| Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
|
|
|
|
|
|
Maybe to make any stacktraces thrown by tests shorter.
|
|
|
|
|
|
|
|
As part of the coffeescript port.
|
|
|
|
|
|
|
|
Right now this is triggered on Shift+T. Firefox also does tab matching in their
omnibar. I'm not sure if collapsing the two searches into one UI is a good idea.
|
|
created excessively.
|
|
|
|
|
|
This allows the vomnibar rankings to properly rank sites you've visited recently,
after it populated its original cache.
|
|
|
|
I'm worried about keeping around a reference to history forever, unnecessarily.
|
|
|
|
|
|
|
|
|
|
|
|
Wish this was shorter.
|
|
|
|
One major change between this and the previous implementation is
that all scores are between [0, 1] and "1" means most relevant.
|
|
|
|
The purpose of this refactor is to simplify the contract so it's easier
to modify, and to make some substantial usability improvements.
One of the key differences is that matching is no longer fuzzy. If you
want to search more than one term, separate them by spaces. This matches
the behavior in Firefox.
While fuzzy matching is a nice experience for a limited set of known
items (like files in the current project), it doesn't work well for
a huge messy collection, like URLs in your history. The query "hello"
will match random stuff from your google search results for instance,
and it's hard to prune that noise using ranking intelligence.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
They're aleady being discarded for ranking purposes.
|
|
|
|
are less.
|
|
|
|
|
|
display cleaner.
|
|
|