mersenneforum.org The Appallingly Blue Page
 Register FAQ Search Today's Posts Mark Forums Read

2012-10-26, 18:48   #89
Dubslow

"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

1C3516 Posts

Quote:
 Originally Posted by Greebley Not sure if this is helpful, but when I get sequences, I base them on the value I last got. So for example, for 276 I call: wget "http://factorization.ath.cx/elf.php?seq=1318317059659669408240038846262480443602748610435445388542986041125974944748075928636239695427694073168827641 38573827680383355306759333640241337963062754830971836250952068019746298&type=1" -O final_276.elf This gets me the NEW values only. In general, I can call this for all 47941 sequences (number of all open seqs less than 5 million) without hitting the limit. I only do this rarely (it takes a long time to run) but most sequences are the same so there is only one line to get so not a huge strain on the DB. It also gets past previous data base errors because it doesn't get the bad bits each time. It is vulnerable to new database errors (to get past that I have a program to append the new lines and can then use aliqueit -t to check the full elf). You would need a way to get the seq number and the current last value and to run the above calls. In my case, my stats program generates a bat file for me to run (the program actually does 3 things, generate the stats and bat file, append the final to the existing elf, and check for merges so I can update the current open sequence list) As a side note, someone has been working on the 1-5 million range as they are all up to around 70 digits. That is why I extended to 5 million. Would make your table too big though. Interesting fact: It takes a bit over 6.6 billion bytes to represent all the sequences in text/elf format for seqs up to 5 million with duplicates removed (sequences end when they merge). Of course this is always growing as more of each sequence is discovered. Edit: One thing to mention is that I use the last full line in the elf for above and not the partially factored one, so I end up with elf files with at least 2 lines - the full and partial. It should also work with the final value though and give one line like I mentioned above.
Quote:
 Originally Posted by Dubslow Between that and a new feature I wanted to add, the script has been reorganized to query the last ID and only check the actual sequence page if the ID is fully factored, as henryzz suggested a while ago.
So, for e.g. 276, my query would be "http://factordb.com/index.php?id=1100000000504144112", which is the ID of the last line. If it's fully factored, then I pull a new status the old way, as opposed to with the value of the number itself like you do; since pulling a new status is done much less often than for every sequence, I don't think it makes much difference.

As recently requested, I did add in the ability to work around broken sequences, by catching a broken sequence and instead of querying that sequence, query instead one of the later numbers that I know is good (and throw in a index offset so that it appears in the table correctly). These two lines can currently be found in my script:
Code:
broken = {747720: (1977171370480, 67)}
# A dict of tuples of {broken seq: (new_start_val, offset)}
So 747720 actually queries 1977171370480, which is the 67th line of 747720. I can add sequences as necessary to this variable.

 2012-10-28, 06:04 #90 Dubslow Basketry That Evening!     "Bunslow the Bold" Jun 2011 40
2012-10-31, 14:11   #91
Greebley

May 2009
Dedham Massachusetts USA

3·281 Posts

It appears you are using the full power of the odd prime exponent when calculating class. That is why you get a class of -3 for 2*3^7 term.

The 7th power isn't supported by the 2 so I would just use the 3 and not 3^7.

That would be single powers for all odd primes except for 2 drivers:

2^3*3*5 - supports 3^2
so 2^3*3^2*5 really is class 2

2^5*3*7 - supports 3^2
so 2^5*3^2*7 is class 2 as well.

--------------

Is there a way for me to get the full data? I wanted to compare my results to yours since you do it in a slightly different way. Any difference might indicate a DB error either for you or for me.

Since I couldn't figure out a way to download yours, I attached a zipped version of my 9224 sequences arranged by sequence number. Its tab separated with sequence, size, index, and factors, though the cofactor size is a separate column. There will be small differences for sequences that changed since last weekend when I downloaded. Any sequence not recently added to though should be the same.
Attached Files
 seqTo1Million.zip (102.4 KB, 69 views)

Last fiddled with by Greebley on 2012-10-31 at 14:13

2012-10-31, 19:32   #92
Dubslow

"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

160658 Posts

Quote:
 Originally Posted by Greebley It appears you are using the full power of the odd prime exponent when calculating class. That is why you get a class of -3 for 2*3^7 term.
Indeed. This is a deliberate choice, and is an option in my sequence analysis Python:
Quote:
 >>> a.get_class(2**2*7) -1 >>> print(a.get_guide(4*49*5*13)) 2^2 * 7^2 >>> print(a.get_guide(4*49*5*13, powers=False)) 2^2 * 7 >>> a.get_class(4*49*5*13) 2 >>> a.get_class(4*49*5*13, powers=False) -1
http://www.mersenneforum.org/showpos...7&postcount=47

