3834: Define searchable fields at runtime r=Kerollmops a=ManyTheFish
## Summary
This feature allows the end-user to search in one or multiple attributes using the search parameter `attributesToSearchOn`:
```json
{
"q": "Captain Marvel",
"attributesToSearchOn": ["title"]
}
```
This feature act like a filter, forcing Meilisearch to only return the documents containing the requested words in the attributes-to-search-on. Note that, with the matching strategy `last`, Meilisearch will only ensure that the first word is in the attributes-to-search-on, but, the retrieved documents will be ordered taking into account the word contained in the attributes-to-search-on.
## Trying the prototype
A dedicated docker image has been released for this feature:
#### last prototype version:
```bash
docker pull getmeili/meilisearch:prototype-define-searchable-fields-at-search-time-1
```
#### others prototype versions:
```bash
docker pull getmeili/meilisearch:prototype-define-searchable-fields-at-search-time-0
```
## Technical Detail
The attributes-to-search-on list is given to the search context, then, the search context uses the `fid_word_docids`database using only the allowed field ids instead of the global `word_docids` database. This is the same for the prefix databases.
The database cache is updated with the merged values, meaning that the union of the field-id-database values is only made if the requested key is missing from the cache.
### Relevancy limits
Almost all ranking rules behave as expected when ordering the documents.
Only `proximity` could miss-order documents if all the searched words are in the restricted attribute but a better proximity is found in an ignored attribute in a document that should be ranked lower. I put below a failing test showing it:
```rust
#[actix_rt::test]
async fn proximity_ranking_rule_order() {
let server = Server::new().await;
let index = index_with_documents(
&server,
&json!([
{
"title": "Captain super mega cool. A Marvel story",
// Perfect distance between words in an ignored attribute
"desc": "Captain Marvel",
"id": "1",
},
{
"title": "Captain America from Marvel",
"desc": "a Shazam ersatz",
"id": "2",
}]),
)
.await;
// Document 2 should appear before document 1.
index
.search(json!({"q": "Captain Marvel", "attributesToSearchOn": ["title"], "attributesToRetrieve": ["id"]}), |response, code| {
assert_eq!(code, 200, "{}", response);
assert_eq!(
response["hits"],
json!([
{"id": "2"},
{"id": "1"},
])
);
})
.await;
}
```
Fixing this would force us to create a `fid_word_pair_proximity_docids` and a `fid_word_prefix_pair_proximity_docids` databases which may multiply the keys of `word_pair_proximity_docids` and `word_prefix_pair_proximity_docids` by the number of attributes in the searchable_attributes list. If we think we should fix this test, I'll suggest doing it in another PR.
## Related
Fixes#3772
Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: ManyTheFish <many@meilisearch.com>
3821: Add normalized and detailed scores to documents returned by a query r=dureuill a=dureuill
# Pull Request
## Related issue
Fixes#3771
## What does this PR do?
### User standpoint
<details>
<summary>Request ranking score</summary>
```
echo '{
"q": "Badman dark knight returns",
"showRankingScore": true,
"limit": 10,
"attributesToRetrieve": ["title"]
}' | mieli search -i index-word-count-10-count
```
</details>
<details>
<summary>Response</summary>
```json
{
"hits": [
{
"title": "Batman: The Dark Knight Returns, Part 1",
"_rankingScore": 0.947520325203252
},
{
"title": "Batman: The Dark Knight Returns, Part 2",
"_rankingScore": 0.947520325203252
},
{
"title": "Batman Unmasked: The Psychology of the Dark Knight",
"_rankingScore": 0.6657594086021505
},
{
"title": "Legends of the Dark Knight: The History of Batman",
"_rankingScore": 0.6654905913978495
},
{
"title": "Angel and the Badman",
"_rankingScore": 0.2196969696969697
},
{
"title": "Angel and the Badman",
"_rankingScore": 0.2196969696969697
},
{
"title": "Batman",
"_rankingScore": 0.11553030303030302
},
{
"title": "Batman Begins",
"_rankingScore": 0.11553030303030302
},
{
"title": "Batman Returns",
"_rankingScore": 0.11553030303030302
},
{
"title": "Batman Forever",
"_rankingScore": 0.11553030303030302
}
],
"query": "Badman dark knight returns",
"processingTimeMs": 12,
"limit": 10,
"offset": 0,
"estimatedTotalHits": 46
}
```
</details>
- If adding a `showRankingScore` parameter to the search query, then documents returned by a search now contain an additional field `_rankingScore` that is a float bigger than 0 and lower or equal to 1.0. This field represents the relevancy of the document, relatively to the search query and the settings of the index, with 1.0 meaning "perfect match" and 0 meaning "not matching the query" (Meilisearch should never return documents not matching the query at all).
