Search: Basic shard state #1249
Conversation
Signed-off-by: Vladislav Oleshko <[email protected]>
Signed-off-by: Vladislav Oleshko <[email protected]>
dranikpg
changed the title
Signed-off-by: Vladislav Oleshko <[email protected]>
src/server/search/search_accessor.h
Outdated
| struct BaseAccessor : public search::DocumentAccessor { | ||
| using FieldConsumer = search::DocumentAccessor::FieldConsumer; | ||
|
|
||
| virtual SearchDocData Serialize() const = 0; |
There was a problem hiding this comment.
remind what Serialize is used for?
There was a problem hiding this comment.
Convert underlying type to a map<string, string> to be sent as a reply, added a comment
|
Fixed comments. I also did some naming changes, because there will be around 5 depth levels and each one can be called index. Randomly adding words like shard or search doesn't make it more clear 🙂 I suggest calling "logical indices" just document indices (those defined by ft.create SCHEMA PREFIX...). This indicates that they don't perform search, but just keep track of documents. Shard prefix just means its per shard. Now its like this: DocIndex , ShardDocIndices -> ShardDocIndex -> [in pr2] -> FieldIndices -> BaseIndex -> e.g. NumericIndex |
Signed-off-by: Vladislav Oleshko <[email protected]>
| : query{make_unique<AstExpr>(ParseQuery(query))} { | ||
| bool SearchAlgorithm::Init(string_view query) { | ||
| try { | ||
| query_ = make_unique<AstExpr>(ParseQuery(query)); |
There was a problem hiding this comment.
make a comment explaining why do you try/catch.
catch the exception thrown by parser, ideally vlog it as well.
Signed-off-by: Vladislav Oleshko <[email protected]>
|
@romange Please re-approve, I want to rebase on this |
Context:
This is a preparation step before adding basic indices to core/search. Indices will be managed per shard, need to be constructed when the index is created with ft.create and need to be updated when new entries are added/deleted. Indices will store integer codes (compact & faster) instead of keys for documents.
Goal: