Printf/scanf replacement methods: some feedback/advice needed

Question

I've recently been writing some basic command-line programs (I want to keep my skills sharp over the summer), but printf and scanf have been starting to annoy me. I'm not a wonderful C programmer, and having to get into printf/scanf and their instabilities (or even worse, fgets and their ilk) isn't exactly putting me in a comforting setting (for this reason exactly, I love NSLog, with its comforting default namespace and its automatic NSString and NSObject parsing).

Much to my disappointment, though, NSLog doesn't have a counterpart function, and prints a lot of extra 'junk' (time, function name, etc., along with a newline at the end), which defeats a lot of the purpose in my using it. So I decided to sit down for a different kind of programming exercise and write functions to replace printf and scanf that would meet my needs.

And voila, I came up with my own NSInput.h file, containing two functions: NSPrint(), and NSScan(). These two functions are modeled much after printf and scanf, but also handle NSString's. I know I'm treading on sacred namespace here, but I couldn't resist (IFPrint and IFScan just sound terrible!).

Now, while I'm really happy that I have working code (for which you can find the source here), I know that it's not efficient (much to my surprise, though, NSPrint is several times more efficient than printf under LLDB in Xcode 4, but that's beside the point). I need some advice on how to make the functions better, and more efficient. NSScan, for example, converts the va_list it recieves into an NSPointerArray, and uses NSScanner's to scan through the format and input strings, so I know there's a lot of room for improvement.

Basically, what I want to know is, are there any glaring mistakes I made that could and should be fixed? Is there anything huge that I missed? Should I just be called spoiled and go back to using printf and scanf? Please tell me, I'm looking for input here (pun not intended!)...

Thanks in advance!

What is it you're looking for in a printf-like function that you can't get with [NSString stringWithFormat:]? — Seamus Campbell, Aug 08 '10 at 21:42
The one glaring omission I noticed is the lack of a test suite. — Greg Hewgill, Aug 08 '10 at 21:44
@Seamus: if you look at the source, that's exactly what I do. @Greg: I've actually never done test suites before, so I guess now would be a good time to try, right? Thanks for the responses! — Itai Ferber, Aug 08 '10 at 22:47
Whoops. My brain was thinking sprintf, and I was trying to figure out why you were reinventing the wheel, as a thin wrapper over an existing wheel. :) — Seamus Campbell, Aug 08 '10 at 23:11

score 3 · Accepted Answer · answered Aug 09 '10 at 09:05

3

My thoughts:

Don't call them NSxxxxx, NS is reserved for Cocoa and Foundation.
Both functions should be modified to accept a FILE* i.e. you should be modelling the interface to fprintf() and fscanf() for more flexibility.
Your printf function would probably be better if you used fputs()

e.g.

void NSFPrint (FILE* fp, NSString *format, ...) 
{
    // Create the variable argument list.
    va_list args;
    va_start(args, format);

    // Using NSString, parse the argument list and convert it to a C string.
    fputs([[[[NSString alloc] initWithFormat:format arguments:args] autorelease] UTF8String], fp);
     va_end(args);
}

Consider adding support for input and output in encodings other than UTF-8.
Your scanf replacement mixes C buffered IO and Unix unbuffered IO on stdin. This might be bad.
Your scanf replacement reads up to the end of the line even when it doesn't need to. I haven't checked carefully, but if the scan format does not consume the entire line, it looks like you are discarding input. This might be bad.

answered Aug 09 '10 at 09:05

JeremyP

84,577
15
123
161

Thanks for the suggestions! I'll be implementing them as soon as possible. Just one question, though. What do you mean by my using both the C buffered IO and Unix unbuffered IO, or at least, how can I avoid this problem? – Itai Ferber Aug 09 '10 at 09:21
I've updated my code a little (http://snipt.org/XmL) to reflect your suggestions (I've renamed the functions `IFPrint` and `IFScan`, they now take file pointers as arguments, `IFPrint` now uses `fputs`, and I've replaced `read(0, NULL, 1)` with `fgetc` - is that what you meant with the buffered/unbuffered IO?). Hopefully, I'm heading in the right direction. – Itai Ferber Aug 09 '10 at 11:22
2

Yes. The `read()` function call is a Unix system call and it reads direct from the file descriptor using nothing but Unix IO calls. `fgets()` is a C library function which operates on a `FILE*`. file pointer. The file pointer maintains a buffer and fgets only reads from the physical file when this buffer is exhausted. And when it does read, it reads in a large block. That way, doing `fgetc()` is not as inefficient as you might think because most of the time it gets the character from an internal buffer. – JeremyP Aug 09 '10 at 13:24
Now, how about this: http://pastie.org/1081717? I realized that if I ever had to use `IFPrint` to write to a file, I'd be much more likely to work with an `NSFileHandle` than a `FILE *`, so I worked accordingly. `IFPrint` will keep the same efficient code as before (printing to `stdout`), while `IFFPrint` will print using an NSFileHandle. I split up `IFScan` in the same way, letting the user scan through an `NSFileHandle` with a given offset. Do you think this was worth the effort, or am I just wasting my time? – Itai Ferber Aug 09 '10 at 13:46

Printf/scanf replacement methods: some feedback/advice needed

1 Answers1

Linked