expression evaluation confusion in c++ - expression

Can somebody please explain to me what is going on here?
I have this line of code:
if ( pt->child == NULL && pt->visits < cutoff+1 || depth > 5 )
and I'm getting a g++ compiler warning:
warning: suggest parentheses around ‘&&’ within ‘||’ [-Wparentheses]|
What is it trying to warn me of? And worse, if I put in the "obvious" parens, like so:
if ( ( pt->child == NULL && pt->visits < cutoff+1 ) || depth > 5 )
I get a different behavior -- indicating I really did do something wrong in the first expression. Arithmetic operators have higher precedence over comparisons which have higher precedence over boolean operators where && has higher precedence than ||, right?
What am I missing?

Just like the numeric operators + and *, the logical operators have different precedence. So the order of evaluation of a logical statement will not necessarily be strictly left to right, but will instead evaluate && first and || 2nd. This has been the source of so much confusion and error that the compiler will emit this warning unless you explicitly surround the elements with parentheses.
What going on here? I think in this case it's erroneous because if you make g++ dump it's internal structures
this
if ( pt->child == NULL && pt->visits < cutoff+1 || depth > 5 )
printf("Hello World");
equates to this
D.3367 = pt->child;
if (D.3367 == 0B) goto <D.3368>; else goto <D.3364>;
<D.3368>:
D.3369 = pt->visits;
D.3370 = cutoff + 1;
if (D.3369 < D.3370) goto <D.3365>; else goto <D.3364>;
<D.3364>:
if (depth > 5) goto <D.3365>; else goto <D.3366>;
<D.3365>:
printf ("Hello World");
<D.3366>:
Which does exactly what you would expect.

Related

Coding styles in conditional expression of some programming languages

It's a bit confusing to me about what is the difference between these condition expressions below:
if( 1 == a) {
//something
}
and
if( a == 1 ) {
//something
}
I saw the above one in some scripts I have downloaded and I wonder what's the difference between them.
The former has been coined a Yoda Condition.
Using if(constant == variable) instead of if(variable == constant), like if(1 == a). Because it's like saying "if blue is the sky" or "if tall is the man".
The constant == variable syntax is often used to avoid mistyping == as =. It is, of course, often used without understanding also when you have constant == function_call_retuning_nothing_modifiable.
Other than that there's no difference, unless you have some weird operator override.
Many programming languages allow assignments like a = 1 to be used as expressions, making the following code syntactically valid (given that integers can be used in conditionals, such as in C or many scripting languages):
if (a = 1) {
// something
}
This is rarely desired, and can lead to unexpected behavior. If 1 == a is used, then this mistake cannot occur because 1 = a is not valid.
Well, I am not sure about the trick. Generally, we could say the equal sign is commutative. So, a = b implies b = a. However, when you have == or === this doesn't work in certain cases, for example when on the right side you have a range: 5 === (1..10) vs. (1..10) === 5.

What is the bash "test" command evaluation order?

What is the order in which the test or [ expressions are evaluated in a bash script? Is it the same as the order in which they are written? Are they all evaluated?
In the following example:
if [ expr1 -a -n expr2 ]
will expr2 be evaluated if expr1 is false?
It's been a long time since I used that particular syntax. Nowadays, I prefer:
if [[ expr1 && -n expr2 ]]
since I know that these are short-circuited. The bash man page specifically states:
The && and || operators do not evaluate expression2 if the value of expression1 is sufficient to determine the return value of the entire conditional expression.
It doesn't appear to state clearly one way or the other for the [ and test variant (using -a).
However, in the interests of furthering your knowledge and mine (and an indication as to why open source is a good thing), I went trolling through the bash source code.
It turns out that it's not short-circuited in that case. The particular pieces of code are (compressed a bit for readability):
static int or() {
int value, v2;
value = and();
if (pos < argc && argv[pos][0] == '-'
&& argv[pos][1] == 'o' && !argv[pos][2])
{
advance (0);
v2 = or();
return (value || v2);
}
return (value);
}
static int and() {
int value, v2;
value = term();
if (pos < argc && argv[pos][0] == '-'
&& argv[pos][1] == 'a' && !argv[pos][2])
{
advance (0);
v2 = and();
return (value && v2);
}
return (value);
}
You can see in both those cases that it continues to interpret the expressions regardless of the state of the first expression.
That was from bash 2.5 but I've just checked the 4.1 source as well and it hasn't changed. Source was pulled straight from the horse's mouth.
So. bottom line: it appears that all sub-expressions are evaluated when using -a and -o with test or [.

