project autosave?

[prefetch]

it seems that there is no project autosave feature for .xosm files and associated data.  this seems like a trivial feature and it's strange to me that it doesn't exist.  i've lost progress more than once on random crashes over the last six months.

why doesn't this exist?

[georg.viehoever]

Probably because such a save can take considerable time, and doing it in parallel with GUI work is technically difficult.
Georg

[prefetch]

>Probably because such a save can take considerable time, and doing it in parallel with GUI work is technically difficult.

not be 'that guy' but i'm a software engineer - and PI is incredibly impressive software, and autosave is not technically difficult.  when i manually save it takes about a second of time, and incremental saves should take less and of course multithreading disk i/o in the background is not difficult.

even so, if even small interrupts were not desirable, it'd be relatively simple to check for idleness, and save if/when the user is idle.  i think a lot of people take breaks or leave PI open on their systems for extended periods of time.

[RickS]

not be 'that guy' but i'm a software engineer - and PI is incredibly impressive software, and autosave is not technically difficult.  when i manually save it takes about a second of time, and incremental saves should take less and of course multithreading disk i/o in the background is not difficult.

even so, if even small interrupts were not desirable, it'd be relatively simple to check for idleness, and save if/when the user is idle.  i think a lot of people take breaks or leave PI open on their systems for extended periods of time.

You must have small projects.  Some of mine take several minutes to load or save.  Incremental saves are a nice idea but I'm sure the dev team has no shortage of nice ideas that they don't have the bandwidth to implement 

Cheers,
Rick.

[prefetch]

>Some of mine take several minutes to load or save.

whoa!  this seems crazy to me - maybe i'm not using PI right!  i typically have a handful of saved processes (which are tiny) and maybe a half dozen or so images that i'm doing some kind of processing on, and of course sometimes hundreds of large images on disk, but these are simply references and not actual PI data that would be saved.

could you tell me more about the projects you are working on?  i assume if it does actually take several minutes to save then you must have hundreds of images (gigabytes of data) that you are editing in PI concurrently.  i'm still pretty new to PI and my use cases might be pretty entry level.

[pfile]

it's not so much the # of images (though that is of course a factor) but the processing history of each image.

PI is storing a copy of the image for each processing step that is in the history. so if you have a 32b, 3-plane image from even a smallish sensor like the 8300M, there's about 100MB per step. 10 steps gets you a gigabyte.

while PI is running, these history images are in whatever swap directory you have defined. the process of loading a project back into PI means that all of the processing histories of every image are copied from the project container back into the swap directory. that can take a while if there is a lot of processing history per image.

an unfinished 3x2 mosaic project i have on disk is 25GB.

rob

[prefetch]

ah.  thanks rob.  that makes sense now. i've never done a mosaic before, and about half of my post processing is done in photoshop.  i also use a pretty fast solid state disk - this explains my relatively small project sizes and fast project save times.

i guess differential autosaves would really be a requirement to be useful.

okay, thanks for the discussion everyone - i think my question has been answered!