Larger Library Challenges about icloud-photos-sync

[steilerDev]

Thanks for the feedback - some points from my POV:

1 - Already discussed :)

2 - Great point, will add this to the docs!

3 - The sync error is expected on long downloads - the data initially loaded is invalidated by Apple after about an hour, which requires a metadata reload. This could be done more gracefully but dying and starting should work and it's simple (see point 5 below). However what should not happen is a locking error, something is wrong in that logic and this should be a separate bug

4 - There are some edge cases in which axios and the download process can die - there are some I'm not catching till now

3, 4 and 5 are somewhat related - because I actually tried to design the program in a way, that it would gracefully recover after a sync error. It should load everything that is on disk and compare it to what's in the cloud and make a decision about what to keep and what to remove based on this - so the logs where you are seeing a full wipe would be helpful, since this should not happen :)

6 - Actually the _All-Photos folder is only the storage of the assets, after they are downloaded, the tool will also replicate the folder structure of your albums and restore the file names in this view:

This however only happens after an initial asset sync - the reason why I'm not using the actual names, but the IDs created by Apple is that one asset can be linked in multiple folders (so it would be space inefficient to store them in folders which is why I'm linking them) and there can be equally named files in the library (so when I store them all in a central place there might be a collision if I use the original name) - so I'm going for Apple's UUID

[steilerDev]

I've added documentation around increasing file limits - the other issues are currently hard to investigate for me (especially loosing track because this ticket includes multiple of them).

If you could open a separate issue/discussion for the remaining challenging issues, that would be highly appreciated (please include logs and error codes there).

The locking behaviour is something I'm looking into at the moment.

I'll be closing this for now - but don't hesitate to open a new issue for each of your problems - check out my brand new contributing guidelines in case you are unsure what to do!

[github-actions]

This issue should be resolved with version v1.1.1-beta.2, please confirm.

[lonevvolf]

So now that I've finally sync'ed my 300k+ elements, I definitely wanted to comment on this item. Especially the feedback:

Definitely in the enhancement area: would be nice to have the downloaded assets not dropped into a single directory with random character names of the files. Instead, would be nice to have a hierarchy of folders based on picture date with file names what they are in the underlying library similar to how icloudpd works. I was utilizing icloudpd but the project has largely stalled and the PRs for shared library support thus far haven't been successful. As almost all of our assets are in our share library, it lead me to this project. I do appreciate the objectives of a dedicated photo downloader, etc. that you've enumerated and appreciate your fine work. I also appreciate the challenges of hacking through iCloud website api to figure this stuff out.

100% this would be a great feature, even for smaller libraries. I don't really make use of Albums, but rather browse by dates (Year, Month type of folder structure). This is how the Photos organizes itself, and it's pretty comfortable for browsing old memories. The Albums strategy is ok if that's part of your main workflow, but unfortunately doesn't even work for large libraries apparently (all Albums are empty and have a long ID number instead of a name). I am extremely happy that I finally have a backup of the files, but would love to see some automatic organization applied.

But I did just want to add a huge thanks to @steilerDev - this tool downloaded my entire library in just a few days, compared to my Mac with Photos, working on this task for a few months in the native Photos app.

[steilerDev]

[B]ut unfortunately doesn't even work for large libraries apparently (all Albums are empty and have a long ID number instead of a name)

This sounds like a bug, unless all your items are part of the shared library OR you can't see symlinks through your network share, check out this discussion.

This is how it is supposed to look like:

I'm trying to explain the local file structure in here - I should add some more information around the goals of the design process.

We could 'hide' the folder that is actually holding all the data and create an 'All Photos' meta album that uses date based formatting (and links to the assets). This would add (imho unnecessary) additional logic and complexity, because you should also be able to sort the 'All Photos' album by modified time and browser the pictures in the same order as on the phone (as I'm reading and applying this metadata) - file names would still be ugly, however is AUR0lYIXVYq9SuZQ3OGBqbpEkY9P.arw really worse than IMG_1181.arw?

[steilerDev]

Having a better way to access the library (through a web based application) is something that I've already though about - however this is probably the lowest priority work item at the moment.

Unless someone from the community (with maybe some experience in web technology) wants to take a stab at it, I would not hold my breath for getting this done anytime soon though.

[lonevvolf]

This sounds like a bug, unless all your items are part of the shared library OR you can't see symlinks through your network share, #161.

Ok, that was a good tip - I am indeed running this in a Docker container on Synology. I found out that the file browsing app on the Synology (File Station) doesn't show the links, but they are actually there when browsing in bash or mounting the SMB share. So the album names and links are ok.

However, all my albums are empty due to the
ICLOUD_PHOTOS_COUNT_MISMATCH (WARN): Received unexpected amount of records (expected 55 CPLMaster & 55 CPLAsset records, but got 0 CPLMaster & 0 CPLAsset records for album BA42E172-2EE6-47C9-8344-333F0EAF48AE)
errors (see #231 and common warnings).

Nonetheless, the automatic year/month folder sorting (ie. 2020/05) would be much appreciated. Can certainly also be done with symbolic links, though there's really no issue in that case with duplicate pointers to files as each can be allocated to a single folder.

We could 'hide' the folder that is actually holding all the data and create an 'All Photos' meta album that uses date based formatting (and links to the assets).

This could work, but no need to hide the original folder I suppose. If people want to see it, that's fine.

file names would still be ugly, however is AUR0lYIXVYq9SuZQ3OGBqbpEkY9P.arw really worse than IMG_1181.arw

Not a major issue for me - agree that a nonsense name is a nonsense name. :)

Having a better way to access the library (through a web based application) #120 - however this is probably the lowest priority work item at the moment.

This is not really something needed from my end at least.