A question of indentation!
-
Eddie Velasquez wrote: An exception should always indicate an error... That's what I also thought. But PGP errors are often not errors. I have explained in one of the mis-placed posts here. I don't wanna re-type it :-) The indendation of this thread is now worse than any indentation I could have produced :-) Nish
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
Nish - Native CPian wrote: That's what I also thought. But PGP errors are often not errors. If a PGP error isn't a "real" error (informational return value) then it shouldn't raise an exception or you can catch the exception and just ignore it.
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class Implementation -
Christian Graus wrote: An exception does not necessarily mean an error. An exception should always indicate an error... that's why they're called "exceptions"... because they're exceptional. :) Or did I miss something in your post?
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class ImplementationTo quote Stroustrup: The Exception handling mechanism is a nonlocal control structure based on stack unwinding that can be seen as an alternative return mechanism. There are therefore legitimate uses of exceptions that have nothing to do with errors. Christian I am completely intolerant of stupidity. Stupidity is, of course, anything that doesn't conform to my way of thinking. - Jamie Hale - 29/05/2002
-
Tomasz Sowinski wrote: if (call1() && call2() && call3() ....) That won;'t work. I am not supposed to call call-n unless call-n-1 returns true!
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
Nish - Native CPian wrote: if (call1() && call2() && call3() ....) That won;'t work. I am not supposed to call call-n unless call-n-1 returns true! :confused: Why won't that work then? With C++, you have "short-circuit" evaluation - that is, if the first item in an If statement fails, the rest won't even get evaluated. For instance, I have code like this all the time (well, when I'm not using CStrings... :-D) : if(pStr != NULL && strlen(pStr) > 0) { ... } If pStr is NULL, the first condition fails, and the second condition (strlen) won't even be evaluated. No generalization is 100% true. Not even this one.
-
Tomasz Sowinski wrote: Well, it seems that you need some kind of object wrapper over PGPError. The object could convert that to true/false or even throw some exception Shucks! It always comes down to the basics and theory. If my OOP theory was strong enough I'd not have run into this situation :-( Nish
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
Nish - Native CPian wrote: Shucks! It always comes down to the basics and theory. If my OOP theory was strong enough I'd not have run into this situation Nish, don't get discouraged. The prime motivation for most of us to learn OOP was because of these things. Don't shy away from OOP. In a few months you will be the one telling others to wrap their functions into objects. C++ Programmers do it in class Drinking In The Sun Forgot Password?
-
Indentation is nice. In fact code that is not indented is an absolute pain to even look at. But then sometimes you get into absurd levels of indentation. Right now I am working with the PGP SDK. For certain operations I need to call about 7-10 functions sequentially. The problem is that each of these functions can be called ONLY if all the previous functions are successful. Thus I have something like this.
if(call1())
{
if(call2())
{
if(call3())
{
if(call4())
{
if((call5())
{
if(call6())
{That's just a sample, just the tip of the large iceberg. Often it get's a LOT more deeply nested than I have shown above! In such situations can we actually do away with indentation at least partially? For example would it be considered okay to do this.
if(call1())
{
if(call2())
{
if(call3())
{
if(call4())
{
if((call5())
{
if(call6())
{I have maintained a little indentation, but it's not perfectly done! Your comments are welcome
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
Excuse me Nish, I can't resist! Post it in the right forum ! :) :) Crivo Automated Credit Assessment
-
Eddie Velasquez wrote: if(!call3()) return; I can't do that. Since I need to do some cleaning up as well :( Nish
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
do {
if (!call1())
break;
if (!call2())
break;
...
} while (false);Cleanup();
cheers, Chris Maunder
VC++ - the language that doesn't say 'no'
-
Eddie Velasquez wrote: Well, I was talking in a general sense and didn't realize you were referring to the PGP SDK in particular. My bad. PGP calls return a PGPError and you need to call IsPGPError(...) on it to figure out whether it was an error or not. Some cases you actually get an error and it's not really an error overall! Like some encryption algorithms won't work when it's an RSA key but you are supposed to ignore that since it won't cause any problem Nish
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
Nish - Native CPian wrote: PGP calls return a PGPError and you need to call IsPGPError(...) on it to figure out whether it was an error or not. That is the worst API I've ever heard of. Jason Gerard
-
To quote Stroustrup: The Exception handling mechanism is a nonlocal control structure based on stack unwinding that can be seen as an alternative return mechanism. There are therefore legitimate uses of exceptions that have nothing to do with errors. Christian I am completely intolerant of stupidity. Stupidity is, of course, anything that doesn't conform to my way of thinking. - Jamie Hale - 29/05/2002
I never thought of it this way but try/catch would be a good "control structure" to use instead of nested if's. Its also 100x easier to read. Is Stroustrup trying to say it should only be used as a return mechanism if it isn't used in error handling? Like it or not, I'm right.
