the code work both for basketball and golf?

# ? Jun 14, 2019 13:52

Beef: Jul 26, 2004

And how else am I going to make absolutely sure the compiler will inline it?

# ? Jun 14, 2019 15:01

VikingofRock: Aug 24, 2008

MononcQc posted:

How else are you going to make the code work both for basketball and golf?

Easy.

code:


#define GOLF 1
#define BASKETBALL 2

Now the code correctly shows that BASKETBALL BETTER_THAN GOLF.

# ? Jun 14, 2019 16:07

Munkeymon: Aug 14, 2003; Motherfucker's got an
armor-piercing crowbar! Rigoddamndicu𝜆ous.

VikingofRock posted:

Easy.
code:
#define GOLF 1
#define BASKETBALL 2
Now the code correctly shows that BASKETBALL BETTER_THAN GOLF.

You can just go ahead and #define GOLF LONG_MIN

# ? Jun 14, 2019 19:03

Bonfire Lit: Jul 9, 2008; If you're one of the sinners who caused this please unfriend me now.

Jeb Bush 2012 posted:

I don't think so, but it is different from return (x < 0.0) in exactly one situation

-0.0 is a valid floating point value, and -0.0 < 0.0 evaluates to false

Your standard library may write a NaN value with its sign bit set as "-nan", but since it's a NaN that comparison is still false. So the function differs from a plain compare in at least two situations :eng101:

# ? Jun 14, 2019 19:52

Adbot: ADBOT LOVES YOU

# ? May 17, 2024 15:06

Jeb Bush 2012: Apr 4, 2007; A mathematician, like a painter or poet, is a maker of patterns. If his patterns are more permanent than theirs, it is because they are made with ideas.

Bonfire Lit posted:

Your standard library may write a NaN value with its sign bit set as "-nan", but since it's a NaN that comparison is still false. So the function differs from a plain compare in at least two situations

maybe. python prints NaN with a set sign bit as just "nan", but I haven't tried writing it in C (and it could be implementation-dependent anyway I guess)

# ? Jun 14, 2019 20:06

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Cavern of COBOL > Coding Horrors: You can gather all your technical debt into one easy framework!

«‹›1503 »