Dear friends,

I am wirting a program that involves a great deal of writing to files. In order to constantly checking the file size , I have included a lot lines of fflush(), so that I can use ftell(File) in the debug window to monitor the size of the file on the fly. I am not sure if this is good practice, is there a down side of using fflush so many times? Thanks in advance for your replies!

Best regards,
Dave

fflush flushes the buffer and forces a write to file.
You can use unbuffered write calls to avoid needing to rely on a call to fflush. However, the reliability will come at the cost of what the buffering provides.

Speaking generally here: The less bytes you write between each fflush() the slower your program is. fflush() is best used only before closing the file to ensure that everything from the write buffer gets sefely to the file.

The file IO operations are relatively slow operations (e.g. compared to memory) involving seek time, write speed and read speed. Therefore it's better to use buffers to make IO more effective. A larger chunk of bytes gets written once. If every small piece of data was to be written separately it follows that for every piece the harddisk spends time in searching the correct place to continue the writing (==seek time). And that is exactly what happens if you force the system with repetitive use of fflush().

If you are appending to an existing file then you can use ftell() once right after opening the file so that your program knows the initial size of the file. Then you should keep a counter variable which you increase by the amount of written bytes every time you write to the file.

It would be also interesting to know what exactly your program needs to monitor the filesize for. Is it because of targetting for some limitation like floppysize 1.44 MB ? If so then keep in mind that you need to take into account some overhead for the actual size in the filesystem. One reason is the allocation unit size used by the filesystem.

Why would fflush() be best used before closing the file? Closing the file already flushes the file! fflush() is best used when you wish to make certain data gets into the file when you do not want to close the file.


Generally, this is very true. But if your data is critical, and losing some because of a problem would be a disaster, the output should be written as soon as it's available.

I agree with the concept of keeping track of the file size by incrementing counters. That way in the debugger all you need to do is watch the counters and not execute the ftell() function.

__________________

Cow: You're a lawyer too?
Mooseblood (mosquito): Ma'am, I was already a bloodsucking parasite. All I needed was a briefcase!

Dear friends,

Thanks a lot for all your responses.

The reason I want to examine the file size is I want to fit the generated stream into packets in order to transmit over the Internet.

It seems now that inserting a counter is a better way. I think I will go from here, thanks again guys!

David