-
Christian Graus wrote: Try/Thow/catch ( and finally in C#, dunno about C++ ) are the obvious way to deal with your situation. I don't even know how to implement exceptions :-( Might have to do some reading Actually after reading all the suggestions, I liked peterchens do{} while(0) idea :-) Nish
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
// Bad example, but it shows you how to use them
for (int count=0; count<10; count++)
{
try
{
if (count%2 == 0)
throw "Even";
else
throw "Odd";
}
catch(LPSTR str)
{
MessageBox(NULL, str, "This number is:", MB_OK);
}
catch(...) // all other errors
{
MessageBox(NULL, "Unhandled exception.", "Error", MB_OK);
}
}Like it or not, I'm right.
-
To quote Stroustrup: The Exception handling mechanism is a nonlocal control structure based on stack unwinding that can be seen as an alternative return mechanism. There are therefore legitimate uses of exceptions that have nothing to do with errors. Christian I am completely intolerant of stupidity. Stupidity is, of course, anything that doesn't conform to my way of thinking. - Jamie Hale - 29/05/2002
Christian Graus wrote: There are therefore legitimate uses of exceptions that have nothing to do with errors. Well, I can't pretend to teach C++ to Bjarne! :-O But as quoted, these "legitimate uses that have nothing to do with errors" are the exception, not the rule. (pun intended!) I'm not really thinking about this too hard, but I fail to come up with a case where I would raise an exception that doesn't somehow indicate an error condition. :confused:
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class Implementation -
// Bad example, but it shows you how to use them
for (int count=0; count<10; count++)
{
try
{
if (count%2 == 0)
throw "Even";
else
throw "Odd";
}
catch(LPSTR str)
{
MessageBox(NULL, str, "This number is:", MB_OK);
}
catch(...) // all other errors
{
MessageBox(NULL, "Unhandled exception.", "Error", MB_OK);
}
}Like it or not, I'm right.
I want to add that it's a bad idea to throw exceptions for anything that doesn't indicate an "exceptional" (uncommon, erroneous) condition because in C++ exception support has a lot of overhead that's too costly for idioms that can be expressed in a cleaner and more efficient way using othe language constructs. Jason Henderson wrote: catch(...) // all other errors { MessageBox(NULL, "Unhandled exception.", "Error", MB_OK); } Just nitpicking... the message should be "unexpected exception" not "unhandled exception" because you just handled the exception!
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class Implementation -
I never thought of it this way but try/catch would be a good "control structure" to use instead of nested if's. Its also 100x easier to read. Is Stroustrup trying to say it should only be used as a return mechanism if it isn't used in error handling? Like it or not, I'm right.
Jason Henderson wrote: I never thought of it this way but try/catch would be a good "control structure" to use instead of nested if's. Its also 100x easier to read. The problem is that for most implementations, exception support has too much overhead for practical use as a "control structure" that's not directly related to exceptional (uncommon, errroneous) conditions. Check out Vishal Kochhar's excellent How a C++ compiler implements exception handling article for more information.
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class Implementation -
Eddie Velasquez wrote: I fail to come up with a case where I would raise an exception that doesn't somehow indicate an error condition Assume that you have a dozen of functions searching for some data in smart card, registry, current directory, directories listed in some env variable, etc, etc. So instead of nested constructs like this...
if (!SearchForFooInGoo(...))
{
if (!SearchForFooInBoo(...))
{
}
}... you just throw FooFoundException when one of the calls returns true. Then your exception handler does the job. Perverse, but still possible :-D Tomasz Sowinski -- http://www.shooltz.com
- It's for protection
- Protection from what? Zee Germans?Yeah I thought about this but I think it's a bad idea. Check out my reasons why this use of exceptions is a bad idea post
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class Implementation -
Christian Graus wrote: There are therefore legitimate uses of exceptions that have nothing to do with errors. Well, I can't pretend to teach C++ to Bjarne! :-O But as quoted, these "legitimate uses that have nothing to do with errors" are the exception, not the rule. (pun intended!) I'm not really thinking about this too hard, but I fail to come up with a case where I would raise an exception that doesn't somehow indicate an error condition. :confused:
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class ImplementationEddie Velasquez wrote: I fail to come up with a case where I would raise an exception that doesn't somehow indicate an error condition Assume that you have a dozen of functions searching for some data in smart card, registry, current directory, directories listed in some env variable, etc, etc. So instead of nested constructs like this...
if (!SearchForFooInGoo(...))