- The `sort` and `geosort` ranking rules do not influence the `_rankingScore`.
<details>
<summary>Request detailed ranking scores</summary>
```
echo '{
"q": "Badman dark knight returns",
"showRankingScoreDetails": true,
"limit": 5,
"attributesToRetrieve": ["title"]
}' | mieli search -i index-word-count-10-count
```
</details>
<details>
<summary>Response</summary>
```json
{
"hits": [
{
"title": "Batman: The Dark Knight Returns, Part 1",
"_rankingScoreDetails": {
"words": {
"order": 0,
"matchingWords": 4,
"maxMatchingWords": 4,
"score": 1.0
},
"typo": {
"order": 1,
"typoCount": 1,
"maxTypoCount": 4,
"score": 0.8
},
"proximity": {
"order": 2,
"score": 0.9545454545454546
},
"attribute": {
"order": 3,
"attributes_ranking_order": 1.0,
"attributes_query_word_order": 0.926829268292683,
"score": 0.926829268292683
},
"exactness": {
"order": 4,
"matchType": "noExactMatch",
"score": 0.26666666666666666
}
}
},
{
"title": "Batman: The Dark Knight Returns, Part 2",
"_rankingScoreDetails": {
"words": {
"order": 0,
"matchingWords": 4,
"maxMatchingWords": 4,
"score": 1.0
},
"typo": {
"order": 1,
"typoCount": 1,
"maxTypoCount": 4,
"score": 0.8
},
"proximity": {
"order": 2,
"score": 0.9545454545454546
},
"attribute": {
"order": 3,
"attributes_ranking_order": 1.0,
"attributes_query_word_order": 0.926829268292683,
"score": 0.926829268292683
},
"exactness": {
"order": 4,
"matchType": "noExactMatch",
"score": 0.26666666666666666
}
}
},
{
"title": "Batman Unmasked: The Psychology of the Dark Knight",
"_rankingScoreDetails": {
"words": {
"order": 0,
"matchingWords": 3,
"maxMatchingWords": 4,
"score": 0.75
},
"typo": {
"order": 1,
"typoCount": 1,
"maxTypoCount": 3,
"score": 0.75
},
"proximity": {
"order": 2,
"score": 0.6666666666666666
},
"attribute": {
"order": 3,
"attributes_ranking_order": 1.0,
"attributes_query_word_order": 0.8064516129032258,
"score": 0.8064516129032258
},
"exactness": {
"order": 4,
"matchType": "noExactMatch",
"score": 0.25
}
}
},
{
"title": "Legends of the Dark Knight: The History of Batman",
"_rankingScoreDetails": {
"words": {
"order": 0,
"matchingWords": 3,
"maxMatchingWords": 4,
"score": 0.75
},
"typo": {
"order": 1,
"typoCount": 1,
"maxTypoCount": 3,
"score": 0.75
},
"proximity": {
"order": 2,
"score": 0.6666666666666666
},
"attribute": {
"order": 3,
"attributes_ranking_order": 1.0,
"attributes_query_word_order": 0.7419354838709677,
"score": 0.7419354838709677
},
"exactness": {
"order": 4,
"matchType": "noExactMatch",
"score": 0.25
}
}
},
{
"title": "Angel and the Badman",
"_rankingScoreDetails": {
"words": {
"order": 0,
"matchingWords": 1,
"maxMatchingWords": 4,
"score": 0.25
},
"typo": {
"order": 1,
"typoCount": 0,
"maxTypoCount": 1,
"score": 1.0
},
"proximity": {
"order": 2,
"score": 1.0
},
"attribute": {
"order": 3,
"attributes_ranking_order": 1.0,
"attributes_query_word_order": 0.8181818181818182,
"score": 0.8181818181818182
},
"exactness": {
"order": 4,
"matchType": "noExactMatch",
"score": 0.3333333333333333
}
}
}
],
"query": "Badman dark knight returns",
"processingTimeMs": 9,
"limit": 5,
"offset": 0,
"estimatedTotalHits": 46
}
```
</details>
- If adding a `showRankingScoreDetails` parameter to the search query, then the returned documents will now contain an additional `_rankingScoreDetails` field that is a JSON object containing one field per ranking rule that was applied, whose value is a JSON object with the following fields:
- `order`: a number indicating the order this rule was applied (0 is the first applied ranking rule)
- `score` (except for `sort` and `geosort`): a float indicating how the document matched this particular rule.
