Pat wrote:z666zz666z wrote:
Each row has to be of exact same length 225*3=675 charactes
i almost got it into your format
except my file (attached) has an extra TAB separating cells -- sorry
Pat, it does not matter at all that extra TABs, i can use Notepad++ to remove them... as easy a Find&replace "\t" to "".
I have downloaded the file just perfect, thanks at all.
On previous post when i said i did not see it when i go to...:
http://www.setbb.com/sudoku/viewtopic.php?t=1830&mforum=sudoku
I did not see it, just because my Browser show the numbers as one pixel per digit (like it it where an image) and i did not try to select all (i did not even think it could be text), sorry for that... in other words the text was shown as a font with a size too small to be able to read anything of it and no scrollbars at all.
On previous post when i said i did not see it when i go to...:
http://forum.enjoysudoku.com/a-new-100x100-t6395.html
I did not see it, just because i did not understand the format and so not see it as a sudoku (if it would have scrollbas so each row of sudoku would be on a line maybe i would understand the format).
Again, thanks to Pat for putting the file as an Atatachment (a file in a Plain Text format so any text viewer, notepad, etc can read it)... in that way it is very easy to find it... just with one click can be downloaded.
Note for those that says they have no problems to find it on:
http://www.setbb.com/sudoku/viewtopic.php?t=1830&mforum=sudoku
http://forum.enjoysudoku.com/a-new-100x100-t6395.html
I went again to them and i still do no see a download link there, where is it?
As i said i do not see any link where i can download any file, like Pat has done on his post, so in my opinion it is not trivial to find it there, at least not as trivial as on Pat post... a direct link... correct that there is a lot of text i can Copy&Paste, but my browser (Google Chrome) show it with a so small font size i can not even notice that was the text of sudoku cells, so i never think on doing a Copy&Paste.
Again thanks to Pat for putting a "direct link" to a "Plain Text format file" (no matter that tabs, i can remove them or adapt my algorithm to suport them)...
Note about the format and why i choose that one:
The format is a very common and standar format for "Information Exchange" on all over the world... a Plain text format, each cell with a fixed size... so the format is formally called a "ASCII fixed column", the Pat ones is also "tab delimited format".
I used that other format (similar to Pats one but without Tabs) since i want to test my program not only with 225x225 ones, also with 10000x10000 and whit a "OneMillon x OneMillon" one
Just for a Madnness extreme i did a test with a "OneMillon x OneMillon" one... i did not let it to finish since it was only just to check if it would work of just hang, crash, etc...
Note about the size of the file for that one, the total madness "OneMillon x OneMillon" one:
The file on plain text without Tabs has a size of 5.457TB, yes TeraBytes... so no PC has enough ram to hold it fully on RAM on "ASCII Plain text" format... neither on binary format since it takes at least 2.73TeraBytes.
If i include such tabs the file goes up to 6.366TB, so the tabs will take One TeraByte extra, that matters a lot...
I know it was a Really Full Madness One ... it needs a big rack of HDDs to hold the info... near 6 extrenal 2TB HDD (USB3.0/Sata-III) on Raid stripping ... and for solving it and extra 2TB HDD for temporal info ... i also calculated that if i was too madnees to try to print it on paper it would maybe need more paper than there is on the world, well not really so much, but something really near 400 millon ( yes, 400000000 ) pages of paper... i am not as mad to print that, je je... neither to try to solve it manually of course, just to check the algorithms... i neither let it to finish... when it solved up to 25% i shut it down, takes too, too much time... estimated more or less a month of a High speed home-super-computer with a 4x4 (four quadcore computers).
Yes, i know i am a little paranoid when testing my own made algorithms... why prove it with small amunt of data if i can test it with huge branch of huge data!!! ... with that ultra-madness-high-amount of data you can also check extrange and rare conditions where data is not readed on time, bottlenecks, etc... and one very important thing... how much more time needed for extra size.
Oh! i am also more madness...
I did some cals for a 100 times that (a ten millon x ten millon one) but unfortunatly i do not have enought HDDs to aford that... a faked estimation tells me it would took to "find a valid one & solve it" about a full year of computer... i am not as mad to pay the cost of electricity for that task... so i desist on trying that one... but just do a simple Maths... since values can go from 0000000 to 9999999, that is 7 chars per cell, 10 millon cells per row (i removed the line breaks, more easy to process with direct file access) and ten millon rows, so 7*10^7*10^7 Bytes... that is 700TeraBytes, wow!!! that much just to hold one step.
Actually i have on External 2TB HDDs near 20Terabytes... (ten 2TB HDDs)... for huge test tasks like this ones, but i would need about 40 times what i have... i am not so rich, i can not even pay for that... maybe someone give them to me as a gift or let me free a "cluster super-computer"... but then where could i "storage" the "solution", je je... nowhere near me!!!
So finally i desist trying it, je je... with that Maths i saw the bigger one i could try was near "One millon x One Millon", so i test it not to end!!! It was great to see it did not crash and was really fast on getting some cells values... at least my algorith works with such huge data, je je... maybe it can not solve some of 9x9 but at least i think i had done things well so it can solve any one of any size if have enought HDD and time, it does not works on RAM, it uses HDD direct file access... a huge file working as RAM does (you can read/write any byte at any time) but slower of course, the file does not change on size, so never get fragmented, etc... speed is allways the same for the whole proccess... that is very important since if you let file get fragmented, read/write speed can decrease.
Well, again thanks a lot to Pat, now i have it... now i will try to solve it with pencil and paper... then after that, maybe i will pass it as input to my algorithm if i get "cycles" on the home-super-computer, now it is working on other tasks queues (and i am lot allow to use it till ends).
Thanks a lot to all of you.