{
if (!SearchForFooInBoo(...))
{
}
}... you just throw FooFoundException when one of the calls returns true. Then your exception handler does the job. Perverse, but still possible :-D Tomasz Sowinski -- http://www.shooltz.com
- It's for protection
- Protection from what? Zee Germans? -
Indentation is nice. In fact code that is not indented is an absolute pain to even look at. But then sometimes you get into absurd levels of indentation. Right now I am working with the PGP SDK. For certain operations I need to call about 7-10 functions sequentially. The problem is that each of these functions can be called ONLY if all the previous functions are successful. Thus I have something like this.
if(call1())
{
if(call2())
{
if(call3())
{
if(call4())
{
if((call5())
{
if(call6())
{That's just a sample, just the tip of the large iceberg. Often it get's a LOT more deeply nested than I have shown above! In such situations can we actually do away with indentation at least partially? For example would it be considered okay to do this.
if(call1())
{
if(call2())
{
if(call3())
{
if(call4())
{
if((call5())
{
if(call6())
{I have maintained a little indentation, but it's not perfectly done! Your comments are welcome
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
Nish - Native CPian wrote: In fact code that is not indented is an absolute pain to even look at. But then sometimes you get into absurd levels of indentation Try generating HTML tables with perl's CGI.pm. (Yes I'm using the MDE to edit perl files... this was before I learned vi) - Jason (SonorkID 100.611) In the beginning, teachers taught the 5 W's: who, what, where, when, why. Now it's just a big damn G
-
Yeah I thought about this but I think it's a bad idea. Check out my reasons why this use of exceptions is a bad idea post
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class ImplementationI don't buy performance-related arguments, at least not in 100%. While there are the situations in which this could matter (thight loops etc), in my example you'd access disk. For sure this operation would be orders of magnitude slower than throwing/catching exceptions. The same applies to displaying anything on the screen or accepting user input. In short, the cost associated with exception handling would be totally invisible. Tomasz Sowinski -- http://www.shooltz.com
- It's for protection
- Protection from what? Zee Germans? -
Nish - Native CPian wrote: PGP calls return a PGPError and you need to call IsPGPError(...) on it to figure out whether it was an error or not. That is the worst API I've ever heard of. Jason Gerard
Jason Gerard wrote: That is the worst API I've ever heard of Try Windows ;) ERROR_MORE_DATA Well, actually you have the same kind of return values in almost any API - commonly you find this things in "evolved" API's where conditions changed but not the environment. Thus, the function returns a error which must be interpreted to find out that its actually just a information. Sometimes its just a misconception as for ERROR_MORE_DATA. Vote against software patents in europe
-
do {
if (!call1())
break;
if (!call2())
break;
...
} while (false);Cleanup();
cheers, Chris Maunder
VC++ - the language that doesn't say 'no'
What about this? :)
switch (1)
{
case 1: if (!call1()) break;
case 2: if (!call2()) break;
case 3: if (!call3()) break;
...
}Tomasz Sowinski -- http://www.shooltz.com
- It's for protection
- Protection from what? Zee Germans? -
(Using my best Yoda voice...) Ummm. Strong in the force you are... but OOP you must.;)
Eddie Velasquez: A Squeezed Devil
Checkout General Guidelines for C# Class ImplementationEddie Velasquez wrote: but OOP you must. Yeah. I am gonna OOP more from now on ... Nish :-)
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
-
do {
if (!call1())
break;
if (!call2())
break;
...
} while (false);Cleanup();
cheers, Chris Maunder
VC++ - the language that doesn't say 'no'
Thanks Chris M. PeterChen had already suggested this to me. Though he used a 0 instead of a false as you did :-) By the way your do while brace-poitioning is ugly Dont use do { like a java guy. Put it as do { like a C++ guy Nish
Regards, Nish Native CPian. Born and brought up on CP. With the CP blood in him.
