Friday, May 1, 2009

Ancestry.com Response to Census Quirk Problem

I've exchanged several emails today with Tony Macklin from Ancestry.com who confirmed that my post yesterday, titled Ancestry.com Quirk - Different Results for Old Search and New Search is a real problem for Ancestry's New Search option.

Tony said I could pass this information on:

"It turns out that you’ve uncovered a very specific bug in the code for lifespan filter (that we launched on Weds) that had managed to slip by our QA process.

"What is happening with your search is that on new search, lifespan filter SHOULD be looking for records where birth date could be anywhere from 1821 to 1825. However, because of the bug, it is ACTUALLY only returning results where the person was born BEFORE 1821 and lived until AFTER 1825. Unfortunately, this only impacts new search as this code only appears in the new search UI. Once we correct this, you should see the results returning exactly as you expected.

"We are correcting this as a matter of priority, and will push it live to the site as soon as we can within the next week. I’ll let you know as soon as I have an exact date."


And then later, he explained it a bit more:

"The original search results you had were:

........................... Name ............ Birth ... Possible lifespan
1850 census Issac Seaber 1824 1824-1924
1860 census Issac Seabury 1821 1821-1921
1860 census Issac Seaver 1824 1824-1924
1870 census Issac Seaver 1824 1824-1924
1880 census Issac Seaver 1824 1824-1924
1900 census Issac Seaver 1823 1823-1923

"The search should be asking for records where the person could have been born from 1821 onwards –

"Instead it is incorrectly asking for records for people that could have been alive in 1823 – so only two of these are returned.

"We are working on this right now, and are aiming to have the fix live within a matter of days."

I really appreciate Tony taking the time to provide a clear reason why this "quirk" happened. As a former FORTRAN programmer, I can clearly see how this happened - someone coded it wrong and the QA process didn't catch it.

This came up when they rolled out the improvement on Wednesday that limits search results to a range of 107 years if the user puts in only a birth or death year. That complicated the coding and a mistake was made so that only persons alive in the year specified were included in the search results. Tony's explanation clearly demonstrates how this happened in my case.

Clearly, this problem occurs in more than the census searches. The message for the next few days is clear - use the "Old Search" which handles the birth year range correctly. If you've done searches in Ancestry's "New Search" in the past three days, and gotten no matches, you should go back and use "Old Search."

If you are "stuck" in "New Search," you can access "Old Search" at http://search.ancestry.com/search/default.aspx?new=0.

In "Old Search," I use the Advanced Search box to input information and choose which items I want to make "Exact."

2 comments:

Unknown said...

Clearly seems to be the word of the dae. Yes, most programmers have coded something similar, clearly hoping that testing might catch the inadvertant error. Often it is the user who finds the bug -- and clearly it would remain if unreported.

Happy Dae·
http://ShoeStringGenealogy.com

Geolover said...

Randy, thank you for posting this information.

There is one teeny error at the end of your post, where you say:

In "Old Search," I use the Advanced Search box to input information and choose which items I want to make "Exact."

1) One cannot specify what items one wishes to make 'exact' in Old Search. One can make all of the items you enter "exact" or not. If you choose the "exact" mode in Old Search, however, you can also select "soundex" for the name -- which is not available in New Search UI.

2) The only function of "advanced" in New Search is to disclose the "exact" options for search parameters. This is a bit silly, and that option should be visible from the default page (as many have pointed out).