Quote:
 Originally Posted by Greebley The 7th power isn't supported by the 2 so I would just use the 3 and not 3^7. That would be single powers for all odd primes except for 2 drivers: 2^3*3*5 - supports 3^2 so 2^3*3^2*5 really is class 2 2^5*3*7 - supports 3^2 so 2^5*3^2*7 is class 2 as well.
I don't know what you mean (specifically, "support"). My personal knowledge of how drivers/guides work is very limited; my module above is based on an algorithmic reading of Clifford's analysis, without any particular personal comprehension. As you can see above, I am aware that the powers affect the class, but I don't know how/why/if it's good to know. (I think it's good to know, but that's just me )

Quote:
 Originally Posted by Greebley Is there a way for me to get the full data? I wanted to compare my results to yours since you do it in a slightly different way. Any difference might indicate a DB error either for you or for me. Since I couldn't figure out a way to download yours, I attached a zipped version of my 9224 sequences arranged by sequence number. Its tab separated with sequence, size, index, and factors, though the cofactor size is a separate column. There will be small differences for sequences that changed since last weekend when I downloaded. Any sequence not recently added to though should be the same.
If by full data you mean not just AllSeq.txt, then you should know that AllSeq.html is really just a pretty-looking HTML/JS frontend to AllSeq.json, which contains the data as output by the script (in JSON format obviously). I could perform a comparison with your data, if you still want me to.

2012-10-31, 23:22   #93
kar_bon

Mar 2006
Germany

2×1,433 Posts

Quote:
 Originally Posted by Greebley There will be small differences for sequences that changed since last weekend when I downloaded. Any sequence not recently added to though should be the same.
As expected there're no great differences in both data files.

There're many seq. sizes, which seems not correct.

Example:
Code:
564	3373	175	2^2*3^2*7*13*71*4292236942619	175
The last 175 should be a 157 here. So the "Cofactor size" is often the same as "Size" in your data.

 2012-11-17, 17:24 #94 fivemack (loop (#_fork))     Feb 2006 Cambridge, England 7·911 Posts Which sequence finished? I noticed (the site is now down) that the count of sequences on http://dubslow.tk/aliquot/AllSeq.html went from 9224 to 9223; which is the lucky number?
2012-11-17, 18:25   #95
Batalov

"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

926110 Posts

Call me old-fashioned, but I still have 9225.
Attached Files
 ALL9225.zip (55.8 KB, 64 views)

2012-11-17, 20:42   #96
Dubslow

"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts

Quote:
 Originally Posted by fivemack I noticed (the site is now down) that the count of sequences on http://dubslow.tk/aliquot/AllSeq.html went from 9224 to 9223; which is the lucky number?

I left my dorm yesterday for the week, and left the computer on along with a note to maintenance to please leave it on. If the site's down, then most likely is that someone shut it or the router off (or unplugged them or some such variation). It was working last night when I got home (three hours after leaving), so it was definitely up when I left.

There's a small (small) chance it may go back up, but the overwhelming probability is that it will stay down for the next 8 days, until I get back.

Quote:
 Originally Posted by Batalov Call me old-fashioned, but I still have 9225.
I obviously can't check my own database, but the last merge was the one that Clifford reported via you. (Perhaps fivemack could check his daily snapshots of my data?)

 2012-11-18, 00:00 #97 fivemack (loop (#_fork))     Feb 2006 Cambridge, England 7·911 Posts Unfortunately the thing I have daily snapshots of is the statistics table rather than the data. Code: >>> sum([t[1] for t in a["aSizes"]]) gives 9223 for the 20121117 statistics table and 9224 for the 20121115 one.
 2012-11-18, 00:25 #98 kar_bon     Mar 2006 Germany 2·1,433 Posts Seq. 622560 is missing in the 2012-11-16 data file... not terminated in FactorDB. Last fiddled with by kar_bon on 2012-11-18 at 00:26
2012-11-18, 03:14   #99
Dubslow

"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3×29×83 Posts

Quote:
 Originally Posted by kar_bon Seq. 622560 is missing in the 2012-11-16 data file... not terminated in FactorDB.
Thanks, I'll look into it in 8 days. Sorry guys, I don't know there's anything I can do until then.

 Similar Threads Thread Thread Starter Forum Replies Last Post xilman Science & Technology 24 2015-10-16 00:03 retina Lounge 32 2015-03-28 13:08 fetofs Puzzles 28 2005-11-03 15:50 ixfd64 Hardware 9 2005-11-01 11:34 dave_0273 Hardware 17 2005-05-19 14:33

All times are UTC. The time now is 05:25.

Thu Jan 21 05:25:20 UTC 2021 up 49 days, 1:36, 0 users, load averages: 1.62, 1.96, 2.22