Should I test if equal to 1 or not equal to 0?

I was coding here the other day, writing a couple of if statements with integers that are always either 0 or 1 (practically acting as bools). I asked myself:
When testing for positive result, which is better; testing for int == 1 or int != 0?
For example, given an int n, if I want to test if it's true, should I use n == 1 or n != 0?
Is there any difference at all in regards to speed, processing power, etc?
Please ignore the fact that the int may being more/less than 1/0, it is irrelevant and does not occur.
Human's brain better process statements that don't contain negations, which makes "int == 1" better way.
It really depends. If you're using a language that supports booleans, you should use the boolean, not an integer, ie:
if (value == false)
or
if (value == true)
That being said, with real boolean types, it's perfectly valid (and typically nicer) to just write:
if (!value)
or
if (value)
There is really very little reason in most modern languages to ever use an integer for a boolean operation.
That being said, if you're using a language which does not support booleans directly, the best option here really depends on how you're defining true and false. Often, false is 0, and true is anything other than 0. In that situation, using if (i == 0) (for false check) and if (i != 0) for true checking.
If you're guaranteed that 0 and 1 are the only two values, I'd probably use if (i == 1) since a negation is more complex, and more likely to lead to maintenance bugs.
If you're working with values that can only be 1 or 0, then I suggest you use boolean values to begin with and then just do if (bool) or if (!bool).
In language where int that are not 0 represents the boolean value 'true', and 0 'false', like C, I will tend to use if (int != 0) because it represents the same meaning as if (int) whereas int == 1 represents more the integer value being equal to 1 rather than the boolean true. It may be just me though. In languages that support the boolean type, always use it rather than ints.
A Daft question really. If you're testing for 1, test for 1, if you're testing for zero, test for zero.
The addition of an else statement can make the choice can seem arbitrary. I'd choose which makes the most sense, or has more contextual significance, default or 'natural' behaviour suggested by expected frequency of occurrence for example.
This choice between int == 0 and int != 1 may very well boil down to subjective evaluations which probably aren't worth worrying about.
Two points:
1) As noted above, being more explicit is a win. If you add something to an empty list you not only want its size to be not zero, but you also want it to be explicitly 1.
2) You may want to do
(1 == int)
That way if you forget an = you'll end up with a compile error rather than a debugging session.
To be honest if the value of int is just 1 or 0 you could even say:
if (int)
and that would be the same as saying
if (int != 0)
but you probably would want to use
if (int == 1)
because not zero would potentially let the answer be something other than 1 even though you said not to worry about it.
If only two values are possible, then I would use the first:
if(int == 1)
because it is more explicit. If there were no constraint on the values, I would think otherwise.
IF INT IS 1
NEXT SENTENCE
ELSE MOVE "INT IS NOT ONE" TO MESSAGE.
As others have said, using == is frequently easier to read than using !=.
That said, most processors have a specific compare-to-zero operation. It depends on the specific compiler, processor, et cetera, but there may be an almost immeasurably small speed benefit to using != 0 over == 1 as a result.
Most languages will let you use if (int) and if (!int), though, which is both more readable and get you that minuscule speed bonus.
I'm paranoid. If a value is either 0 or 1 then it might be 2. May be not today, may be not tomorrow, but some maintenance programmer is going to do something weird in a subclass. Sometimes I make mistakes myself [shh, don't tell my employer]. So, make the code say tell me that the value is either 0 or 1, otherwise it cries to mummy.
if (i == 0) {
... 0 stuff ...
} else if (i == 1) {
... 1 stuff ...
} else {
throw new Error();
}
(You might prefer switch - I find its syntax in curly brace language too heavy.)
When using integers as booleans, I prefer to interpret them as follows: false = 0, true = non-zero.
I would write the condition statements as int == 0 and int != 0.
I would say it depends on the semantics, if you condition means
while ( ! abort ) negation is ok.
if ( quit ) break; would be also ok.
if( is_numeric( $int ) ) { its a number }
elseif( !$int ) { $int is not set or false }
else { its set but its not a number }
end of discussion :P
I agree with what most people have said in this post. It's much more efficient to use boolean values if you have one of two distinct possibilities. It also makes the code a lot easier to read and interpret.
if(bool) { ... }
I was from the c world. At first I don't understand much about objective-c. After some while, I prefer something like:
if (int == YES)
or
if (int == NO)
in c, i.e.:
if (int == true)
if (int == false)
these days, I use varchar instead of integer as table keys too, e.g.
name marital_status
------ --------------
john single
joe married
is a lot better than:
name marital_status
------ --------------
john S
joe M
or
name marital_status
------ --------------
john 1
joe 2
(Assuming your ints can only be 1 or 0) The two statements are logically equivalent. I'd recommend using the == syntax though because I think it's clearer to most people when you don't introduce unnecessary negations.

