202
edits
BrettWilson (talk | contribs) |
BrettWilson (talk | contribs) |
||
| Line 14: | Line 14: | ||
If a significant amount of history will be stored, a much more efficient method of searching will be required (the current implemtation searches all of history for matches). | If a significant amount of history will be stored, a much more efficient method of searching will be required (the current implemtation searches all of history for matches). | ||
The database can be queried for URLs matching multiple criteria: recency, popularity, and type-edness. We can probably encode some of the ranking in a query. For example, select all pages visited in the last <i>n</i> days OR where (visit count - days last visited ago > 1). This will save us from ranking most of the pages in the user's history. | The database can be queried for URLs matching multiple criteria: recency, popularity, and type-edness. We can probably encode some of the ranking in a query. For example, select all pages visited in the last <i>n</i> days OR where (visit count - days last visited ago > 1). This will save us from ranking most of the pages in the user's history. | ||
We might want to consider another column in the histroy DB for autocomplete purposes. This would just contain the URL minus the protocol type and with the common prefixes stripped (the corrent implemetation strips "www" and "ftp", and we'll probably just continue this). Then an index could be created on this column and we can quickly find matching pages without schlepping through all of them comparing strings. | We might want to consider another column in the histroy DB for autocomplete purposes. This would just contain the URL minus the protocol type and with the common prefixes stripped (the corrent implemetation strips "www" and "ftp", and we'll probably just continue this). Then an index could be created on this column and we can quickly find matching pages without schlepping through all of them comparing strings. | ||
edits