C/C++ Programming Questions Not Worth Their Own Thread

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Cavern of COBOL > C/C++ Programming Questions Not Worth Their Own Thread

«‹›640 »

Paniolo: Oct 9, 2007; Heads will roll.

niteice posted:

XInput may also work for you, though it has very specific limitations.

XInput is designed specifically for an Xbox 360 controller so it probably won't work with his racing wheel.

Just use DirectInput.

# ? Apr 5, 2010 01:28

Adbot: ADBOT LOVES YOU

# ? Jun 1, 2024 23:25

Null Pointer: May 20, 2004; Oh no!

It might work if it's a 360 wheel. I've never tried, but XInput technically should support them (XINPUT_DEVSUBTYPE_WHEEL.)

# ? Apr 5, 2010 05:13

Rocko Bonaparte: Mar 12, 2002; Every day is Friday!

I'm trying to work with the unordered_map in Boost for doing some hashing. I think my problem below is more an issue with templates and how to make them behave. This code is from snippets in unordered_map documentation uses to demonstrate the structure. You'll see that when the unordered_map is typedef'd the code is happy. If I try to spell out the full type, it will not compile.

code:

/* 
 * File:   sandbox.cpp
 * Author: rocko
 *
 * Created on April 3, 2010, 9:43 PM
 */

#include <iostream>
#include <stdio.h>
#include <stdlib.h>
#include <boost/lexical_cast.hpp>
#include <boost/unordered_map.hpp>
#include <boost/foreach.hpp>

using namespace std;

/*
 * 
 */
int main(int argc, char** argv) {
    typedef boost::unordered_map<std::string, int> map;
    map x;
    x["one"] = 1;
    x["two"] = 2;
    x["three"] = 3;

//    BOOST_FOREACH(map::value_type i, x) {         // works
    BOOST_FOREACH(boost::unordered_map< std::string, int>::value_type i, x) {   // BOOST_FOREACH passed 3 arguments, but takes just 2
        std::cout<<i.first<<","<<i.second<<"\n";
    }
    
    assert(x.at("one") == 1);
    assert(x.find("missing") == x.end());
    return (EXIT_SUCCESS);
}

I have some of my own code based off this and I get burnt right there. I will be defining my hash through a typedef to get around it, but for the sake of knowing C++'s pedantic nougat, I wondered if there was a way to spell out the type there w/o it failing to compile.

# ? Apr 5, 2010 05:14

OddObserver: Apr 3, 2009

Comma confusing a macro, perhaps?

# ? Apr 5, 2010 05:16

raminasi: Jan 25, 2005; a last drink with no ice

Yeah, BOOST_FOREACH can't handle that.

# ? Apr 5, 2010 07:54

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

Put the type in parens. The preprocessor doesn't recognize < and > as grouping things and so it splits on the comma in your template. Also don't use that god-awful foreach macro (though it's admittedly better than rolling your own).

# ? Apr 5, 2010 13:03

Mustach: Mar 2, 2003; In this long line, there's been some real strange genes. You've got 'em all, with some extras thrown in.

Why do you hate it?

# ? Apr 5, 2010 13:17

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

Mustach posted:

Why do you hate it?

I'll let the source file speak for itself: http://www.boost.org/doc/libs/1_38_0/boost/foreach.hpp

# ? Apr 5, 2010 14:24

Mustach: Mar 2, 2003; In this long line, there's been some real strange genes. You've got 'em all, with some extras thrown in.

That's grosser than I expected.

# ? Apr 5, 2010 14:53

Jose Cuervo: Aug 25, 2004

I am trying to sort an array of numbers and I am using the sort function in C++. Unfortunately, I cannot think of a way to also come up with an array of sorted indices, eg

If my array is initially [9, 8, 6, 17], the sorted array using sort will be [6, 8, 9, 17], and I would also like to get the array of sorted indices: [3, 2, 1, 4].

Is it possible to get this?

