That was my original idea - and to take things to extremes, you might even want to recurse directories: using only a couple of recursiveness levels (1-1000/1-1000/1-1000.jpg) would ensure 1 bil. files without getting more than 1000 in any singe directory. The really nice thing about this approach is that you can start with a single level and in the unlikely event you get more than 1 mil. (you wish! :)) then you can add another in no time, with a down time of probably ten minutes.
In a slightly different train of thoughts, since this approach is so flexible, I'd recommend going for 500 files per directory - that's a limit which guarantees no slowdown. (And yes, I've read Jay's argument, but I have heard so many counter-arguments throughout time that I personally prefer to stay on the safe side - it's no big deal after all). Just my 2c, obviously. Bogdan Justin French wrote: > So hash them into groups of 1000 images???? > > 1-1000 > 1001 - 2000 > 2001 - 3000 > > etc etc, and just name the files 1.gif, 2.jpeg, etc etc > > Justin French > > > > on 22/08/02 12:41 AM, Scott Houseman ([EMAIL PROTECTED]) wrote: > > >>Hi there. >> >>So what you are suggesting is using an AUTO_INCREMENT field, possibly the >>image's Primary Key as an identifier >>for that image file., which is fine by me, but surely one should store files >>across directories, as 10000 images >>in a single directory might slow down access to those images in the >>filesystem, or not so? >> >>Thanks for your input. >> >>Regards >> >>-Scott >> >> >>>-----Original Message----- >>>From: DL Neil [mailto:[EMAIL PROTECTED]] >>>Sent: 21 August 2002 04:31 >>>To: [EMAIL PROTECTED]; Bogdan Stancescu >>>Subject: Re: [PHP] Image library >>> >>> >>>Scott (confirming Bogdan), >>> >>>Libraries of all types have had this concern for years - even though books >>>are uniquely identified by ISBN, that is still not good enough for library >>>purposes (eg multiple copies of a single title). So they, exactly >>>as Bogdan >>>suggests, use an "Accession" number sequence - which can be >>>implemented very >>>neatly in MySQL (from PHP) as an AUTO_INCREMENT field. >>> >>>Regards, >>>=dn >>> >>> >>>>I've seen this kind of random approach several times and I keep >>>>wondering why not counting the files instead. Yes, it may take a little >>>>longer when uploading but I personally think the safety of the approach >>>>is worth the insignificant speed sacrifice. >>>> >>>>Bogdan >>>> >>>>Scott Houseman wrote: >>>> >>>>>Hi all. >>>>> >>>>>This confirms what I suspected. >>>>> >>>>>The hash algrithm: >>>>> >>>>>I have a directory structure: dirs 0 - f, and within each of >>>> >>>these, the >>>same >>> >>>>>dir structure 0 - f. >>>>>When an image gets uploaded into the library, do an md5sum of >>>> >>>the file, >>>take >>> >>>>>the first 2 chars of that hash >>>>>and there's your path. e.g >>>>>$PICDBPATH.'/a/7/a7b8be10b0e69fe3decaa538f1febe84' >>>>> >>>>>I'm not sure what the mathematical randomness of this is, but I'm sure >>>> >>>it's >>> >>>>>pretty random, and the chances >>>>>of collision should be virtually null, the only time you should >>>> >>>overwrite a >>> >>>>>file is if you upload the exact same file(?) >>>>>Is there a better way of doing this? >>>>> >>>>>Cheers >>>>> >>>>>-Scott >>>>> >>>>> >>>>> >>>>>>-----Original Message----- >>>>>>From: Justin French [mailto:[EMAIL PROTECTED]] >>>>>>Sent: 21 August 2002 03:25 >>>>>>To: [EMAIL PROTECTED]; PHP General >>>>>>Subject: Re: [PHP] Image library >>>>>> >>>>>> >>>>>>on 21/08/02 9:45 PM, Scott Houseman ([EMAIL PROTECTED]) wrote: >>>>>> >>>>>> >>>>>> >>>>>> >>>>>>>Which way would be the most efficient/fastest to access images >>>>>> >>>>>>from an image >>>>>> >>>>>> >>>>>>>library. >>>>>>>A) Store image files in a hash directory structure AND storing >>>>>> >>>>>>each file's >>>>>> >>>>>> >>>>>>>information in a mysql table >>>>>>>OR >>>>>>>B) Storing image information in mysql table AND storing the >>>>>> >>>>>>image in a BLOB >>>>>> >>>>>> >>>>>>>field in that table. >>>>>> >>>>>>>From all accounts I've read on this list, a database is not >>>>>> >>>>>>usually faster >>>>>>than a filesystem. And for large amounts of files, like 1000's, >>>>>>a hash will >>>>>>speed it up more. >>>>>> >>>>>> >>>>>> >>>>>> >>>>>>>The way I see it, considerations to be taken into acount: >>>>>>>- Is it quicker/better to retrieve image from table & then stream out >>>>>> >>>to >>> >>>>>>>browser OR simply direct the browser to the file? >>>>>>>i.e <IMG SRC="/imagelib/image.php?iImageID=10"> OR <IMG >>>>>>>SRC="/imagelib/5/f/10"> >>>>>>>- Will a database OR filesystem be more scalable i.e. which >>>>>> >>>wil perform >>> >>>>>>>better when there are 10000 images in the libary? >>>>>> >>>>>>Filesystem should be quicker. You need to think about how >>>>> >>>you hash the >>> >>>>>>files up for the most even spread of files in each directory I guess. >>>>>> >>>>>> >>>>>>Justin >>>>>> >>>>>> >>>>> >>>>> >>>> >>>> >>>>-- >>>>PHP General Mailing List (http://www.php.net/) >>>>To unsubscribe, visit: http://www.php.net/unsub.php >>>> >>>> >>> >>> >>>-- >>>PHP General Mailing List (http://www.php.net/) >>>To unsubscribe, visit: http://www.php.net/unsub.php >>> >>> >> > -- PHP General Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php
