Class SnarfPacker

Class Methods

Overrides: DiskManager class

Overrides: DiskManager class

Instance Methods

Add another flockInfo object to myFlocks with info about another location for shep.

Find a snarf in which to fit shep. Then assign it to
that location, and mark that snarf as changed.

Overrides: DiskManager

Perform the sanity check of the moment. Beware the compile cost of changing this comment.

Used by ResetCommit bomb

Commit by destroying the current view and creating a new one.

Send the snarf over a transmitter into a stream that just counts the bytes put
into it.

Overrides: DiskManager

Return the current urdiView.

Destroy all forgotten flocks that are no longer in memory.

Queue destroy of the given flock. The destroy will happen later.

Overrides: DiskManager

Destroy all objects imaged from this snarf.

Overrides: DiskManager

Overrides: DiskManager

Turn the flock designated by info into a Pumpkin. It should have completed all dismantle actions.

Overrides: DiskManager

The flock is being removed from memory. For now, this is an error
if the flock has been updated. If the flock has been forgotten,
then it will be dismantled when next it comes in from disk.
Because of forwarding, there may be many FlockInfos refering
to the flock if it is not new.

Overrides: DiskManager

Overrides: DiskManager

If something is already imaged at that location, then return it. If there is already
an existing stub with the same hash at a different location, follow them till we
know that they are actually different objects.

Overrides: DiskManager

Return true if the object is on disk but not in memory.

Actually write all the changed and newly assigned flocks to the disk.

Remember that there are no more persistent pointers to the shepherd
represented by info. If it gets manually deleted, dismantle it immediately.
If it gets garbage collected, remember to dismantle it when it comes back
in from the disk.

Overrides: DiskManager

Return the set of indices to flocks in snarf snarfID that are forgotten.

Shep has grown too large for its current place. Treat it as just a new flock and give it another place.

Return a flock at a particular location. This needs to register
the flock if it doesn't exist already.

Return the starting object for the entire backend. This will be the 0th
flock in the first snarf following the snarfInfo tables.

Overrides: DiskManager

Get the read handler on the snarf.

Return the snarfRecord for snarfID. The SnarfRecord must exist if there are
changed flocks imaged out of that snarf, but might not otherwise. Create it if necessary.

Overrides: DiskManager

Overrides: DiskManager

The virtual image in memory is now in a consistent state. Write the image of
all changed or new Shepherds out to the disk in a single atomic action. The
atomicity only happens on top of a real Urdi, however.

Retrieve from the disk the flock at index within the specified snarf. Since
stubs are canonical, and this only gets called by stubs, the existing stub will
*become* the shepherd for the flock.

Overrides: DiskManager

Shepherds use a sequence number for their hash. Return the next one
and increment. This should actually spread the hashes.

Overrides: DiskManager

Flush everything out to disk and remove all purgeable imaged
objects from memory.

Overrides: DiskManager

purge all shepherds that are currently clean, not locked, not dirty,
and purgeable. Purging just turns them into stubs, freeing all their
flocks. Garbage collection can clean up the flocks and any stubs no
longer pointed to by something in memory. Because infos for new
flocks don't appear in myFlocks, this will not throw out any newFlocks
(which will be marked dirty anyway). For each FlockInfo, we check
that its flock refers to that exact instance to get correct information
about its dirty state.

Overrides: DiskManager

The flock represented by info has changed. Record it in the
bookkeeping data-structures. This must be called by all things
that affect whether the flock gets rewritten to disk.

Make sure all flocks that have changed still fit in their snarfs.
Add any that don't to myNewFlocks and return the table
from their current locations to the newShepherds.

Overrides: DiskManager

Release the supplied snarfHandler and destroy it.

There are now persistent pointers to the shepherd help by info.

Overrides: DiskManager

Make sure that the shepherd or stub at that location actually points
at the real location for a shepherd. This will resolve forwarding pointers,
but not instantiate any flocks.

Overrides: DiskManager

Do nothing

Overrides: DiskManager

A turtle just got created! Write out a pseudo-forwarder that has all the protocol information encoded in the snarfID and index.

Overrides: DiskManager

Shep just got created! On some later commit, assign it to a snarf
and write it to the disk.

Overrides: DiskManager