The initial idea for this refactor was motivated by the need to make
PluginQuery.match() have the same method signature as the match() methods on
other queries. That is, it needed to take an *item*, not the pattern and
value. (The pattern is supplied when the query is constructed.) So it made
sense to move the value-to-pattern code to a class method.
But then I realized that all the other FieldQuery subclasses needed to do
essentially the same thing. So I eliminated PluginQuery altogether and
refactored FieldQuery to subsume its functionality. I then changed all the
other FieldQuery subclasses to conform to the same pattern.
This has the side effect of allowing different kinds of queries (even
non-field queries) down the road.
In an attempt to finally address the longstanding SQLite locking issues, I'm
introducing a way to explicitly, lexically scope transactions. The Transaction
class is a context manager that always fully fetches after SELECTs and
automatically commits on exit. No direct access to the library is allowed, so
all changes will eventually be committed and all queries will be completed. This
will also provide a debugging mechanism to show where concurrent transactions
are beginning and ending.
To support composition (transaction reentrancy), an internal, per-Library stack
of transactions is maintained. Commits only happen when the outermost
transaction exits. This means that, while it's possible to introduce atomicity
bugs by invoking Library methods outside of a transaction, you can conveniently
call them *without* a currently-active transaction to get a single atomic
action.
Note that this "transaction stack" concepts assumes a single Library object per
thread. Because we need to duplicate Library objects for concurrent access due
to sqlite3 limitation already, this is fine for now. Later, the interface should
provide one transaction stack per thread for shared Library objects.
As part of this, the BaseLibrary class was also adapted to include a notion of
albums. This is reflected by the new BaseAlbum class, which the Album class
(formerly _AlbumInfo) completely replaces in the concrete Library. The BaseAlbum
class just fetches metadata from the underlying items.
current metadata to be correct if it's complete
Previously, we were using the Munkres algorithm (minimum bipartite matching) to
order tracks intelligently only as a fallback if the current metadata was
paradoxical or incomplete. This was because of a concern about the performance
of the potentially-O(n^3) Munkres solver. However, it was found that (a) the
performance is actually not bad, taking on the order of 0.02 to perform a
matching, and (b) there was no recourse for the tagger to reorder tracks that
were legitimately in the wrong order. Now, we get intelligent reordering of
badly tagged music even when the metadata seems to be complete.
To retain some of the functionality of the old orderer, the track distance
metric was expanded to include a component reflecting the track index.
In doing this, another bug was discovered in the UI that showed the track name
differences based on an arbitrary ordering. Now, the tag_album function returns
a reordered items list with every candidate.
Also, new organization for tests and automatic loader. Fixed bugs uncovered by new tests.
--HG--
extra : convert_revision : svn%3A41726ec3-264d-0410-9c23-a9f1637257cc/trunk%4069
