On Mar 14, 2011, at 16:09 , Greg Von Kuster wrote:
On Mar 14, 2011, at 10:52 AM, Assaf Gordon wrote:
Do you really want to delete all the user's records from the database ? I think the database size is tiny compared to the actual files on disk, keeping all database records forever shouldn't be such a problem.
You're right, of course. There would be no problem with keeping the user's credentials and other records.
Regarding files, It's my understanding (galaxy people, correct me if I'm wrong), that once a dataset is marked as "deleted" in "history_dataset_association" table, it's as if the user deleted the dataset by himself.
So if you run a query that sets "deleted=true", the galaxy clean-up scripts will take it from there and will eventually delete the dataset ("eventually", because there are couple of clean up steps and scripts).
For the above, I recommentd the approach discussed in my previous email ( uncomment the Delete / Undelete / Purge operation buttons in the admin controller. This will do what you want, and you'll not need to execute any sql commands manually, which can be very dangerous.
Yes, I believe that's how we can do it. But there should be some kind of feedback for a user when this is about to happen... and a way to prevent it.
This is actually something I would like to request as a feature: "reproducibility" doesn't require all the files, all the time - only the first file (let say: a FASTQ file) and the meta-data for downstream files (jobs, tools, parameters) are needed.
For the above, determining whether a dataset is shared is currently available, but what is your definition of a "published dataset"?
A history item that is used on a Galaxy page, as supplementary material to a publication, for example, or a history that is listed on the "published histories" of Galaxy.
It would be great if there was a way for users to see the datasets (and the jobs, parameters, etc.) of all their datasets (ever), even if I deleted the underlying physical file.
Where are you talking about "seeing the datasets, jobs, parameters, etc" whose underlying file has been removed from disk? Would this be in the history, where you can currently see "deleted" datasets, but not "purged" datasets?
In an ideal world, we would be able to eventually remove all the intermediates and keep only the initial dataset from which all the results were generated. For that to work 100%, you would need a snapshot of the galaxy installation with all the tools and their versions at the time of execution... it sounds very messy to implement and test, though. The dataset without the physical file could get its own color and all analysis steps could be re-run transparently if someone really wanted to access a specific intermediate file. Thanks for discussing this with us! -- Sebastian