Speed of Relationship calculator
Brought to you by:
canajun2eh,
yalnifj
Menu
▾
▴
#357 Speed of Relationship calculator
Perhaps the feature I like best is the relationship
calculator. Unfortunately, its also the one feature
that IMHO is a little slow. I have a 70,000+
individual gedcom, and it often takes upwards of 6
minudes to determine a relationship on a 3 GHz machine
with a gig of memory and RAID drives.
Also, unless I have missed something, whenever I hit
"Relationship to me", it goes off looking for a direct
relationship. There is no way to avoid thiis, unless
one goes to the relationship chart and keys in the
identifiers for each of the individuals. It may take
two or three minutes to come back and respond that
there are no direct relationships, at which time one
can then then check for relationships by marriage.
Legacy in a couple of seconds can determine all direct
relationships to an individual. Why so long in
phpgedview? Is it the difference between php and C++?
Even with 70,000+ individuals, the database is still
quite small as compared to physical memory, so I can't
believe it's disk time.
Last time I checked, looking for a relationship by
marriage does not detect direct relationships. Is that
what is intended?
My suggestion on the "Relationship to me" is that it
either gives you the option on which type of
relationship you are interested in prior to searching,
or left mouse click = 1 type and ctrl +click = another.
Just my thoughts on improving what is already a great
product.
Ron
Logged In: YES
user_id=300048
Hi Ron,
I agree that the relationship calculator is slow. The greatest
bottleneck is the database. If you turn on the statistics, you
will see that it takes thousands of database queries to try
and find the relationship. This is because every person that
it checks has to be retrieved from the database and every
family that it passes through also has to be retrieved.
In a normal application you don't have this problem because
the database is in memory. So you can go searching through
memory to follow relationships without waiting for a database
query.
My algorithm is probably not the greatest. I based it on
search algorithms that I learned in an artificial intelligence
class I took in college. I wonder if other genealogy
relationship algorithms are out there somewhere in the vast
internet. I haven't ever looked.
What I would like to do, and where I think we would find the
greatest benefit in this area is to have some sort of
numbering scheme for the people in the gedcom. An example
of how we might go about this is to do something like this:
Father = 1
Child 1 = 1.1
Child 2 = 1.2
Child 3 = 1.3
Child 1 of Child 1 = 1.1.1
Child 2 of Child 1 = 1.1.2
Child 1 of Child 2 = 1.2.1
With this I could get all of the descendants of Father just by
finding all of the people whose numbers start with "1." which
also means that anyone who starts with a "1." is related.
Anyway, this particular scheme has a lot of holes in it and
generating the numbers to start with would be difficult.
--John
Logged In: YES
user_id=998633
Hi John,
Sorry for the delay - I thought I had responded.
I run windows, and although it does a lot of mysql accesses, I
notice that most of the CPU time, and it is cpu bound, is
taken up in apache. I believe that means that most of the
time is taken up in the php code.
I guess php does not generate object code, but isn't there a
way of distributing php projects in psuedo code that works a
lot quicker?? If so, perhaps full dot revs could be compiled
and distributed in that manner?! Means that users won't be
able to avail themselves of CVS, however speed has it's
advantages, especially with large gedcoms.
regards,
Ron
Logged In: YES
user_id=1061833
It would also be a nice addition to this feature if a
summary of the relationship appeared at the top. GEDCOMIT
reports the following (not bragging you understand) and the
calculation took about 12 seconds whereas it timed out after
120 seconds in PGV:
"George Walker Bush is your sixth cousin once removed." when
reporting the relationship to me.
stephen
Logged In: YES
user_id=1463438
I also found this feature slow, and decided to do something
about it. Turns out the algorithm to find direct-line
relatives can be very speedy - even eith 34,000+ names in a
file. I would share my patch - if I knew how.
See it in action at http://www.ournetwork.net/FamilyConnections
using pids I409 and I5994, or try other combinations.
For a real treat, try I409 with himself, or with his wife I410!
Haven't got all the kinks worked out yet, but for me this is
waaaaay better than what is in the 3.3.8 release.
Logged In: YES
user_id=1566627
Would be good if there actually WAS an "relationship to me"
function in 4.01
Best I can find is a chart naming the relationship at each
intermediate connecting link between two individuals - no
textual description of the overall relationship between the
two people, which is a BIG omission!
Logged In: YES
user_id=634811
Originator: NO
Can this be closed?
Logged In: YES
user_id=1061833
Originator: NO
While my comment was resolved with the now existant relationship verbage, the speed issue still exists. On a VERY fast machine with our 46,000 INDI's, many calculations still time out or take more than 3 minutes. Practically useless.
-Stephen
Logged In: YES
user_id=1463438
Originator: NO
I have a fix I use in 3.3.8, and am working on adding it to the project's current dev release. My fix finds up to 200 or more direct-line relationships in seconds, and names the overall relationship when each one is displayed.
Will need help with translations of the relationship descriptions. Also - I am aware that there are (at least) 2 ways of describing a relationship; one used widely in North America and one used in France (and perhaps elsewhere). Any ideas on how to tackle this?
Barry