How to create conditional breakpoint with std::string

Suppose I have this function:
std::string Func1(std::string myString)
{
//do some string processing
std::string newString = Func2(myString)
return newString;
}
How do I set a conditional break when newString has a specific value? (without changing the source)
Setting the condition newString == "my value" didn't work. The breakpoints were disabled with an error overloaded operator not found.
There is a much easier way in Visual Studio 2010/2012.
To accomplish what you are looking for in ANSI use this:
strcmp(newString._Bx._Ptr,"my value")==0
And in unicode (if newString were unicode) use this:
wcscmp(newString._Bx._Ptr, L"my value")==0
There are more things you can do than just a compare, you can read more about it here:
http://blogs.msdn.com/b/habibh/archive/2009/07/07/new-visual-studio-debugger-2010-feature-for-c-c-developers-using-string-functions-in-conditional-breakpoints.aspx
In VS2017, I was able to set the condition as:
strcmp(&newString[0], "my value") == 0
Some searching has failed to turn up any way to do this. Suggested alternatives are to put the test in your code and add a standard breakpoint:
if (myStr == "xyz")
{
// Set breakpoint here
}
Or to build up your test from individual character comparisons. Even looking at individual characters in the string is a bit dicey; in Visual Studio 2005 I had to dig down into the member variables like
myStr._Bx._Buf[0] == 'x' && myStr._Bx._Buf[1] == 'y' && myStr._Bx._Buf[2] == 'z'
Neither of these approaches is very satisfactory. We should have better access to a ubiquitous feature of the Standard Library.
In VS2017 you can do
strcmp(newString._Mypair._Myval2._Bx._Buf,"myvalue")==0
While I've had to work around this using something similar to Brad's answer (plus using DebugBreak() to break right from the code), sometimes editing/recompiling/re-running a bit of code is either too time consuming or just plain impossible.
Luckily, it's apparently possible to spelunk into the actual members of the std::string class. One way is mentioned here -- and though he calls out VS2010 specifically, you can still access individual chars manually in earlier versions. So if you're using 2010, you can just use the nice strcmp() functions and the like (more info), but if you're like me and still have 2008 or earlier, you can come up with a raggedy, terrible, but functional alternative by setting a breakpoint conditional something like:
strVar._Bx._Ptr[0] == 'a' && strVar._Bx._Ptr[1] == 'b' &&
strVar._Bx._Ptr[2] == 'c'
to break if the first three characters in strVar are "abc". You can keep going with additional chars, of course. Ugly.. but it's saved me a little time just now.
VS2012:
I just used the condition below because newString._Bx._Ptr ( as in OBWANDO's answer ) referenced illegal memory
strcmp( newString._Bx._Buf, "my value")==0
and it worked...
#OBWANDO (almost) has the solution, but as multiple comments rightly point out, the actual buffer depends on the string size; I see 16 to be the threshold. Prepending a size check to the strcmp on the appropriate buffer works.
newString._Mysize < 16 && strcmp(newString._Bx._Buf, "test value") == 0
or
newString._Mysize >= 16 && strcmp(newString._Bx._Ptr, "ultra super long test value") == 0
