202
edits
Places:AutoComplete: Difference between revisions
Jump to navigation
Jump to search
→Possible algorithmic changes
Minh Nguyen (talk | contribs) m (→Performance with the new database: Typo) |
BrettWilson (talk | contribs) |
||
| Line 3: | Line 3: | ||
== Possible algorithmic changes == | == Possible algorithmic changes == | ||
Places (Firefox | Places (Firefox 3.0) will <i>not</i> likely see a large change in the AutoComplete algorithm. The current differences are: | ||
* It is faster. The places autocomplete code should be faster than the old code on and normal amount of data. For large histories, it will generally be much faster. | |||
* Suggestion of host names first. If the first URL has not been visited very many times and it has a path component, a new entry is suggested as the first item consisting of just the host portion, even if that URL has never been visited. If the first item has been visited many times, we assume the user actually wants to go to that page and don't do this addition. | |||
* We give a little boost to URLs that are bookmarked. | |||
There are some additional simple low-risk changes that we might consider making. | |||
* Better use of the "typed" flag (set for URLs that have ever been entered in the address bar). The current implementation gives typed URLs a constant boost over non-typed URLs. It would be nice to weight the various parameters by time, so a recent typed URL would still take precidence over something visited several times but many weeks ago. | * Better use of the "typed" flag (set for URLs that have ever been entered in the address bar). The current implementation gives typed URLs a constant boost over non-typed URLs. It would be nice to weight the various parameters by time, so a recent typed URL would still take precidence over something visited several times but many weeks ago. | ||
* When considering the aforementioned "aging" of priority, we might want to take into account browsing behavior. It sucks to come back from a two-week vacation and have all your history and autocomplete expired. At startup, we can detect periods of little or no activity and adjust the aging parameters accordingly. | * When considering the aforementioned "aging" of priority, we might want to take into account browsing behavior. It sucks to come back from a two-week vacation and have all your history and autocomplete expired. At startup, we can detect periods of little or no activity and adjust the aging parameters accordingly. | ||
== Performance with the new database == | == Performance with the new database == | ||