- other fields that are specific to the rule, indicating for example how many words matched for a document and how many typos were counted in a matching document.
- If the `displayableAttributes` list is defined in the settings of the index, any ranking rule using an attribute **not** part of that list will be marked as `<hidden-rule>` in the `_rankingScoreDetails`.
- Search queries that are part of a `multi-search` requests are modified in the same way and each of the queries can take the `showRankingScore` and `showRankingScoreDetails` parameters independently. The results are still returned in separate lists and providing a unified list of results between multiple queries is not in the scope of this PR (but is unblocked by this PR and can be done manually by using the scores of the various documents).
### Implementation standpoint
- Fix difference in how the position of terms were computed at indexing time and query time: this difference meant that a query containing a hard separator would fail the exactness check.
- Fix the id reported by the sort ranking rule (very minor)
- Change how the cost of removing words is computed. After this change the cost no longer works for any other ranking rule than `words`. Also made `words` have a cost of 0 such that the entire cost of `words` is given by the termRemovalStrategy. The new cost computation makes it so the score is computed in a way consistent with the number of words in the query. Additionally, the words that appear in phrases in the query are also counted as matching words.
- When any score computation is requested through `showRankingScore` or `showRankingScoreDetails`, remove optimization where ranking rules are not executed on buckets of a single document: this is important to allow the computation of an accurate score.
- add virtual conditions to fid and position to always have the max cost: this ensures that the score is independent from the dataset
- the Position ranking rule now takes into account the distance to the position of the word in the query instead of the distance to the position 0.
- modified proximity ranking rule cost calculation so that the cost is 0 for documents that are perfectly matching the query
- Add a new `milli::score_details` module containing all the types that are involved in score computation.
- Make it so a bucket of result now contains a `ScoreDetails` and changed the ranking rules to produce their `ScoreDetails`.
- Expose the scores in the REST API.
- Add very light analytics for scoring.
- Update the search tests to add the expected scores.
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
3842: fix some typos r=dureuill a=cuishuang
# Pull Request
## Related issue
Fixes #<issue_number>
## What does this PR do?
- fix some typos
## PR checklist
Please check if your PR fulfills the following requirements:
- [x] Does this PR fix an existing issue, or have you listed the changes applied in the PR description (and why they are needed)?
- [x] Have you read the contributing guidelines?
- [x] Have you made sure that the title is accurate and descriptive of the changes?
Thank you so much for contributing to Meilisearch!
Co-authored-by: cui fliter <imcusg@gmail.com>
3670: Fix addition deletion bug r=irevoire a=irevoire
The first commit of this PR is a revert of https://github.com/meilisearch/meilisearch/pull/3667. It re-enable the auto-batching of addition and deletion of tasks. No new changes have been introduced outside of `milli`. So all the changes you see on the autobatcher have actually already been reviewed.
It fixes https://github.com/meilisearch/meilisearch/issues/3440.
### What was happening?
The issue was that the `external_documents_ids` generated in the `transform` were used in a very strange way that wasn’t compatible with the deletion of documents.
Instead of doing a clear merge between the external document IDs of the DB and the one returned by the transform + writing it on disk, we were doing some weird tricks with the soft-deleted to avoid writing the fst on disk as much as possible.
