Fixed activity type model updates stalling indefinitely when many models need updating simultaneously (e.g., after a flight covering many geographic regions), by limiting concurrent model updates (BIG-352)
Fixed app hanging when scanning for available backups during onboarding, by moving file system operations off the main thread (BIG-346)
Fixed app hanging when opening context menus on timeline items, caused by a synchronous system check for installed map apps (BIG-349)
Fixed crash when viewing altitude charts for locations below sea level (BIG-350)
Fixed crash when saving a place edit in certain navigation states (BIG-351)
Fixed navigation not working after saving a new private place created from search results (BIG-354)
Iāve been ignoring the betas but now I fancy a bit of excitement . How exciting is it likely to be? Is there a significant risk of data loss? Is there a summary somewhere of how the different apps relate to each other and how the new app changes things?
No rush because I appreciate the work being done as-is, but do you have a sense of when youāll have the activity maps working?
Not a huge deal but I like to occasionally look at my progress for the year on the map and the place dots can obscure some interesting stuff at some zoom levels. On the attached picture Iāve spent some time walking/biking at various places along the train line, but since I spend so much time at different stations on the line, itās mostly obscured at a glance by the places.
Really cool visual most of the time, but sometimes I just like to see how Iāve been moving and nothing else
For a full rundown of how Arc Editor relates to Arc Timeline, the initial forum thread goes into detail: First public beta for Arc Editor!
Though the short version is: Thereās zero risk, zero data loss potential, and the new app is better than the old app in almost all ways. Itās also very close to being final - Iām aiming to ship it to the App Store proper ideally before the end of this month.
The new app doesnāt yet have all of the features of the old app - Iām adding them back in in priority order of which ones people mention theyāre missing the most. But itās got enough in there now to be worth switching to entirely. At this point I basically only keep the old app around because itās my job. Otherwise Iād have deleted it already The new app is so much better!
Oh, though definitely keep the old app around for a little while at least. Thereās no need to rush to deleting it or turning it off. The two will run happily side by side - they record to separate databases.
Now that youāve asked, Iāll up the priority of that task Iām prioritising adding features from the old app based on how often they get asked for.
Though for what youāre askingā¦ Iām thinking the old per-activity map views in the old app arenāt necessarily exactly right. I mean, theyād do the job, but the fact that theyāre only per-activity is a bit limiting.
So Iām almost thinking a better fit would be the feature idea of ā¦ map toggles. Like, when the content sheet is slid down there could be some tools/toggles visible on the map that let you turn on and off different kinds of map content, to filter the view to just the kinds of things you want to see.
Iāll make sure Iāve got the right tasks filed in the task tracking system, so that Iām tracking all this and prioritising. But yeah for the per-activity views, Iām loosely thinking those should be in the next major update, so v1.1. And the map toggles feature ideaā¦ Iād like to get that into 1.1 also, but weāll have to see how things play out.
Iāve taken the plunge and itās gone well! My only observation so far of missing something is that it doesnāt tell you the āsourceā of a place, for example Google or Squarespace. Unless Iām missing somethingā¦ I like to ensure frequently visited places are from Google, Iām not really sure why I do that!
I really like that map toggle idea. Another super minor thing for me, but with that, if youād be able to toggle certain activities, it would solve the messy airplane line aesthetic.
This update sounds like you fixed a lot of the hang issues; thank you!
I havenāt managed to reproduce the full UI hang with this build! So thatās definite progress.
But when I update a big activity (like, as you say, a long flight), it still kicks off a ton of simultaneous updates that effectively prevent most of the UI from working.
At the moment I have 191 active (3 timeline, 188 activityTypes) running for 10+ minutes, and thus most of the UI doesnāt work (e.g. maps are blank, activity types canāt be selected, etc.)
Hi @elucubrator! No need to worry. The āpurchasesā in the TestFlight builds are fake - all done in the āApp Store sandboxā, so no real money is involved.
Also the subscriptions for the old app are meant to carry over to the new. No separate purchases or subscriptions required. But it looks like weāve just spotted a bug in that carry over, which will need fixing.
But yeah, regardless of that bug, you can just tap through to do a fake lifetime purchase in the TestFlight build and itāll stop bugging you, and no money will be spent
Are you meaning the Place Merge feature? For when thereās things like āBobās Cafeā and āBobs Cafeā and theyāre the same thing but listed separately?
If so, not yet! But now that youāve asked Iāll up the priority of that feature request. I think itās slated to land in v1.1, so the first major update after the initial v1.0 release (which is targeted for before the end of this month).
Hm. Thatāsā¦ weird. The fix in the last build was to throttle concurrent model updates to a max of 3. So almost all of those 188 must beā¦ classification tasks, presumably. Curious.
Classification tasks are also deduplicated, soā¦ it must be a lot of different classifications. I wouldāve thought thatād be a quite rare situation to get into, but clearly you got into it! So itās not impossible, and possibly easier to get to than I thought.
Did you happen to notice what kinds of activityTypes operations they were? Although if I recall right, there is only two: model updates, and classification. But maybe Iām forgetting some others. Iād better checkā¦
Eek. Well thatās a case of āshould be impossibleā. But there it is, screenshot is build 43, and thereās >3 updateModel() operations. Hmm.
There must be another pathway for those to get in there. Though I thought we put the 3 concurrents check at the singular funnel they all have to go through. Clearly I missed something there.
This might be obvious from earlier in the thread, but for the avoidance of doubt: it seems to be quite reliably kicked off by doing something like setting the activity type of a flight.
Yeah I figured thatād be the worst case scenario. Those cover potentially a whole lot of D2 regions.
Oh for interestās sake: Not sure if youāve picked up on how the models work or not, but itās kinda fun. Thereās three depths: D2, D1, D0, and classification makes use of potentially all three overlapping.
D0 is a single model for the whole world, D1 is city/state sized regions, and D2 is neighbourhood sized regions. So D2 does most of the work, as long as itās already fleshed out with enough data for that region. If not, then classification is weighted to fall back more on the D1. And if even thatās not well fleshed out, weighting will instead lean most heavily on the global D0.
So when you have a flight thatāll cover a whole bunch of D2 regions, thus why it triggers so many model updates. It will usually also cover a handful of D1 regions, depending on how far you flew.
In the Debug Info view, under Activity Classifiers, you can see which ones are currently being used for most recent classification. The C numbers are ācompletenessā and A numbers are āaccuracyā.
If the D2 is already at 1.0 completeness then it takes the full weight and the D1 and D0 arenāt engaged at all.
The BD0 is a fallback ābootstrapā model, for new users with little to no data of their own yet. Itās basically my own CD0 but with the types grouped more simply, to be more generally applicable.
Anyway, tangent aside: hopefully Iāll have a new build this afternoon! This gap that these updates are slipping through shouldnāt be too hard to findā¦
@gsoper The Place Merge feature is slated for the first release after v1.0 goes live. Itās definitely high priority! I want it too. But Iāve had to be brutal about limiting scope for the initial 1.0 release, to make sure I can actually get it shipped this month.
But yeah, if all goes to plan it should be landing within weeks of the 1.0 release. Which should land before the end of this month
. How exciting is it likely to be? Is there a significant risk of data loss? Is there a summary somewhere of how the different apps relate to each other and how the new app changes things?
The new app is so much better!
