Re: [Apc-aa-coders] Ids of strange lineths.

[Honza M. <hon...@ec...>]

On P=E1, 2003-01-31 at 23:08, Mitra wrote:
> Honza
>=20
> Did you make this change to new_id?
> I don't see it in the new_id in the current CVS

I do not send it to CVS, yet. I have quite a lot of changes, which I do
not want to send to CVS at this time. If you can add it here I will
appreciate it:

/**
 * Returns new unpacked md5 unique id, except these which can force
 * unexpected end of string (or other problems)
 */
function new_id ($seed=3D"hugo"){
    do {
       $foo=3Dmd5(uniqid($seed));
    } while (ereg("(00|27)",$foo) || (substr($foo,30,2)=3D=3D'20')); =20
      // '00' is end of string, '27' is ' and packed '20' is space,=20
      // which is removed by MySQL
    return $foo;
}=20

Honza

> - Mitra
>=20
>=20
> At 2:43 PM +0100 25/1/03, Honza Malik wrote:
> > Hi Mitra,
> >=20
> >    I did some more testing on ids and now I know, where the problem
> > is.
> > The new_id() is OK - it returns allways 32 character long id, but
> > sometimes it return id with ...20 at the end. Packed '20' is space
> > and
> > this is problem, because if you insert 16 characters long packed id
> > ending with space into MySQL and then you get it back, the id is 15
> > character long.
> >=20
> > MySQL REMOVES ENDING SPACES FROM INSERTED STRING!
> > I will fix new_id() to not generate such ids.
> >=20
> >=20
> > On P=E1, 2003-01-24 at 12:28, Mitra wrote:
> > > Ok - stranger .... my guess was wrong, md5 returns a 32 character
> > > string, which cannot contain 00 or 27 so cannot end up as a
> > shorter
> > > string.
> > >
> > > To test, I ran a loop through 99999 samples of new_id, which
> > should
> > > have generated about 400 bad ids by your statistics.
> > >
> > > I think the bad ids in your database must be the result of some
> > buggy
> > > earlier code.
> > >
> > > Lets try three things.
> > > 1: Please check my test - appended at the end of this message.
> >=20
> > I test it and all 99999 ids are 32 characters long.
> >=20
> > > 2: Please re-run your test, see if the number of bad ids has gone
> > up,
> > > if there is something bad then as long as you have created more
> > than
> > > 256 ids then statistically it should have increased by 1 or more.
> >=20
> > Increased: [16] =3D> 92254 [15] =3D> 367 [14] =3D> 4
> >=20
> > > 3: Please send the code for your test, I'll run it on my
> > installation
> > > and see how many odd length ids I have.
> >=20
> > require "include/config.php3";
> > require $GLOBALS[AA_INC_PATH]."util.php3";
> > require $GLOBALS[AA_INC_PATH]."locsessi.php3";
> >=20
> > $db =3D new DB_AA;
> >=20
> > $db->query( "SELECT id from item" );
> > while($db->next_record()) {
> >     $fromdb[$len_p =3D strlen($db->f('id'))]++;
> >     $unpdb[$len_u =3D strlen(unpack_id($db->f('id')))]++;
> > //    if( $len_p<16 )
> > //        echo "<br>".$db->f('id');
> > }  =20
> >=20
> > huh($fromdb, 'packed');
> > huh($unpdb, 'unpacked');
> > echo "
> >        Regards
> >                  Honza ";
> > >
> > > - Mitra
> > >
> > >
> > > At 11:55 PM +0100 22/1/03, Honza Malik wrote:
> > > >  >3) Your assumption that packed ids are EXACTLY 16 characters
> > long (and
> > > >  >unpacked are EXACTLY 32 characters long) is unfortunately
> > wrong. I can't
> > > >  >say you why - we generate the ids by new_id() function, but in
> > our
> > > >  >database are packed ids for item 16 (in 91907 items), 15 (366)
> > and 14
> > > >  >(4) characters long.
> > > >
> > > >  Strange - because new_id uses "md5" which has to return 128
> > bits.
> > > >
> > > >  I wonder if these were ids which generated a leading ascii
> > zero, the
> > > >  ratios look reasonable for this, i.e. 919007/366 =3D 251, which
> > is
> > > >  close to a one in 256 chance, i.e. something that occurred when
> > 8
> > > >  bits (2^8=3D256) all were zero.    This makes sense if we assume
> > that
> > > >  md5 returns a number which the ereg in new_id casts as a
> > string, and
> > > >  so ignores the leading zero.
> > > >
> > > >  I've fixed zids.php3 to allow 12..16 character packed ids, and
> > 24..32
> > > >  long, but this could break my tagged ids. I've also fixed
> > > >  util.php3.new_id to only return a string of exactly 16 chars.
> > > >
> > > >  > >5) In zids.php3 you wrote:
> > > >>  >
> > > >>  >function id_t2l(&$str) {
> > > >>  >#TODO: Look online for quicker way to substr last 32 chars -
> > mitra
> > > >>  >         if (ereg("(.*)([0-9a-f]{32})",$str,$parts))
> > > >>  >                 return $parts[2];
> > > >>  >
> > > >>  >The quicker way is: substr($str, -32), but see 3)
> > > >>
> > > >>  Ok - I'll make this 24..32, and I guess we can't use tags that
> > end in
> > > >>  a hex character [0-9a-f)
> > > >
> > > >Or you can use some delimeter (tag:unpacked_id). As I write
> > before, we
> > > >never use tags so far, so the syntax is on your choice.
> > >
> > >
> > > require "../include/config.php3";
> > > require $GLOBALS[AA_INC_PATH]."util.php3";
> > >
> > > $i =3D 0;
> > > $j=3D 99999;
> > > print "<br>Testing start, loops=3D$j: sample=3D";
> > > $foo =3D new_id();
> > > print $foo;
> > > while ($i++ < $j) {
> > > $foo =3D new_id();
> > > if (strlen($foo) !=3D 32) print
> > "<br>FOO:$i:$foo:strlen=3D".strlen($foo);
> > > if (strlen(pack_id($foo)) !=3D 16) print "<br>FOO:$i:$foo:packed
> > > strlen=3D".strlen(\
> > > pack_id($foo));
> > > }
> > > print "<br>Testing end sample =3D ";
> > > $foo =3D new_id();
> > > print $foo;
> > >
> > > print "Done\n";
> > >
> >=20
> >=20
> >=20
> > -------------------------------------------------------
> > This SF.NET email is sponsored by:
> > SourceForge Enterprise Edition + IBM + LinuxWorld
> > http://www.vasoftware.com
> > _______________________________________________
> > Apc-aa-coders mailing list
> > Apc...@li...
> > https://lists.sourceforge.net/lists/listinfo/apc-aa-coders
>=20
>=20
>=20
> --=20
> Mitra Technology Consulting -  www.mitra.biz - mi...@mi...
> 02-6684-8096  or 0414-648-0722=20
>=20
> Life is a Mystery to be Lived, not a Problem to be Solved
>=20