The new algorithm may be a bit slower but is way more straightforward and doesn’t change depending on if the soft deletion was used or not. Here is a list of the changes introduced:
1. We now do a clear distinction between the `new_external_documents_ids` coming from the transform and only held on RAM and the `external_documents_ids` coming from the DB.
2. The `new_external_documents_ids` (coming out of the transform) are now represented as an `fst`. We don't need to struggle with the hard, soft distinction + the soft_deleted => That's easier to understand
3. When indexing documents, we merge the `external_documents_ids` coming from the DB and the `new_external_documents_ids` coming from the transform.
### Other things introduced in this PR
Since we constantly have to write small, very specialized fuzzers for this kind of bug, we decided to push the one used to reproduce this bug.
It's not perfect, but it's easy to improve in the future.
It'll also run for as long as possible on every merge on the main branch.
Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: Loïc Lecrenier <loic.lecrenier@icloud.com>
3835: Add more documentation to graph-based ranking rule algorithms + comment cleanup r=Kerollmops a=loiclec
In addition to documenting the `cheapest_path.rs` file, this PR cleans up a few outdated comments as well as some TODOs. These TODOs have been moved to https://github.com/meilisearch/meilisearch/issues/3776
Co-authored-by: Loïc Lecrenier <loic.lecrenier@icloud.com>
3768: Fix bugs in graph-based ranking rules + make `words` a graph-based ranking rule r=dureuill a=loiclec
This PR contains three changes:
## 1. Don't call the `words` ranking rule if the term matching strategy is `All`
This is because the purpose of `words` is only to remove nodes from the query graph. It would never do any useful work when the matching strategy was `All`. Remember that the universe was already computed before by computing all the docids corresponding to the "maximally reduced" query graph, which, in the case of `All`, is equal to the original graph.
## 2. The `words` ranking rule is replaced by a graph-based ranking rule.
This is for three reasons:
1. **performance**: graph-based ranking rules benefit from a lot of optimisations by default, which ensures that they are never too slow. The previous implementation of `words` could call `compute_query_graph_docids` many times if some words had to be removed from the query, which would be quite expensive. I was especially worried about its performance in cases where it is placed right after the `sort` ranking rule. Furthermore, `compute_query_graph_docids` would clone a lot of bitmaps many times unnecessarily.
2. **consistency**: every other ranking rule (except `sort`) is graph-based. It makes sense to implement `words` like that as well. It will automatically benefit from all the features, optimisations, and bug fixes that all the other ranking rules get.
3. **surfacing bugs**: as the first ranking rule to be called (most of the time), I'd like `words` to behave the same as the other ranking rules so that we can quickly detect bugs in our graph algorithms. This actually already happened, which is why this PR also contains a bug fix.
## 3. Fix the `update_all_costs_before_nodes` function
It is a bit difficult to explain what was wrong, but I'll try. The bug happened when we had graphs like:
<img width="730" alt="Screenshot 2023-05-16 at 10 58 57" src="https://github.com/meilisearch/meilisearch/assets/6040237/40db1a68-d852-4e89-99d5-0d65757242a7">
and we gave the node `is` as argument.
Then, we'd walk backwards from the node breadth-first. We'd update the costs of:
1. `sun`
2. `thesun`
3. `start`
4. `the`
which is an incorrect order. The correct order is:
1. `sun`
2. `thesun`
3. `the`
4. `start`
That is, we can only update the cost of a node when all of its successors have either already been visited or were not affected by the update to the node passed as argument. To solve this bug, I factored out the graph-traversal logic into a `traverse_breadth_first_backward` function.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
3757: Adjust the cost of edges in the `position` ranking rule by bucketing positions more aggressively r=loiclec a=loiclec
This PR significantly improves the performance of the `position` ranking rule when:
1. a query contains many words
2. the `position` ranking rule needs to be called many times
3. the score of the documents according to `position` is high
These conditions greatly increase:
1. the number of edge traversals that are needed to find a valid path from the `start` node to the `end` node
2. the number of edges that need to be deleted from the graph, and therefore the number of times that we need to recompute all the possible costs from START to END
As a result, a majority of the search time is spent in `visit_condition`, `visit_node`, and `update_all_costs_before_node`. This is frustrating because it often happens when the "universe" given to the rule consists of only a handful of document ids.
