- Do not crash when a version information is missing in bower.json.
- A few documentation fixes.
- Allow content_type of response to be None, e.g. for HTTP-401.
- Python 3 compatibility fixes: a few things broke with 0.8.
Added references to Morepath and Pyramid extensions that integrate BowerStatic.
Don’t try to pick up any directories that start with a period (
.) as bower packages.
Added an example on how to integrate BowerStatic with a Flask application.
Fixed documentation describing WebOb integration to new reality as of 0.7.
Ability to support custom renderers:
Breaking change: renderers will receive the
resourceargument instead of the url. Renderers can be used to render resources inline.
The publisher and injector have been refactored into
Publisher, and an
Injector. The Tween versions are the ones intended for use by web frameworks that already use WebOb (such as Morepath and Pyramid) as an alternative to using WSGI-based integration.
The WSGI-based integration has remained unchanged; you can still use
- Added Python 3 compatibility. Fixes issue #25.
- Bower components and local components can be created based on a path
relative to the directory in which the components are created using
- The injector and the publisher can now also be used directly by providing a WebOb request and response. Third-party frameworks that already use WebOb (such as Morepath and Pyramid) can make use of this to integrate with on the level of their own request and response objects, instead of on the WSGI level.
On some platforms and filesystems (such as Linux ext3, Mac OS X)
os.path.getmtime()returns timestamps with the granularity of seconds instead of subseconds such as Linux ext4. We go for second granularity now by default for autoversioning as this should be good enough during development.
The test for autoversioning was assuming sub-second granularity and this test failed. This test is now skipped on Mac OS X. What didn’t help was weird code in BowerStatic that cut off the last bit of the microsecond isoformat – this was removed.
This fixes bug #20. (Thanks to Michael Howitz for the bug report)
Display a nicer error messages when a component depends on another one that doesn’t exist. Thanks for Michael Howitz for the improvement.
Internal toposort module was not imported relative to package, which could lead to errors in some circumstances. Thanks TylorS for reporting! Fixes issue #24.
- There was a bug in the new
component(name)method to obtain the component, because it wouldn’t work for local components. Fixing this properly took a significant refactoring:
- the ComponentCollection gains its own fallback behavior, much simpler to implement than in LocalComponentCollection.
- UrlComponent is now gone and Component gains that functionality; it keeps a reference to the collection itself now.
- add a lot of free-standing functions to the methods.
- an earlier hack passing the component collection through an argument is now gone.
- The new
.component()API to get a component from a collection explicitly in order to get its URL is gone again as it conflicted with an earlier API on local component collections. Instead use
- BowerStatic failed to initialize if a component was discovered
main. These are handled now – if you try to include a path to such a resource, no extra inclusions are generated. Fixes #5.
mainwas an array, only the first such resource was loaded. The correct behavior should be to load all these resources. This required a reworking of how resources get created; instead of creating a single resource for a path, a list of resources is created everywhere. Fixes #6 (and was mentioned in #5).
- Introduce a
component(name)method on the components object. Given a component name it will give an object that has a
url()method. This can be used to obtain the URL of a component directory, which is sometimes useful when client-side code needs to construct URLs itself, such as for templates. Fixes issue #8.
- You can register a renderer for a particular extension type using,
bower.renderer('.js', render_js), where
render_jstakes a URL and should return a string with a HTML snippet to include on the page.
- Even if the same resource is included multiple times, it will only be included once. Thanks Ying Zhong for the bug report and suggested fix.
- Initial public release.