Add interface for prefix cache indexer #657
Conversation
| type PrefixCacheIndexer interface { | ||
| // MatchPrefix matches the longest prefix sequence for input request (passed as input tokens) | ||
| // and returns matched prefix (as tokens), remaining unmatched input request (as tokens) and pods matching the prefix | ||
| MatchPrefix(inputTokens []int, model string, pods []*v1.Pod) (matchedTokes []int, unMatchedTokens []int, matchedPods []*v1.Pod) |
There was a problem hiding this comment.
@DwyaneShi pleas help review the routing indexer interface.
There was a problem hiding this comment.
@gangmuk You'd like to review this as well since you will join the indexer implementation.
There was a problem hiding this comment.
typo I think matchedTokes -> matchedTokens
There was a problem hiding this comment.
what's the usage of unMatchedTokens? is it better to return directly or just matched and caller could know the unMatched automatically
There was a problem hiding this comment.
For hash table usecase, we only need to add unmatched tokens hence returned unmatchedtokens.
Technically caller can know unmatched tokens from returned matched tokens but it will add extra computation to figure out.
| type PrefixCacheIndexer interface { | ||
| // MatchPrefix matches the longest prefix sequence for input request (passed as input tokens) | ||
| // and returns matched prefix (as tokens), remaining unmatched input request (as tokens) and pods matching the prefix | ||
| MatchPrefix(inputTokens []int, model string, pods []*v1.Pod) (matchedTokes []int, unMatchedTokens []int, matchedPods []*v1.Pod) |
There was a problem hiding this comment.
typo I think matchedTokes -> matchedTokens
* Add interface for prefix cache indexer * address review comments * refactor dir layout Signed-off-by: Varun Gupta <[email protected]>
No description provided.