By limiting the number of possible edges between two nodes from `20` to `10`, we:
1. reduce the number of possible costs from START to END
2. reduce the number of edges that will be deleted
3. make it faster to update the costs after deleting an edge
4. reduce the number of buckets that need to be computed
In terms of relevancy, I don't think we lose or gain much. We still prefer terms that are in a lower positions, with decreasing precision as we go further. The previous choice of bucketing wasn't chosen in a principled way, and neither is this one. They both "feel" right to me.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
Co-authored-by: meili-bors[bot] <89034592+meili-bors[bot]@users.noreply.github.com>
3755: Re-add final dot r=curquiza a=ManyTheFish
I removed the final dot of the error message in my last PR, this one re-adds it.
related to https://github.com/meilisearch/meilisearch/pull/3749
> Oups 😬
Co-authored-by: ManyTheFish <many@meilisearch.com>
3741: Add ngram support to the highlighter r=ManyTheFish a=loiclec
This PR fixes a bug introduced by the search refactor, where ngrams were not highlighted.
The solution was to add the ngrams to the vector of `LocatedQueryTerm` that is given to the `MatchingWords` structure.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
3749: Fix back: sort error message r=ManyTheFish a=ManyTheFish
This PR reintroduces the error message modified in https://github.com/meilisearch/milli/pull/375.
However, this added double-quotes around `sort` in the message. I don't think another message contains double-quotes, so I have added a separate commit replacing the double-quotes with back-ticks, which seems more consistent with the other error messages, this last change can be reverted easily.
## Detailed changes
#### v1.2-rc0
```
The sort ranking rule must be specified in the ranking rules settings to use the sort parameter at search time.
```
#### [Reintroduce fix (previous and expected behavior)](23d1c86825)
```
You must specify where "sort" is listed in the rankingRules setting to use the sort parameter at search time
```
#### [Replace double-quotes with back-ticks (my suggestion)](4d691d071a)
```
You must specify where `sort` is listed in the rankingRules setting to use the sort parameter at search time
```
## Related
Fixes#3722
## Reviewers
- technical review: `@irevoire`
- to validate the replacement: `@macraig`
Co-authored-by: ManyTheFish <many@meilisearch.com>
3726: Fix prefix highlighting r=loiclec a=ManyTheFish
The prefix queries were not properly highlighted, this PR now highlights only the start of a word when it matched with a prefix
Co-authored-by: ManyTheFish <many@meilisearch.com>
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
3687: Allow to disable specialized tokenizations (again) r=Kerollmops a=jirutka
In PR #2773, I added the `chinese`, `hebrew`, `japanese` and `thai` feature flags to allow melisearch to be built without huge specialed tokenizations that took up 90% of the melisearch binary size. Unfortunately, due to some recent changes, this doesn't work anymore. The problem lies in excessive use of the `default` feature flag, which infects the dependency graph.
Instead of adding `default-features = false` here and there, it's easier and more future-proof to not declare `default` in `milli` and `meilisearch-types`. I've renamed it to `all-tokenizers`, which also makes it a bit clearer what it's about.
Co-authored-by: Jakub Jirutka <jakub@jirutka.cz>
In PR #2773, I added the `chinese`, `hebrew`, `japanese` and `thai`
feature flags to allow melisearch to be built without huge specialed
tokenizations that took up 90% of the melisearch binary size.
Unfortunately, due to some recent changes, this doesn't work anymore.
The problem lies in excessive use of the `default` feature flag, which
infects the dependency graph.
Instead of adding `default-features = false` here and there, it's easier
and more future-proof to not declare `default` in `milli` and
`meilisearch-types`. I've renamed it to `all-tokenizers`, which also
makes it a bit clearer what it's about.