# ? Apr 5, 2010 19:42

haveblue: Aug 15, 2005; Toilet Rascal

Initialize a second array to [0, 1, 2, 3, 4...] and perform every swap operation on it as well?

# ? Apr 5, 2010 19:43

emf: Aug 1, 2002

I'm not familiar with sort, and I don't see a man page for it, but if it is like qsort you specify a comparison function for use by the sorting function. Make the array a vector of two elements, with the second element the index. Have your comparison function only use the first element of the vector, but pass the size of the entire vector to the sorting function so that they are swapped together.

# ? Apr 5, 2010 20:18

Jose Cuervo: Aug 25, 2004

emf posted:

I'm not familiar with sort, and I don't see a man page for it, but if it is like qsort you specify a comparison function for use by the sorting function. Make the array a vector of two elements, with the second element the index. Have your comparison function only use the first element of the vector, but pass the size of the entire vector to the sorting function so that they are swapped together.

I ended up doing something very similar to this. Thank you.

# ? Apr 5, 2010 20:20

That Turkey Story: Mar 30, 2003

Avenging Dentist posted:

Put the type in parens. The preprocessor doesn't recognize < and > as grouping things and so it splits on the comma in your template. Also don't use that god-awful foreach macro (though it's admittedly better than rolling your own).

Sorry, but BOOST_FOREACH owns, not really much reason to not use it until range-based for. Even with 0x lambdas, BOOST_FOREACH is generally ahead of the game as opposed to std::for_each. Also, simply putting the type in parenthesis will not work. The way the parenthesis trick usually works with other macros is that inside the macro, the type parameter is referred to as something along the lines of function_traits< void your_type_param_name >::arg1_type. It can be made to account for [basic] cases of void by just having a metafunction that yields void when passed a function type with no args. The downsides of this are that it means you always have to pass your type in parenthesis regardless of whether or not is has commas, there are some cases it doesn't handle, at least not as intended (function types, array types, typedefed void, etc), and it internally may need to use typename cases where it's a dependent name.

If you're using a compiler that supports the 0x use of auto (GCC works fine), just use auto with BOOST_FOREACH instead of specifying the type, unless you need to be portable to compilers that are not able to handle it yet. More portably, just use a typedef as described in the docs.

# ? Apr 6, 2010 08:11

Zaxxon: Feb 14, 2004; Wir Tanzen Mekanik

ok this is likely a simple question but after a while of googling around I can't find a satisfactory example.

I am creating a series of instrument objects which have parameters that need to be accessed and changed by a uniform system. So each object needs to have a set of parameters and each parameter needs a description, but the descriptions need to be unique and constant.

So I want each instrument base class to define some constant parameter descriptions, however I get compiler errors whenever I try and declare a constant in the class interface. It looks a bit like this

code:


struct parameter_description
{
  int size;
  int min;
  int max

  parameter_description(int size, int min, int max); //this does the obvious
};

class instrument : instrument_base
{
  //blah blah blah
  protected:
  static const parameter_description PD_one(1,0x0,0xF);
  static const parameter_description PD_two(2,0x0,0xFF);

};

I suppose I could just make those descriptions static and only access them through constant references. but I'd rather just have those descriptions be constant and have each parameter just point to them.

# ? Apr 6, 2010 19:57

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

Construct the objects outside of the class.

# ? Apr 6, 2010 19:59

Zaxxon: Feb 14, 2004; Wir Tanzen Mekanik

Avenging Dentist posted:

Construct the objects outside of the class.

is that really the best practice?

I was doing that originally but I thought it made more sense to have the descriptions linked to the class. Especially since a bunch of different classes are going to have parameters with similar names like amp or frequency that may have different bounds.

# ? Apr 6, 2010 23:02

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

Zaxxon posted:

is that really the best practice?

