Better Auto Save

Report · Jan 27, 2018

I've been working on a project in DaVinci Resolve 14.2 trying to learn the software. Today I had my first crash ever with Resolve and naturally my first thought was "Damn, how much of that work am I going to have to redo?"

The answer...None! The project opened up exactly as it was when the software crashed, even though I hadn't saved the project at all during the editing session. How is this possible, you ask? How can a project retain all my hard work even without saving?

In Resolve the feature is called Live Save. Put simply, Resolve saves every change you make to the hard drive instantly. There's no need to ever manually save the project because everything you do is automatically saved in the background.

Let's finally get this done for Premiere Pro, eh?

FR: Instant AutoSave

Report · Feb 04, 2018

Sounds like a good idea Jim, I saw the other thread, will put in a feature request.

Check out the FILE/Save command in Adobe Prelude. Perhaps we don't have to go that far, but continuous save seems like a good way.

Report · Mar 13, 2018

I agree it's a hugely important feature, but Resolve can do it because everything about a project is stored in a disk-based or socket-based PostgreSQL database (like Lightroom does with SQLite, hence no need for a "save catalog" button). Updating the project data store in Resolve is a very fast INSERT ROW, and requires a few hundred bytes of disk traffic. In Premiere everything is neatly packed into the often-quite-chunky PRPROJ file, and re-writing that every time you click a checkbox would cripple performance.

The database approach is also why Resolve can support almost unlimited real-time collaboration as standard, and that's a big selling point for studio pipelines. I could imagine a day when Premiere switches to database storage for that reason alone.

The drawback is that the database can itself be eaten by the logic weevils, and then you can lose all your projects in one go. Keeping (non-incremental) backups is vitally important, and in Resolve they're a bit more obscure for the average Joe to set up because it's not as obvious where the data lives.

Report · Mar 13, 2018

Resolve can do it because everything about a project is stored in a disk-based or socket-based PostgreSQL database

That's an option. I don't use it.

Additionally, Lightworks has had the feature for years.

We can either accept excuses for Adobe not doing it, or keep requesting they get it done, whatever it takes.

I'm in favor or the latter. You?

Report · Mar 13, 2018

Jim_Simon wrote
That's an option. I don't use it.

The option is whether to install PostgreSQL as a system service and talk to it through sockets, or whether Resolve writes directly to disk (using a folder tree of small XML files). The abstracted data being written is the same, which is why you can switch between modes. If you ask Resolve to back up the database, putting everything into a single compressed file, it can take a while - but updating the store when you edit a grade or re-cut a clip is trivial, as you only need to touch the XML file that refers to the item in question. The file structure is designed for speed over size and portability - to all intents and purposes it's instantaneous. In contrast a PRPROJ file, although it's XML data under the hood, would have to be written to disk in its entirety even if all you did was toggle a checkbox. Some PRPROJ files are quite small on disk, but that's because they are GZipped, which eats up CPU cycles during the save operation. A 100kb PRPROJ file has around 25,000 lines of XML, compressed 10:1. On a fast machine the compress-and-save loop is quite fast, but imagine how bad Premiere would be to use if it paused for half a second every time you clicked something.

Report · Nov 11, 2019

Thats very clever indeed, saving all the changes to a database, and you're right, all projects are at risk from a single database corruption or failure. That said, the combination of file saving as well as database runtime changes would be the best way to go. Editing and saving to the database at runtime, and at the autosave time limit reached, write the autosave info to the normal project file. -Best of both worlds, with of course the expense of doubling the amount of disk space needed for saving a single project, but I'm sure that if this combination of autosave could be implemented we would not really be fussed about the extra disk space we use. 🙂

Report · Nov 11, 2019

-