Conflicts | resolution
----------|-----------
Cargo.lock | added mimalloc
Cargo.toml | took origin/main version
milli/src/search/criteria/exactness.rs | deleted after checking it was only clippy changes
milli/src/search/query_tree.rs | deleted after checking it was only clippy changes
3571: Introduce two filters to select documents with `null` and empty fields r=irevoire a=Kerollmops
# Pull Request
## Related issue
This PR implements the `X IS NULL`, `X IS NOT NULL`, `X IS EMPTY`, `X IS NOT EMPTY` filters that [this comment](https://github.com/meilisearch/product/discussions/539#discussioncomment-5115884) is describing in a very detailed manner.
## What does this PR do?
### `IS NULL` and `IS NOT NULL`
This PR will be exposed as a prototype for now. Below is the copy/pasted version of a spec that defines this filter.
- `IS NULL` matches fields that `EXISTS` AND `= IS NULL`
- `IS NOT NULL` matches fields that `NOT EXISTS` OR `!= IS NULL`
1. `{"name": "A", "price": null}`
2. `{"name": "A", "price": 10}`
3. `{"name": "A"}`
`price IS NULL` would match 1
`price IS NOT NULL` or `NOT price IS NULL` would match 2,3
`price EXISTS` would match 1, 2
`price NOT EXISTS` or `NOT price EXISTS` would match 3
common query : `(price EXISTS) AND (price IS NOT NULL)` would match 2
### `IS EMPTY` and `IS NOT EMPTY`
- `IS EMPTY` matches Array `[]`, Object `{}`, or String `""` fields that `EXISTS` and are empty
- `IS NOT EMPTY` matches fields that `NOT EXISTS` OR are not empty.
1. `{"name": "A", "tags": null}`
2. `{"name": "A", "tags": [null]}`
3. `{"name": "A", "tags": []}`
4. `{"name": "A", "tags": ["hello","world"]}`
5. `{"name": "A", "tags": [""]}`
6. `{"name": "A"}`
7. `{"name": "A", "tags": {}}`
8. `{"name": "A", "tags": {"t1":"v1"}}`
9. `{"name": "A", "tags": {"t1":""}}`
10. `{"name": "A", "tags": ""}`
`tags IS EMPTY` would match 3,7,10
`tags IS NOT EMPTY` or `NOT tags IS EMPTY` would match 1,2,4,5,6,8,9
`tags IS NULL` would match 1
`tags IS NOT NULL` or `NOT tags IS NULL` would match 2,3,4,5,6,7,8,9,10
`tags EXISTS` would match 1,2,3,4,5,7,8,9,10
`tags NOT EXISTS` or `NOT tags EXISTS` would match 6
common query : `(tags EXISTS) AND (tags IS NOT NULL) AND (tags IS NOT EMPTY)` would match 2,4,5,8,9
## What should the reviewer do?
- Check that I tested the filters
- Check that I deleted the ids of the documents when deleting documents
Co-authored-by: Clément Renault <clement@meilisearch.com>
Co-authored-by: Kerollmops <clement@meilisearch.com>
3641: Bring back changes from `release v1.1.0` into `main` after v1.1.0 release r=curquiza a=curquiza
Replace https://github.com/meilisearch/meilisearch/pull/3637 since we don't want to pull commits from `main` into `release-v1.1.0` when fixing git conflicts
Co-authored-by: ManyTheFish <many@meilisearch.com>
Co-authored-by: bors[bot] <26634292+bors[bot]@users.noreply.github.com>
Co-authored-by: Charlotte Vermandel <charlottevermandel@gmail.com>
Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
Co-authored-by: curquiza <clementine@meilisearch.com>
Co-authored-by: Clément Renault <clement@meilisearch.com>
Co-authored-by: Many the fish <many@meilisearch.com>
Previously, if the primary key was set and a Settings update contained
a primary key, an error would be returned.
However, this error is not needed if the new PK == the current PK.
This commit just checks to see if the PK actually changes
before raising an error.
3568: CI: Fix `publish-aarch64` job that still uses ubuntu-18.04 r=Kerollmops a=curquiza
Fixes#3563
Main change
- add the usage of the `ubuntu-18.04` container instead of the native `ubuntu-18.04` of GitHub actions: I had to install docker in the container.
Small additional changes
- remove useless `fail-fast` and unused/irrelevant matrix inputs (`build`, `linker`, `os`, `use-cross`...)
