514: Stop flattening every field r=Kerollmops a=irevoire
When we need to flatten a document:
* The primary key contains a `.`.
* Some fields need to be flattened
Instead of flattening the whole object and thus creating a lot of allocations with the `serde_json_flatten_crate`, we instead generate a minimal sub-object containing only the fields that need to be flattened.
That should create fewer allocations and thus index faster.
---------
```
group indexing_main_e1e362fa indexing_stop-flattening-every-field_40d1bd6b
----- ---------------------- ---------------------------------------------
indexing/Indexing geo_point 1.99 23.7±0.23s ? ?/sec 1.00 11.9±0.21s ? ?/sec
indexing/Indexing movies in three batches 1.00 18.2±0.24s ? ?/sec 1.01 18.3±0.29s ? ?/sec
indexing/Indexing movies with default settings 1.00 17.5±0.09s ? ?/sec 1.01 17.7±0.26s ? ?/sec
indexing/Indexing songs in three batches with default settings 1.00 64.8±0.47s ? ?/sec 1.00 65.1±0.49s ? ?/sec
indexing/Indexing songs with default settings 1.00 54.9±0.99s ? ?/sec 1.01 55.7±1.34s ? ?/sec
indexing/Indexing songs without any facets 1.00 50.6±0.62s ? ?/sec 1.01 50.9±1.05s ? ?/sec
indexing/Indexing songs without faceted numbers 1.00 54.0±1.14s ? ?/sec 1.01 54.7±1.13s ? ?/sec
indexing/Indexing wiki 1.00 996.2±8.54s ? ?/sec 1.02 1021.1±30.63s ? ?/sec
indexing/Indexing wiki in three batches 1.00 1136.8±9.72s ? ?/sec 1.00 1138.6±6.59s ? ?/sec
```
So basically everything slowed down a liiiiiittle bit except the dataset with a nested field which got twice faster
Co-authored-by: Tamo <tamo@meilisearch.com>
515: Improve the README r=curquiza a=Kerollmops
This PR closes#512 by adding more content to the README. We listed all of the subcrates of the repository, changed the descriptions of the subcrates, and added a simple example usage in the README.
Co-authored-by: Kerollmops <clement@meilisearch.com>
Co-authored-by: Clémentine Urquizar - curqui <clementine@meilisearch.com>
509: Remove pr_status from bors settings r=Kerollmops a=curquiza
Because of multiple issue we had with bors.
https://github.com/bors-ng/bors-ng/issues/1492
Co-authored-by: Clémentine Urquizar <clementine@meilisearch.com>
505: normalize exact words r=curquiza a=MarinPostma
Normalize the exact words, as specified in the specification.
Co-authored-by: ad hoc <postma.marin@protonmail.com>
499: fix min-word-len-for-typo not reset properly r=Kerollmops a=MarinPostma
fix min word len for typo not resettign properly, as reported in https://github.com/meilisearch/meilisearch/issues/2330
Co-authored-by: ad hoc <postma.marin@protonmail.com>
483: Enhance matching words r=Kerollmops a=ManyTheFish
# Summary
Enhance milli word-matcher making it handle match computing and cropping.
# Implementation
## Computing best matches for cropping
Before we were considering that the first match of the attribute was the best one, this was accurate when only one word was searched but was missing the target when more than one word was searched.
Now we are searching for the best matches interval to crop around, the chosen interval is the one:
1) that have the highest count of unique matches
> for example, if we have a query `split the world`, then the interval `the split the split the` has 5 matches but only 2 unique matches (1 for `split` and 1 for `the`) where the interval `split of the world` has 3 matches and 3 unique matches. So the interval `split of the world` is considered better.
2) that have the minimum distance between matches
> for example, if we have a query `split the world`, then the interval `split of the world` has a distance of 3 (2 between `split` and `the`, and 1 between `the` and `world`) where the interval `split the world` has a distance of 2. So the interval `split the world` is considered better.
3) that have the highest count of ordered matches
> for example, if we have a query `split the world`, then the interval `the world split` has 2 ordered words where the interval `split the world` has 3. So the interval `split the world` is considered better.
## Cropping around the best matches interval
Before we were cropping around the interval without checking the context.
Now we are cropping around words in the same context as matching words.
This means that we will keep words that are farther from the matching words but are in the same phrase, than words that are nearer but separated by a dot.
> For instance, for the matching word `Split` the text:
`Natalie risk her future. Split The World is a book written by Emily Henry. I never read it.`
will be cropped like:
`…. Split The World is a book written by Emily Henry. …`
and not like:
`Natalie risk her future. Split The World is a book …`
Co-authored-by: ManyTheFish <many@meilisearch.com>
493: Use smartstring to store the external id in our hashmap r=Kerollmops a=irevoire
We need to store all the external id (primary key) in a hashmap
associated to their internal id.