Tried to use strcmp in gdb8.1 under ubuntu18.04, but it doesn't work:
(ins)(gdb) p strcmp("a", "b")
$20 = (int (*)(const char *, const char *)) 0x7ffff5179d60 <__strcmp_ssse3>
According to this answer, strcmp, is a special IFUNC, one can setup condition like this:
condition 1 __strcmp_ssse3(camera->_name.c_str(), "ping")==0
It's pretty ugly, don't want to do it the second time.
This answer gives a much better solution, it use std::string::compare :
condition 1 camera->_name.compare("ping") == 0
In VS2015 you can do
newstring[0]=='x' && newString[1]=='y' && newString[2]=='z'
Comparing string works better than comparing characters
strcmp(name._Mypair._Myval2._Bx._Buf, "foo")==0
This works, but is very inconvenient to use and error prone.
name._Mypair._Myval2._Bx._Buf[0] == 'f' &&
name._Mypair._Myval2._Bx._Buf[1] == '0' &&
name._Mypair._Myval2._Bx._Buf[2] == '0'
You could convert it into a c string using c_str() like so:
$_streq(myStr.c_str(), "foo")
To set a conditional breakpoint in std::string you need to set it on real internal members of std::string. What you see on watch window is simplified.
You can display real structure of a variable in the watch window by using ,! suffix. In your example:
newString,!
For MSVC 2015 – 2019 you can use:
For string that were never longer than 15 characters:
(newString._Mypair._Myval2._Myres < 16) ?
strcmp(newString._Mypair._Myval2._Bx._Buf, "short") == 0 :
false
For (even historically) longer strings:
(newString._Mypair._Myval2._Myres < 16) ? false :
strcmp(newString._Mypair._Myval2._Bx._Ptr, "My_test_str_value_longer_than_16_chars") == 0
Beware:
The variable name is written twice in each condition!
You need whole expression on single line. Use the copy-paste versions bellow.
Universal condition needs to put the test value twice and variable name three times:
(newString._Mypair._Myval2._Myres < 16) ?
strcmp(newString._Mypair._Myval2._Bx._Buf, "My_test_string") == 0 :
strcmp(newString._Mypair._Myval2._Bx._Ptr, "My_test_string") == 0
Notes: use wcscmp instead of strcmp if you are working with std::wstring.
Find more info on small string optimization in C++ https://vorbrodt.blog/2019/03/30/sso-of-stdstring/ includes sample code to find size of string's internal buffer.
All std:string and std::wstring single line versions for your copy paste convenience:
(newString._Mypair._Myval2._Myres < 16) ? strcmp(newString._Mypair._Myval2._Bx._Buf, "short") == 0 : false
(newString._Mypair._Myval2._Myres < 16) ? false : strcmp(newString._Mypair._Myval2._Bx._Ptr, "My_test_str_value_longer_than_16_chars") == 0
(newString._Mypair._Myval2._Myres < 16) ? strcmp(newString._Mypair._Myval2._Bx._Buf, "My_test_string") == 0 : strcmp(newString._Mypair._Myval2._Bx._Ptr, "My_test_string") == 0
(newString._Mypair._Myval2._Myres < 16) ? wcscmp(newString._Mypair._Myval2._Bx._Buf, L"short") == 0 : false
(newString._Mypair._Myval2._Myres < 16) ? false : wcscmp(newString._Mypair._Myval2._Bx._Ptr, L"My_test_str_value_longer_than_16_chars") == 0
(newString._Mypair._Myval2._Myres < 16) ? wcscmp(newString._Mypair._Myval2._Bx._Buf, L"My_test_string") == 0 : wcscmp(newString._Mypair._Myval2._Bx._Ptr, L"My_test_string") == 0
All above copy/paste samples tested on MSVC version 16.9.10 and program for Windows 10.

