Single user editing and commit feature

Hi all,

I need to develop an Rails app where all changes should be done on a
“shadow copy” of the data and then committed all at once. Only one user
may edit the data at any given time. Other users should not see the
changes until the commit is done. Before commiting the changes, the user
should be able to review a summary of the changes made since the editing
began.

These are basically ACID requirements, but the fact that the changes
will span multiple HTTP requests -and potentially multiple editing
sessions spread out over several days - I’m not sure that database
transactions are the best solution (or even feasible)

There is not a huge amount of data involved, so copying the entire
database when a user enters edit mode is a possibility.

Does anybody have any input on this?

On 15 Apr 2008, at 10:53, Abigail Headroom wrote:

editing
began.

Perhaps acts_as_versioned could be subverted for this ?

Fred

I am assuming that ALL data needs to be locked. That is, every row in
every table is locked.

A simple way to do it would be to have a check_outs table. It would
only ever have one row, and it would have a user_id column.

If nobody had anything checked out, then user_id would be nil.

Upon checkout the current user’s user_id would be stored in this
table.

You can then add a validation to every model (except the CheckOut
model) that disallows the model from being updated except when
current_user.id == CheckOut.user_id.

Does that make sense?

Paul

On Apr 15, 2:53 am, Abigail Headroom <rails-mailing-l…@andreas-

Thank you both for helping out, I really appreciate it.

Fred: I think acts_as_versioned could be helpful in a couple of ways
here. It could implement the snapshot which the read-only users will see
while someone else is in edit mode. Basically it’s just a matter of
storing a list of all object types and a corresponding version number.

The revision history per object which acts_as_versioned provides can
perhaps be used to keep track of changes made by a user since entering
the edit session. Finally, the revert_to method could be used to
rollback changes if the user in edit mode decides not to commit
anything.

Paul: Yes, all data needs to be locked. The check_outs table is a good
idea for implementing the concurrency lock, I’ll probably use it!

I would also need acts_as_paranoid to be able to rollback deletes, and
by the looks of it, also do some hacking of both these plugins:
http://ward.vandewege.net/blog/2007/02/27/116/

The more I think about it, this journalling alternative feels kinda
messy and risky. I would really prefer a solution where edits didn’t
affect the real data until the user commits.

The extreme case scares me: the user enters edit mode, deletes
everything and creates a completely different set of objects, and
finally decides not to commit… :-/

Another motivator for another solution (which I accidently omitted
yesterday) is that external systems will read the data using SQL and I
really don’t want them to be aware of any journalling and snapshot
stuff.

Does anybody have any alternative approaches to this problem? Just
thinking out loud - would it be possible to temporarily “short circuit”
AR to only operate on in-memory objects for example? (The changes made
during an edit session can be considered expendable until committed)

On 21 Apr 2008, at 07:50, Abigail Headroom wrote:

I’m giving this thread a last chance. I feel stuck and would really
appreciate any hints on how to solve this.

Don’t save changes at all, just persist them in something like
memcached until ready? it’s definitely not trivial (for example do you
need to worry if the changes made during the edit session don’t apply
any more (eg user edits records a,b,c. while they are editing b gets
deleted. do you still apply remaining changes to a & c ?)

Fred

I’m giving this thread a last chance. I feel stuck and would really
appreciate any hints on how to solve this.

Basically what I’m looking for is the semantics of a full database copy
for the duration of the edit session, without having to, ermm… copy the
database.

Frederick C. wrote:

Don’t save changes at all, just persist them in something like
memcached until ready?

Thanks again for your input, Fred. Yeah… Working with in-memory objects
during the edit session is probably the way to go. I’ll make an attempt
to overload ActiveRecord save() etc to see what happens.

do you
need to worry if the changes made during the edit session don’t apply
any more (eg user edits records a,b,c. while they are editing b gets
deleted. do you still apply remaining changes to a & c ?)

No, I don’t have to deal with that since two users cannot be in edit
mode simultaneously.

This forum is not affiliated to the Ruby language, Ruby on Rails framework, nor any Ruby applications discussed here.

| Privacy Policy | Terms of Service | Remote Ruby Jobs