The smartstring remove heap allocation / memory usage and should
improve the cache locality.
I ran the benchmarks to measure the impact of this PR on the indexing time.
I think we should merge it whatever happens thought because it'll decrease the memory consumption.
---------
This improve really sliiiiiightly the performances but improve the memory usage thus it should be merged.
```
group indexing_main_6b073738 indexing_use-smartsring_3f343511
----- ---------------------- --------------------------------
indexing/Indexing geo_point 1.02 25.2±0.20s ? ?/sec 1.00 24.8±0.13s ? ?/sec
indexing/Indexing movies in three batches 1.00 18.2±0.10s ? ?/sec 1.00 18.2±0.23s ? ?/sec
indexing/Indexing movies with default settings 1.00 17.5±0.09s ? ?/sec 1.01 17.7±0.11s ? ?/sec
indexing/Indexing songs in three batches with default settings 1.00 68.3±1.01s ? ?/sec 1.00 68.0±0.95s ? ?/sec
indexing/Indexing songs with default settings 1.00 63.2±0.78s ? ?/sec 1.00 63.0±0.58s ? ?/sec
indexing/Indexing songs without any facets 1.02 59.6±1.00s ? ?/sec 1.00 58.5±1.03s ? ?/sec
indexing/Indexing songs without faceted numbers 1.00 62.8±0.38s ? ?/sec 1.00 62.6±1.02s ? ?/sec
indexing/Indexing wiki 1.01 1009.2±25.25s ? ?/sec 1.00 998.1±11.27s ? ?/sec
indexing/Indexing wiki in three batches 1.01 1142.0±9.97s ? ?/sec 1.00 1134.4±11.21s ? ?/sec
```
Co-authored-by: Tamo <tamo@meilisearch.com>
We need to store all the external id (primary key) in a hashmap
associated to their internal id during.
The smartstring remove heap allocation / memory usage and should
improve the cache locality.
496: Improve the performances of the flattening subcrate r=irevoire a=Kerollmops
This PR adds some benchmarks to the _flatten-serde-json_ crate, this crate is responsible for transforming the original documents into flat versions that the engine can understand. It can probably be speed-up and this is why I added benchmarks to it.
I make some interesting performance improvements when I replaced the `json!` macro calls.
```
flatten/simple time: [452.44 ns 453.31 ns 454.18 ns]
change: [-15.036% -14.751% -14.473%] (p = 0.00 < 0.05)
Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
2 (2.00%) high mild
Benchmarking flatten/complex: Collecting 100 samples in estimated 5.0007 s (4.9M i flatten/complex time: [1.0101 us 1.0131 us 1.0160 us]
change: [-18.001% -17.775% -17.536%] (p = 0.00 < 0.05)
Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
5 (5.00%) high mild
1 (1.00%) high severe
```
---
_I removed this particular commit from this PR._ The reason is that the two other commits were enough for this PR to give enough impact and be merged. We will continue to explore where we can get performances later.
But when I changed the flattening function to accept an owned version of the objects, we lost a lot of performances. Yes, I rewrote the benchmarks (locally) to clone the input object (and measured both, previous and new versions, with the cloning benchmarks). Maybe cloning the benchmark inputs is not the right thing to do...
```
Benchmarking flatten/simple: Collecting 100 samples in estimated 5.0005 s (6.7M it flatten/simple time: [746.46 ns 749.59 ns 752.70 ns]
change: [+40.082% +40.714% +41.347%] (p = 0.00 < 0.05)
Performance has regressed.
Benchmarking flatten/complex: Collecting 100 samples in estimated 5.0047 s (2.9M i flatten/complex time: [1.7311 us 1.7342 us 1.7368 us]
change: [+40.976% +41.398% +41.807%] (p = 0.00 < 0.05)
Performance has regressed.
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) low mild
```
Co-authored-by: Kerollmops <clement@meilisearch.com>
489: fix distinct count bug r=curquiza a=MarinPostma
fix https://github.com/meilisearch/meilisearch/issues/2152
I think the issue was that we didn't take off the excluded candidates from the initial candidates when returning the candidates with the search result.
Co-authored-by: ad hoc <postma.marin@protonmail.com>
491: remove the unused key warning r=curquiza a=irevoire
When I copy-pasted my flatten crate I forgot to remove the key used to publish the package and that throw a warning.
Co-authored-by: Tamo <tamo@meilisearch.com>
