Fixes#1264: When a re-import triggers fetchart, the existing album art is no
longer resized if it already has the correct dimensions. This avoids creating
new album art files with unique filenames ("cover.n.jpg").
Because DBpedia seems to be down (it's responding very slowly with errors),
this seems like a good defensive action. It narrows the default sources to
those that resolve quickly, with only a couple of HTTP requests. We can
re-enable this again in the future if things seem to stabilize over at
DBpedia.
requests/urllib3 is throwing an exception due to an internal problem triggered
by some sort of timeout. This change catches the TypeError so that beets
reports "error fetching art" instead of crashing when this happens.
Artists with non-typical casing (e.g., alt-J, dEUS) would not get matched on
DBPedia, as the RDFS:label uses arbitrary casing, and SPARQL provides only exact
matches. The FOAF:name attribute is always title-cased (e.g., Alt-J, Deus).
Due to a bug in DBPedia, the cover filename is truncated when it contains
parentheses, (e.g., 'Foo bar (band).jpg' gets truncated to 'Foo bar .jpg').
To work around this, an additional Wikipedia call gets made for all its
images, in which we try to match our truncated image.
The Wikipedia art source now catches the correct exceptions, instead of
a broad catch-all.
Wikipedia album images can be gifs, so these are now added to the list of
accepted content types.
The `enforce_ratio` and `minwidth` options depend on PIL or ImageMagick.
Previously it silently fails. Now it will log a warning, and accept the
image.
Tests concerning these options are skipped when no imaging backend is available.
Fix#1460
Using -500 URLs for coverartarchive.org will only ever return images
where the biggest dimension is (width or height) is 500 pixels,
regardless of what fetchart settings are otherwise set.
This commit removes the -500 from the URL entirely rather than using it
conditionally, since a maxwidth of 500 will allow for a 600 high and 500
wide image, but CAA.org/...-500 would return a 500x417 image instead, so
not enforcing a size is the only way to ensure the user's {max,min}width
settings are properly respected.
- Minimum image width can be specified via minwidth (default `0`)
- The image ratio can be enforced to 1:1 using `enforce_ratio` (default `no`)
See #1058
E.g., `colorize('text_success', 'hello world')`
To ensure compatibility with 3rd party plugins, a valid color ('red') can still be passed,
but it will be logged.
- Colors are mapped on to a dictionary using abstract names (e.g., text_success)
- Add `colors` option under `ui` to allow users to choose their own color scheme
- Move configuration option `color` from top-level to `ui`
- Show deprecation warning if top-level `color` configuration is used (but respect it)
Fix#1238
Include import of __future__ features division, absolute_imports and
print_function everywhere. Don't add unicode_literals yet for it is
harder to convert.
Goal is smoothing the transition to python 3.
Breaking changes: plugins should set set _import_stages instead of
import_stages. From outside the latter is replaced by import_stages().
This is because it is now wrapped with log level-getting & -setting
statements.
I like the CAA as a first-priority search because the images are generally
high-quality and there's no metadata ambiguity (we always find the right
release if it's in the catalog).
* Ensure that `config.as_str_seq()` returns a list of unicode objects.
* Map these to bytestring paths so we can compare them to other paths.
Fixes#887
This reverts commit c56412a1d5.
- The Requests library advertises itself as thread safe.
- Cookies aren't so interesting, but keep-alive /
connection pooling would be nice to have.
This might have been perfectly correct, but I read that requests.Session
objects share cookies between requests. That makes me nervous for two reasons:
- thread safety (does requests lock the cookie jar internally? who knows?!?)
- leaking cookies could have privacy/performance implications; we don't need
'em anyway
