Jump to content


ljwolfe

Member Since 26 Jun 2018
Offline Last Active Sep 07 2018 02:58 PM
-----

Posts I've Made

In Topic: Compare files doesn't work

28 August 2018 - 02:43 PM

OK, I can't seem to change the subject. Unsurprising, really, and understandable. So consider it to read "...doesn't work properly." Because it doesn't, as others besides myself have noted.


In Topic: Compare files doesn't work

28 August 2018 - 02:42 PM

The title of this post is subjectively misleading. Of course it works !
What about the express purpose of the menu item... to compare NON-EXACT copies of database files? The OP is off in the weeds about the 96-99% matches of an exact copy of the database.

Thanks for the insult </sarcasm> And I did get those results. I could post a series of screen shots (way too many to get in one screen) if you don't believe me.

My point was to give an example of the far-too-numerous false negatives that come up in file compares. If I compare two files that are similar, but NOT exact, to see what the differences are (which is the "express" purpose of the function), it would take hours to weed through all of the people who are identical but the compare flags as different, in order to find the people who actually are different and need to be addressed. As I said earlier, in this case (comparing "NON-EXACT copies of database files") the feature is broken. It's almost faster to print out a report and compare everyone visually.

Looking for duplicates in a single file (Duplicate Search Merge) is a different use, and one that is marginally useful, but I wasn't posting about that function. RM is OK at finding people who are mostly the same for that use. But it's terrible at just finding people who are different.

But, for the pedants, I'll see if I can change the subject line.


In Topic: Compare files doesn't work

27 August 2018 - 04:37 PM

"You do have some that score 100%, don't you?"

Yes, that's why I said 100s instead of 1000s.

 

"It is going to test a number of fact types, e.g., Birth, Death plus Name and possibly the parents and spouses of a given person for a match. Let's say there are 5 such fact types and each is weighted the same. If the person has 4 of those 5 fact types and they obviously match on the comparison database, then the score would be 80%."

And when both people have 5 of 5 facts with the same info - same dates, locations, notes, and sources -- then shouldn't it be 100%? But it isn't on far too many people.

 

"When a person has multiple facts of the same type they are compared against each other. This will cause the match score to lower."

Which, again, makes the file comparison virtually useless. Say two people both have five census facts -- 1900, 1910, 1920, 1930, 1940 -- with identical info for each relevant year. If RM can't tell that those people are identical, then I have to manually check each person, and then what's the point of having a comparison? Same with children. Two people have exactly the same kids, in the same order, but the comparison swaps the order of children on one set (or both, but differently) and then flags the parent as different. :-(

 

I'm with Jerry -- as it stands, for most reasons a person might want to compare files, the built-in comparison is useless. If something happens and I want to compare my current file to one from a little while ago, or if someone sends me a file with mostly the same family and I just want to see who's actually different (as opposed to having to slog through 100s of people who RM says are different even though they aren't), I might as well just create reports and then go through them by hand.

IMO, they need to either make it work, or take it out.

And by the way, if the files are identical, then that means that every single person in the file is identical -- same facts, same family, same sources, same color code, same groups, identical. I'm not sure how people could think identical files could have people who are different between the two files.