It is not so much a "best practice" as "required by the standard". Note that you can keep them as class members.

code:

class parameter_description {
  ...
  static const parameter_description PD_one;
  ...
};

// This has to go in a unique translation unit, i.e. not in a header.
const parameter_description parameter_description::PD_one(1,0x0,0xF);

# ? Apr 6, 2010 23:17

Ciaphas: Nov 20, 2005; > BEWARE, COWARD

Got another couple of questions. I think I know the answers to both, but I want to be sure.

1) When using an ofstream, is there any way to write binary values using the stream operators, or do I have to use write() with a size? I suspect the latter, and though wrapping around that is no big deal, it seems strange. I woulda thought the ios::binary flag would change the behavior of the stream operators.

2) When templating a function, is it possible to make a particular attempted instantiation raise a compile-time error as below? (I suspect the answer is, again, "no, make it fail at runtime with an assert in a specialization or something")

code:

template <typename T>
void foo(T bar)
{
 // blah
}

int main()
{
  int a = 5;
  double b = 3;
  char[3] c = "baz";

  foo(a); // ok
  foo(b); // ok
  foo(c); // can this be made a compile-time error?
}

Thanks!

# ? Apr 7, 2010 00:07

Zaxxon: Feb 14, 2004; Wir Tanzen Mekanik

rjmccall posted:

It is not so much a "best practice" as "required by the standard". Note that you can keep them as class members.
code:
class parameter_description {
  ...
  static const parameter_description PD_one;
  ...
};

// This has to go in a unique translation unit, i.e. not in a header.
const parameter_description parameter_description::PD_one(1,0x0,0xF);

ah! Thanks, thats what he meant! I feel dumb because I was trying to do something like this at first

code:

class Instrument{
  ...
  static const parameter_description PD_one;
  ...
};

// This has to go in a unique translation unit, i.e. not in a header.
const Instrument::PD_one(1,0x0,0xF);

and was getting errors, so I thought I was doing something not allowed and moved them into the header. Then what I thought avenging dentist meant is that I had to declare them as outside the class definition.

# ? Apr 7, 2010 00:19

Vanadium: Jan 8, 2005

Ledneh posted:

Got another couple of questions. I think I know the answers to both, but I want to be sure.

1) When using an ofstream, is there any way to write binary values using the stream operators, or do I have to use write() with a size? I suspect the latter, and though wrapping around that is no big deal, it seems strange. I woulda thought the ios::binary flag would change the behavior of the stream operators.

iostreams are for formatted text output. The binary flag refers to opening a file in binary mode, that is not converting lineendings or whatever funny mechanics your OS imposes on text files.

quote:

2) When templating a function, is it possible to make a particular attempted instantiation raise a compile-time error as below? (I suspect the answer is, again, "no, make it fail at runtime with an assert in a specialization or something")

This works for me:

code:

#include <boost/type_traits.hpp>
#include <boost/utility.hpp>

template <typename T>
typename boost::disable_if<typename boost::is_same<T, char*>::type>::type
foo(T bar)
{
  // blah
}

int main() {
  int a = 5;
  double b = 3;
  char c[] = "baz";

  foo(a); // ok
  foo(b); // ok
  foo(c); // can this be made a compile-time error?
}

# ? Apr 7, 2010 00:23

Ciaphas: Nov 20, 2005; > BEWARE, COWARD

Vanadium posted:

iostreams are for formatted text output. The binary flag refers to opening a file in binary mode, that is not converting lineendings or whatever funny mechanics your OS imposes on text files.

Oh, okay. Should I be using something other than fstream for binary file writing/reading then, or is .write() and .read the best I'm going to get?

And thanks for the template trick

# ? Apr 7, 2010 00:29

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

Vanadium posted:

This works for me:

code:

#include <boost/type_traits.hpp>
#include <boost/utility.hpp>

template <typename T>
typename boost::disable_if<typename boost::is_same<T, char*>::type>::type
foo(T bar)
{
  // blah
}

int main() {
  int a = 5;
  double b = 3;
  char c[] = "baz";

  foo(a); // ok
  foo(b); // ok
  foo(c); // can this be made a compile-time error?
}

http://codepad.org/HKgcutnR

# ? Apr 7, 2010 00:39

Ciaphas: Nov 20, 2005; > BEWARE, COWARD

Avenging Dentist posted:

http://codepad.org/HKgcutnR

I'm so terrible at this profession. :doh:

# ? Apr 7, 2010 00:44

floWenoL: Oct 23, 2002

Avenging Dentist posted:

http://codepad.org/HKgcutnR

That's a link-time error. :colbert:

# ? Apr 7, 2010 00:58

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

floWenoL posted:

That's a link-time error.

