* autofill places

[PeterR]

It would be even better if any such list of partial matches was presented in descending order of frequency of existing use.

[tatewise]

The original most-frequently-used strategy seems to have been restored in the Upgrade to FH V6.2.1 for Place name auto completion.

HOWEVER, although it appears to work correctly in most cases, there is an exception I noticed.
In the Family Historian Sample Project the Place name Bristol, England is by far the most used B at 10 times.
BUT when entering a Place starting with B then Brighton,Sussex, England is auto completed!?!
THEN if you continue and enter r FH changes its mind and correctly auto completes with Bristol, England despite both entries starting with Bri!?!

[AdrianBruce]

Yes, it's usually a bit wonky for me, too.

I have many more "Bristol, Gloucestershire, England" entries than Brighton, but also got Brighton suggested at "Bri" - OK when I got up to "Bris". ("Br" gave a suggestion that was also way out but I'll not worry about that). NB - there are other "Bris" places earlier in the alphabet that were correctly not suggested, so it is doing something.

Basically, it seems to be a bit wonky until it's got 4 letters to work with - I don't have a problem with that. Although....

"Crewe reg. district, Cheshire, England" incorrectly gets suggested vice "Crewe, Cheshire, England" even at "Crewe" - I have to go to the comma to get that one right.

[RogerF]

I'm attempting to illustrate why the FH Place autofill mechanism isn't working for me. For some background on why I use this style for Places, see http://www.firthworks.com/genealogy/RRR/index.html#R7. It's worked well for me for the past eight or so years.

I'm giving three consective example of the FH response as I type in a Place field, with the FH response in "", and my comments in (). The examples were generated on a newly-loaded FH.

Example 1

w => "Wakefield (Ardsley), WRY" (nearly the first alphabetically "W" place)
wo => "Wolstanton, STS" (nearly the first "Wo" place)
wor => "Worksop (Anston), WRY" (the first "Wor" place)
wort => "Wortley (Clayton West), WRY" (the most popular "Wortley (" place)
wortley, => "Wortley, WRY" (the most popular "Wortley" place, and what I'd wanted)

Example 2

y => "Yeovil, SOM" (nearly the first "Y" place)
yo => "York (Escrick), WRY" (no idea why this one)
york (c => "York Copmanthorpe), WRY" (what I'd wanted)

Example 3

w => "Wakefield (Ardsley), WRY" (as before)
wo => "Wolverhampton, STS" (what!?)
wor => "Worksop, NTT" (what!??)
wort => "Wortley (Cawthorne), WRY" (what!???)
wortley (s => "Wortley (Stocksbridge), WRY" (what!???? this is the smallest "Wortley (S" entry)
wortley (t => "Wortley (Thurgoland), WRY" (what!????? this is neither the largest nor smallest "Wortley (T" entry)

Sigh of despair

[AdrianBruce]

Maybe I should amend what I said - it's always a bit wonky. I just tried entering my first place after opening FH 6.2.1 - at "O" it offered me "Oakland, California, USA" which is the 3rd in the alpha list and frequent but not that frequent. At "Ov" I got "Over, Cheshire, England", which is what I wanted. It has improved, since under the alpha first rule, I believe that I would have got "Over (reg. district), Cheshire, England" next, as the space after the "r" collates earlier than the comma.

I can cope with "Ov" since my eyes are on the keyboard as I'm typing so one letter or so makes no difference. But, from a point of view of conformance to the (presumed) specification - it doesn't.

The initial place-name offered doesn't seem quite consistent - 2nd, 3rd, even 6th in the alpha list. Whether there is a specific object count that has Oakland as the most numerous "O" place in my data, I've no idea.

[tatewise]

Just as a reminder:

• The same strategy should apply to all auto-completion fields such as Place, Address, Occupation, Religion, etc.
• The most-frequently-used strategy applies immediately after opening FH until names of the same type have been entered.
• The most-recently-entered strategy applies during one FH session, and takes precedence over most-frequently-used names.
• The usage counts should be those listed in Tools > Work with Data and in Places tab of Records Window of FH V6.

In FH V5 this all worked perfectly, but in FH V6.0 the most-frequently-used strategy became 'broken' for Place names, and is now 'better' in V6.2.1 but still incorrect and different from V5.

[AdrianBruce]

I've just been trying to get some insight as to what the algorithm might be - and failing. NB after each test I close down FH to avoid resetting the behaviour to "most used".

I said above that at the 2 letter stage, I got "Over, Cheshire, England", which is what I wanted - the usage count triumphs. Good.

But if I type "Cr", I get "Crewe (reg. district), Cheshire, England" instead of "Crewe, Cheshire, England" - the alpha collating sequence triumphs, assuming that the first space in the one I get, beats the comma, as I believe it does.

So one chooses on usage, the other on alpha. Confused?

"Crewe, Cheshire, England" does beat "Crewe (reg. district), Cheshire, England" on usage count at the record level of all types and at the event / where-used count level....

I've gone to the places tab first to force a count calculation - and still "Cr" gives the wrong answer.

I'm sure it's better but, as Mike says, it's still not doing what it should and I'm afraid I can't come up with any theory where it might be going wrong.

[tatewise]

Calico Pie admit there was a mistake in V6.2.1 which appears to have been fixed in V6.2.2 unless you guys can prove otherwise!

[RogerF]

I'm not (really; well ok, yes I am) having a go at CP here, while reflecting on recent history.

As far as I can recollect, FH version 6 was (pre-) released around December 2014 (that is, 18 months ago). Almost immediately, problems with autofill on Places were reported; the response was a resounding silence. Despite consistently vocal complaints that FH6 had broken the autofill model and caused significant inconvenience to long-time users, CP did exactly nothing to address those complaints until 6.2.1, released about a week ago. And 6.2.1 did virtually zippo to deal with the problem; 6.2.2 looks to be more promising...

My question are these:

1. why did it take so long to get ANY response to the problem, given 18 months of detailed complaint of EXACTLY what had gone wrong;

2. why did the 6.2.1 release not FIX the problem? There's hundreds of us out here who could have happily reviewed a 6.2.1 pre-release and phoned in the bad news; why didn't CP test-drive 6.2.1 on trusted users like Mike and Adrian?

IMHO, the whole issue has been a classic mismatch between developer aspirations and user requirements. Please let some lessons be learnt?

[AdrianBruce]

I can confirm that I now get "Over, Cheshire, England" (correctly) on typing the first letter as "O" and "Crewe, Cheshire, England" (correctly) on typing the first letter as "C".

So both my obvious examples auto-fill OK, thanks.

[tatewise]

I tend to sympathise with Roger.
The problem was clearly introduced in FH V6 Beta due to Place Records upsetting the usage count mechanism, even though the Tools > Work with Data > Places > Usage count was correct.
Despite being reported then, and repeatedly since, it has taken far too long to make what must have been a relatively simple correction.