I've looked through all contributions to the docs. Any contributions not by
Thomas Adamcik or myself are trivial (e.g. typo fixes) or have mostly been
replaced by later changes. Thus I'm comfortable with changing the license with
the agreement of Thomas Adamcik.
By changing the documentation license to be the same as the source code
license, we avoid ambiguities around the license of e.g. docstrings which are
written as part of the source code but are included in the generated docs.
I've looked through all contributions to the docs. Any contributions not by
Thomas Adamcik or myself are trivial (e.g. typo fixes) or have mostly been
replaced by later changes. Thus I'm comfortable with changing the license with
the agreement of Thomas Adamcik.
By changing the documentation license to be the same as the source code
license, we avoid ambiguities around the license of e.g. docstrings which are
written as part of the source code but are included in the generated docs.
If we accept unicode and try to encode using sys.getfilesystemencoding() then
it may work most of the time, but will fail if we get non-ASCII chars in the
unicode string and the file system encoding is e.g. ANSI-something because the
locale is C. Thus, I figure it is better to always fail if we try to serialize
Path from unicode strings. Paths should be maintained as bytes all the time.
This version of the scanner switches to using the updater API, in other words
we are now fairly close to being able to plug in alternate libraries for local
files and populate them with mopidy-scan.
Idea behind this class is to have a clear sub-set of our library API that is
only used for updating libraries. This makes it very clear what methods are
needed for just libaries and additionally updatable libraries.
Next couple of commits will build on this idea taking us closer to plugable
local libraries.
Tag cache is now output to a tmp file residing in the same folder as the real
one. Once generated the tmpfile is moved over the original file for an atomic
updated.
