Log
-
Add basic benchmarking code 💬 by Caio 5 years ago
This commit intentionally leaves Cargo.lock untouched so that rebasing is easier in the future.
-
Use the new collector api by Caio 5 years ago
-
Change collector api to a `DocId,Query,Result` receiver 💬 by Caio 5 years ago
Slightly less cute, a lot more convenient
-
Revert "Build against tantivy's upstream repository" 💬 by Caio 5 years ago
This reverts commit 3392d944b5536feff32db026231dee2dceb3e9c2.
-
Use ordered_by_f64_from_u64_fast_field by Caio 5 years ago
-
Add functionality to sort u64 fields as f64 💬 by Caio 5 years ago
Somewhat of a partial revert of b59efa1da
-
Store floats in a u64 field 💬 by Caio 5 years ago
I still hit the fast field crash when indexing fast float fieds in parallel. My guess is that somewhere along the code it assumes as fast field readers are loaded (maybe prior to the second merging thread?) but floats are relatively rare in model and then it blows up. Anyway, I'll investigate that later - let's get this working first.
-
Revert "Make float fields FAST too" 💬 by Caio 5 years ago
This reverts commit 59e03d738ebda6c5ef1702adca19e2a44c97b301.
-
Add support for sorting by float features 💬 by Caio 5 years ago
The search is now feature complete, I think.
-
Expose ordered_by_f64_fast_field by Caio 5 years ago
-
Make float fields FAST too by Caio 5 years ago
-
Build against tantivy's upstream repository 💬 by Caio 5 years ago
This is hopefully temporary. I wanna push the float-related stuff forward because otherwise I'll keep coming up with interesting-but- useless-for-me-now cases like what I just did with the database :P I had to touch the QueryParser code due to changes in the public api: * `Field` now has a constructor instead of a public inner field * `BoxedTokenizer` is a struct wrapping the actual box dance (thanks!) Direct link to build head: https://github.com/tantivy-search/tantivy/commit/b9896c496207ecff14a1ffc713dc81d009500b10
-
Typo: s/queries/queried/ by Caio 5 years ago
-
Merge branch 'db_reader_writer' 💬 by Caio 5 years ago
This is good enough for now. I don't think I'll follow through with a borrow-heavy model because it's a lot of work for very little gain - hard to justify. So, maybe one day as a personal challenge...
-
Reorganize the database mod internal structure by Caio 5 years ago
-
Drop BincodeDatabase, use Database{Reader,Writer} 💬 by Caio 5 years ago
Very smooth transition. Thanks compiler!
-
Drive the database via DbReader and DbWriter 💬 by Caio 5 years ago
Now I have a very similar database implementation, but supporting borrowed types. Next steps: * Clean up the interface, forward serde errors * Replace BincodeDatabase with DbWriter+DbReader * Make model::Recipe a borrow type Then I'll be avoiding a ton of useless copying I would be doing when rendering search results.
-
ConfigDb is now MmapDatabase 💬 by Caio 5 years ago
It's a non-remappable, offset-driven thin layer over MmapMut that knows about what kind of data is stored inside of it. But I think I'll split this further: There's nothing fancy ever happening on the `add` by design, so I won't have segments, I won't parallelize writes not anything - it's easier then to have a DbWriter and DbReader, so that the writer doesn't need to care about mmap and the reader doesn't care about `Encoder`
-
Split Config into Encoder and Decoder traits 💬 by Caio 5 years ago
At this point the only useful part of having `add` instead of `add_bytes` is being able to do something with the item prior to storing
-
Experiment holding the Config impl 💬 by Caio 5 years ago
That's a way nicer API, but it doesn't do what I want on `add`. I lose control of `first` and `second` in the test: I can't wrap it in an optional and test it against a database result, can't even drop it without having the whole database. So it looks like it's better to split the config in two and then I can hold the decoder config and require the encoder as a type parameter like in the first try.
-
WIP: Db that can borrow from its mmap 💬 by Caio 5 years ago
Took me forever to figure out how to express this...
-
Extract log functionality into StructuredLog<T> by Caio 5 years ago
-
More thorough pagination test 💬 by Caio 5 years ago
Previously we were checking if we walk through INDEX_SIZE results, without caring what they were. This patch changes it so that we expect to see INDEX_SIZE *distinct* ids in the result set.
-
Verify that sorting works by Caio 5 years ago
-
Add very simple integration tests by Caio 5 years ago
-
Move business logic out of cantine by Caio 5 years ago
-
Alias RecipeId<=>u64 by Caio 5 years ago
-
Move Cantine and related into the index mod 💬 by Caio 5 years ago
The tests are a-coming
-
Use TantivyError instead of &str for Fields by Caio 5 years ago
-
Transform `tantivy_addresses_to_ids!` into a method 💬 by Caio 5 years ago
This is the easy one. I'm not looking forward to tackling the other two macros in the `basic_search` code...