I like linking the errors are more fun.

# ? Apr 7, 2010 01:10

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

code:

class Forbidden { ~Forbidden(); };

template <class T> void foo(T);
Forbidden foo(char *);

void test() {
  char a[] = "help";
  foo(a);
}

The amusing thing about this is that I checked it with clang and found a bug.

# ? Apr 7, 2010 01:11

Flobbster: Feb 17, 2005; "Cadet Kirk, after the way you cheated on the Kobayashi Maru test I oughta punch you in tha face!"

The correct answer is to use C++0x to add your own meaningful, smug diagnostic:

code:

#include <type_traits>

template <typename T>
void foo(T value)
{
    static_assert(!std::is_same<T, char*>::value,
        "don't call this with a char* you rear end in a top hat");

    // Do something
}

int main()
{
    int a = 5;
    double b = 3;
    char c[] = "baz";

    foo(a); // ok
    foo(b); // ok
    foo(c); // can this be made a compile-time error?
      
    return 0;
}

code:

$ g++-4.5 -std=c++0x test.cpp
test.cpp: In function �void foo(T) [with T = char*]�:
test.cpp:20:10:   instantiated from here
test.cpp:6:5: error: static assertion failed: "don\'t call this with a char* you rear end in a top hat"

EDIT: or this!

code:

#include <type_traits>

template <typename T>
void foo(T value)
{
    // Do something
}

void foo(char*) = delete;

Flobbster fucked around with this message at 03:47 on Apr 7, 2010

# ? Apr 7, 2010 01:53

Contero: Mar 28, 2004

This isn't a question, but I think this is probably the best place for it.

I've gotten really sick of YACC, ANTLR and Spirit so I decided to roll my own parser generator that's contained entirely in C++ like Spirit is. The idea is that at every assignment for a rule all of the parsed results get handed off to your function where you can process them and eventually return any type you want. In this case it's an int. I think the end result is simple, clean, easy to set up and gives you a lot of freedom to construct whatever you want.

code:

enum { LP, RP, PLUS, MULT, NUM };

const char *program = "(4 + 6) * 7 + 3";

int doAdd(std::vector<RuleResult> results)
{
   int ret = results[0].getVal<int>();

   for (unsigned i = 2; i < results.size(); i += 2)
      ret += results[i].getVal<int>();

   return ret;
}

...

int main ()
{
   Lexer lexer;

   lexer.type(LP, "\\(");
   lexer.type(RP, "\\)");
   lexer.type(PLUS, "\\+");
   lexer.type(MULT, "\\*");
   lexer.type(NUM, "[0-9]+");
   lexer.ignore("\\s");

   Rule<int> addExpr(doAdd), multExpr(doMult), valExpr(doVal);

   addExpr = multExpr + *( TR(PLUS) + multExpr );

   multExpr = valExpr + *( TR(MULT) + valExpr );

   valExpr = TR(LP) + addExpr + TR(RP)
           | TR(NUM);

   addExpr.check();
   multExpr.check();
   valExpr.check();

   int result = addExpr.start(lexer.lex(program));

   cout << "result: " << result << endl;

   return 0;
}

The lexer uses boost::regex. I made the lexer and parser separate so you can easily use just the lexer if you want or modify the token stream returned by lex() to do interesting things like removing nested block comments: /* /*... */ */.

It's a little rough around the edges of course. I've just implemented throw away (!) and optional (~) rules, but haven't tested them out yet.

# ? Apr 7, 2010 21:30

Ciaphas: Nov 20, 2005; > BEWARE, COWARD

I'm (not) proud of this template exclusion solution. http://codepad.org/W2H3T5Ea

# ? Apr 7, 2010 22:02

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

You can reduce that entire second part to is_pointer. Also you forgot about volatile variables. http://codepad.org/MvuXbKPK

# ? Apr 7, 2010 22:16

That Turkey Story: Mar 30, 2003

Avenging Dentist posted:

You can reduce that entire second part to is_pointer. Also you forgot about volatile variables. http://codepad.org/MvuXbKPK

You're all over-complicating things :smug:

http://codepad.org/lg4Cp1ab

# ? Apr 7, 2010 22:33

Ciaphas: Nov 20, 2005; > BEWARE, COWARD

Well this post series was certainly a learning experience. Thanks for the education and correction

Ciaphas fucked around with this message at 00:33 on Apr 8, 2010

# ? Apr 8, 2010 00:29

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

I wish type_traits had a base type trait to strip out all arrays refs, pointers, and cv-qualifiers.

# ? Apr 8, 2010 01:39

Ciaphas: Nov 20, 2005; > BEWARE, COWARD

That's what I was looking for at first when I started reading about the enable/disable_if constructs, yeah.

Actually, on that last codepad, is the is_pointer clause even necessary? It looks like the star in the second is_convertible parameter covers that, making the is_pointer redundant, unless I'm missing something important here (read: I'm missing something).

# ? Apr 8, 2010 03:12

That Turkey Story: Mar 30, 2003

Ledneh posted:

That's what I was looking for at first when I started reading about the enable/disable_if constructs, yeah.

Actually, on that last codepad, is the is_pointer clause even necessary? It looks like the star in the second is_convertible parameter covers that, making the is_pointer redundant, unless I'm missing something important here (read: I'm missing something).

You are missing the fact that types can have an overloaded conversion operator to a [qualified] char pointer. Without the is_pointer check, it would incorrectly disallow people from using the template with such a type.

# ? Apr 8, 2010 03:22

Avenging Dentist: Oct 1, 2005; oh my god is that a circular saw that does not go in my mouth aaaaagh

TTS explained it but here's the code: http://codepad.org/sMZ1yblu

# ? Apr 8, 2010 03:25

Ciaphas: Nov 20, 2005; > BEWARE, COWARD

Slowly I continue to learn things. Thanks you two

# ? Apr 8, 2010 03:30

CaptainMcArrr: Oct 4, 2003; PACWANKER HAS HORSE COCK

Hi guys, I hope you can help with what (I think) is a relatively stupid C question but for some reason I can't quite wrap my head around it. The classes I've done have sort of glossed over headers so they're still a bit confusing.

Basically I've got a program with multiple (3) c files. I've got a structure that's used across all of them. Where do I define it and how do I import it?

I tried creating a types.h file and putting it in there, but then the other header files threw a hissy fit because, I guess, they couldn't see it?

I hope I'm explaining myself okay - can anyone help?

# ? Apr 9, 2010 07:41

Adbot: ADBOT LOVES YOU

# ? Jun 1, 2024 23:25

UraniumAnchor: May 21, 2006; Not a walrus.

That's pretty much exactly what you want to do (put the structure declaration in a .h file), so you might have to be more specific about what is going wrong.

If another header refers to your struct you need to include the struct header beforehand.

Edit:

code:

foo.h
-----

struct foo {
 int x;
};

bar.h
-----

#include "foo.h"

struct bar {
 struct foo f;
};

Or, optionally, in any file that includes bar.h you need to include foo.h in that file before, such as:

code:

main.c
-----

#include "foo.h"
#include "bar.h"

Hopefully that makes it more clear.

UraniumAnchor fucked around with this message at 07:52 on Apr 9, 2010

# ? Apr 9, 2010 07:50

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Cavern of COBOL > C/C++ Programming Questions Not Worth Their Own Thread

«‹›640 »