- Remove useless step in job
Proof of work with this CI triggered on this current branch: https://github.com/meilisearch/meilisearch/actions/runs/4366233882
3569: Enhance Japanese language detection r=dureuill a=ManyTheFish
# Pull Request
This PR is a prototype and can be tested by downloading [the dedicated docker image](https://hub.docker.com/layers/getmeili/meilisearch/prototype-better-language-detection-0/images/sha256-a12847de00e21a71ab797879fd09777dadcb0881f65b5f810e7d1ed434d116ef?context=explore):
```bash
$ docker pull getmeili/meilisearch:prototype-better-language-detection-0
```
## Context
Some Languages are harder to detect than others, this miss-detection leads to bad tokenization making some words or even documents completely unsearchable. Japanese is the main Language affected and can be detected as Chinese which has a completely different way of tokenization.
A [first iteration has been implemented for v1.1.0](https://github.com/meilisearch/meilisearch/pull/3347) but is an insufficient enhancement to make Japanese work. This first implementation was detecting the Language during the indexing to avoid bad detections during the search.
Unfortunately, some documents (shorter ones) can be wrongly detected as Chinese running bad tokenization for these documents and making possible the detection of Chinese during the search because it has been detected during the indexing.
For instance, a Japanese document `{"id": 1, "name": "東京スカパラダイスオーケストラ"}` is detected as Japanese during indexing, during the search the query `東京` will be detected as Japanese because only Japanese documents have been detected during indexing despite the fact that v1.0.2 would detect it as Chinese.
However if in the dataset there is at least one document containing a field with only Kanjis like:
_A document with only 1 field containing only Kanjis:_
```json
{
"id":4,
"name": "東京特許許可局"
}
```
_A document with 1 field containing only Kanjis and 1 field containing several Japanese characters:_
```json
{
"id":105,
"name": "東京特許許可局",
"desc": "日経平均株価は26日 に約8カ月ぶりに2万4000円の心理的な節目を上回った。株高を支える材料のひとつは、自民党総裁選で3選を決めた安倍晋三首相の経済政策への期待だ。恩恵が見込まれるとされる人材サービスや建設株の一角が買われている。ただ思惑が先行して資金が集まっている面 は否めない。実際に政策効果を取り込む企業はどこか、なお未知数だ。"
}
```
Then, in both cases, the field `name` will be detected as Chinese during indexing allowing the search to detect Chinese in queries. Therefore, the query `東京` will be detected as Chinese and only the two last documents will be retrieved by Meilisearch.
## Technical Approach
The current PR partially fixes these issues by:
1) Adding a check over potential miss-detections and rerunning the extraction of the document forcing the tokenization over the main Languages detected in it.
> 1) run a first extraction allowing the tokenizer to detect any Language in any Script
> 2) generate a distribution of tokens by Script and Languages (`script_language`)
> 3) if for a Script we have a token distribution of one of the Language that is under the threshold, then we rerun the extraction forbidding the tokenizer to detect the marginal Languages
> 4) the tokenizer will fall back on the other available Languages to tokenize the text. For example, if the Chinese were marginally detected compared to the Japanese on the CJ script, then the second extraction will force Japanese tokenization for CJ text in the document. however, the text on another script like Latin will not be impacted by this restriction.
2) Adding a filtering threshold during the search over Languages that have been marginally detected in documents
## Limits
This PR introduces 2 arbitrary thresholds:
1) during the indexing, a Language is considered miss-detected if the number of detected tokens of this Language is under 10% of the tokens detected in the same Script (Japanese and Chinese are 2 different Languages sharing the "same" script "CJK").
2) during the search, a Language is considered marginal if less than 5% of documents are detected as this Language.
This PR only partially fixes these issues:
- ✅ the query `東京` now find Japanese documents if less than 5% of documents are detected as Chinese.
- ✅ the document with the id `105` containing the Japanese field `desc` but the miss-detected field `name` is now completely detected and tokenized as Japanese and is found with the query `東京`.
- ❌ the document with the id `4` no longer breaks the search Language detection but continues to be detected as a Chinese document and can't be found during the search.
