Any better way than time()

Craig321 · 24 Jan 2006 at 19:28

Hi,

Currently, infront of the name of files uploaded onto my site I put a time stamp (time()). Sometimes the time stamp is up to 10 digits out compared to the link meaning sometimes they don't work and I have to change the name of the file to make it work.

Is there anything better than time() I can use that will avoid this problem please?

[Edit]
Also, I put a time stamp in to avoid people uploading a file named the same as another file, as this would overwrite the other file.

Thanks
Craig.

Augmented · 24 Jan 2006 at 19:43

You can create more random strings than simply using time, e.g. combining uniqid()/rand() etc. However, collisions don't seem to be your issue. When you say the 'link' is not the same as the timestamp, how are you generating the link and the timestamp?

You should be storing the unique value you're creating (using time() in your case) in a single variable and use this to create the link and the filename. If you run time() once, do some processing, and then run time() again, the second value of time() will be larger than the first because of the time taken to do the processing!

When I started this post the time was 7:40, and when I finish writing it will be 7:43

.

Craig321 · 24 Jan 2006 at 19:56

True, but most of the time it works, just sometimes the link is a few out for some reason.

I'll check out rand and uniqeid and see if I like them

Thanks
Craig

matja · 24 Jan 2006 at 20:01

On my image uploader, I md5-sum the data uploaded and write to filename using that md5-sum. A md5 hash collision probability is pretty small.

Craig321 · 24 Jan 2006 at 20:05

So do you rename the whole image to a random MD5? Or just a md5 at the beginning?

Beansprout · 24 Jan 2006 at 20:19

I think what people do commonly is hash the file and then add the filename to the end, to make it "look funky-but-recognisable" to Joe Bloggs - and of course provide a 0.000000000000000000000000000000001% collision chance

PS - this is what I would do, or if the filename was too long I'd truncate it and add it to the end

Craig321 · 24 Jan 2006 at 20:26

Has as in http://uk2.php.net/manual/en/function.hash-file.php ?
I'm not quite sure what you mean by "or if the filename was too long I'd truncate it and add it to the end". ?

Thanks
Craig.

Craig321 · 25 Jan 2006 at 16:41

OK, I noticed that on the upload part i was doing time() . before the filename and then on the SQL part I was doing time() . before the filename meaning sometimes they might have been slightly out... which they were. So instead I did this:

$time = time();

And then instead of using time() before the filename I used $time

It hasn't gone wrong since

Thanks for the help everyone
Craig

Bloo_Fish · 25 Jan 2006 at 19:15

You may want to take a little more advice from this thread as the method you are using will cause problems when two people do it @ the exact same time.

Try adding something unique to the name, like a hash of the users ip or something.

Craig321 · 26 Jan 2006 at 01:45

Good idea, I think I'll add a md5 of the users IP instead of a timestamp

Thanks
Craig

robmiller · 26 Jan 2006 at 01:47

But what happens when a user uploads two files from the same IP?

md5(uniqid().time())

Craig321 · 26 Jan 2006 at 02:37

Was thinking that, that md5 uniqid time thingy is pretty good but it's a little long, makes the filename real long

Could I use rand() on it's own? Would there be any down sides to that?

Thanks
Craig.