IF statement formatting best-practise, what's your style?

Looking to improve my IF statement, and I want to keep my code looking pretty
This is what I am currently doing, is it readable, any room for improvement?
SomeObject o = LoadSomeObject();
if( null == o
||
null == o.ID || null == o.Title
||
0 == o.ID.Length || 0 == o.Title.Length
)
I don't have anyone to ask around. That's why I came here in first place. Please don't close my question :(
I always try and avoid complex boolean expressions for the sake of the next guy, but if I had to write an expression that didn't easily go on one line I would format it as follows:
if (value1 == value2 ||
value3 == value4 ||
value5 == value6 ||
value7 == value8) {
executeMyCode();
}
Your verbosity is leading to a less readable code, I think the following format is best:
if ( null == o || null == o.ID || null.Title || 0 == o.ID.Length || 0 == o.Title.Length )
{
// do stuff
}
We all have high resolution/widescreen displays for a reason, there's no reason to lock your code at some horribly short syntax. Also, I would simply create a function named IsIDEmpty so that the code could look like
if ( IsIDEmpty(o) )
{
// do stuff
}
to keep the code simpler & cleaner. The function would perform the actual checks and return a boolean. I'm sure this is something you might have re-use for anyways, plus it serves as a simple way for code to be more self documenting/commenting.
For the simplest formatting of what you have, I would go with one per line.
if(null == o
|| null == o.ID
|| null == o.Title
|| 0 == o.ID.Length
|| 0 == o.Title.Length)
Even better would be if you could refactor the condition such that it fits on one line. I find that a large number of || or && is usually difficult to read. Perhaps you can refactor it out into a function and be left with:
if(myFunction(...))
My rule of thumb: Avoid any format that a semi-smart auto-formatter can't reproduce.
Having a defined set of formats and a automated tool/template/configuration that actually produces code in that format is a big plus, in my opinion.
And if your code is still unreadable after it has been auto-formatted, then chances are that you need to refactor anyway.
I would either put it all on one line if it fits (which this one clearly doesn't). With this, I would put the || consistently at the start or end of line:
if( null == o ||
null == o.ID ||
null == o.Title ||
0 == o.ID.Length ||
0 == o.Title.Length
)
or
if( null == o
|| null == o.ID
|| null == o.Title
|| 0 == o.ID.Length
|| 0 == o.Title.Length
)
You could have >1 condition on a line, the positioning of || is more important I think.
I'm ignoring the fact that null.Title doesn't seem to make much sense
I find it quite distracting, to be honest. Mostly because the '||' start making funny patters.
I much rather prefer something like
if ( o == null || o.ID == null || null.Title ||
o.ID.Length == 0 || o.Title.Length )
or even better, keep it in a single line if possible.
I think it's pretty unreadable.
The affection of putting the constant first always seems a bit odd to me - most compilers can be persuaded to warn if they find an assignment in a conditional.
Then you're testing for null for two different things, then for zero length for two different things - but the important thing is not the length check, but the member you're checking. So I'd write it as
if (o == null ||
o.ID == null || o.ID.length == 0 ||
o.Title == null || o.Title.Length == 0)
Rather than making up a standard for just this one question - I would suggest adopting an existing coding standard for whatever language you are using.
For example:
GNU Coding Standards
http://www.gnu.org/prep/standards/
Code Conventions for the Java Programming Language
http://java.sun.com/docs/codeconv/
.NET Framework General Reference Design Guidelines for Class Library Developers
http://msdn.microsoft.com/en-us/library/czefa0ke.aspx
Generally I'm with TravisO on this, but if there are so many conditions in your if() statement that it just gets crazy long, consider putting in its own little function instead:
bool wereTheConditionsMet()
{
if( NULL == 0 )
return true;
if( NULL == o.ID )
return true;
: : // and so on until you exhaust all the affirmatives
return false;
}
if ( wereTheConditionsMet() )
{
// do stuff
}
It is a lot easier to convey the intent of a well-named predicate function than an endless string of ||s and &&s.
It is not readable.
This is how I do Really Long Ifs(or those I have to twiddle a lot).
if(
o == null ||
o.ID == null ||
o.Title == null ||
o.ID.Length == 0 ||
o.Title.Length == 0
)
For yours, I would do a single line.
if(o == null || o.ID == null || o.Title == null || o.ID.Length == 0 || o.Title.Length == 0)
Or, if you are using C++, I'd do something like this:
if(!o)
{}
if(! (o.ID && o.Title && o.Length))
{}
...since it separates creation from correctness.
However, caveat emptor, I've been accused of bloated LOC due to my fondness for newlines.
Use an automated code formatter, and configure it appropriately.
Write a method like isPresent(String) that checks the String argument for not null and for not empty (zero length).
Rewrite the original conditional to use the new isPresent(String) method, probably all on one line.
I usually do something like:
if(x < 0 || x >= width
|| y < 0 || y >= height)
{
/* Coordinate out of range ... */
}
The first y and x line up in a monospace font, which is nice, and I'm not confused by half-indentions.
This method works best when doing similar comparisons. Otherwise, I usually split up my if's.

Resources