## Related issue
Fixes#3565
## Possible future enhancements
- Change or contribute to the Library used to detect the Language
- the related issue on Whatlang: https://github.com/greyblake/whatlang-rs/issues/122
Co-authored-by: curquiza <clementine@meilisearch.com>
Co-authored-by: ManyTheFish <many@meilisearch.com>
Co-authored-by: Many the fish <many@meilisearch.com>
3525: Fix phrase search containing stop words r=ManyTheFish a=ManyTheFish
# Summary
A search with a phrase containing only stop words was returning an HTTP error 500,
this PR filters the phrase containing only stop words dropping them before the search starts, a query with a phrase containing only stop words now behaves like a placeholder search.
fixes https://github.com/meilisearch/meilisearch/issues/3521
related v1.0.2 PR on milli: https://github.com/meilisearch/milli/pull/779
Co-authored-by: ManyTheFish <many@meilisearch.com>
3319: Transparently resize indexes on MaxDatabaseSizeReached errors r=Kerollmops a=dureuill
# Pull Request
## Related issue
Related to https://github.com/meilisearch/meilisearch/discussions/3280, depends on https://github.com/meilisearch/milli/pull/760
## What does this PR do?
### User standpoint
- Meilisearch no longer fails tasks that encounter the `milli::UserError(MaxDatabaseSizeReached)` error.
- Instead, these tasks are retried after increasing the maximum size allocated to the index where the failure occurred.
### Implementation standpoint
- Add `Batch::index_uid` to get the `index_uid` of a batch of task if there is one
- `IndexMapper::create_or_open_index` now takes an additional `size` argument that allows to (re)open indexes with a size different from the base `IndexScheduler::index_size` field
- `IndexScheduler::tick` now returns a `Result<TickOutcome>` instead of a `Result<usize>`. This offers more explicit control over what the behavior should be wrt the next tick.
- Add `IndexStatus::BeingResized` that contains a handle that a thread can use to await for the resize operation to complete and the index to be available again.
- Add `IndexMapper::resize_index` to increase the size of an index.
- In `IndexScheduler::tick`, intercept task batches that failed due to `MaxDatabaseSizeReached` and resize the index that caused the error, then request a new tick that will eventually handle the still enqueued task.
## Testing the PR
The following diff can be applied to this branch to make testing the PR easier:
<details>
```diff
diff --git a/index-scheduler/src/index_mapper.rs b/index-scheduler/src/index_mapper.rs
index 553ab45a..022b2f00 100644
--- a/index-scheduler/src/index_mapper.rs
+++ b/index-scheduler/src/index_mapper.rs
`@@` -228,13 +228,15 `@@` impl IndexMapper {
drop(lock);
+ std:🧵:sleep_ms(2000);
+
let current_size = index.map_size()?;
let closing_event = index.prepare_for_closing();
- log::info!("Resizing index {} from {} to {} bytes", name, current_size, current_size * 2);
+ log::error!("Resizing index {} from {} to {} bytes", name, current_size, current_size * 2);
closing_event.wait();
- log::info!("Resized index {} from {} to {} bytes", name, current_size, current_size * 2);
+ log::error!("Resized index {} from {} to {} bytes", name, current_size, current_size * 2);
let index_path = self.base_path.join(uuid.to_string());
let index = self.create_or_open_index(&index_path, None, 2 * current_size)?;
`@@` -268,8 +270,10 `@@` impl IndexMapper {
match index {
Some(Available(index)) => break index,
Some(BeingResized(ref resize_operation)) => {
+ log::error!("waiting for resize end");
// Deadlock: no lock taken while doing this operation.
resize_operation.wait();
+ log::error!("trying our luck again!");
continue;
}
Some(BeingDeleted) => return Err(Error::IndexNotFound(name.to_string())),
diff --git a/index-scheduler/src/lib.rs b/index-scheduler/src/lib.rs
index 11b17d05..242dc095 100644
--- a/index-scheduler/src/lib.rs
+++ b/index-scheduler/src/lib.rs
`@@` -908,6 +908,7 `@@` impl IndexScheduler {
///
/// Returns the number of processed tasks.
fn tick(&self) -> Result<TickOutcome> {
+ log::error!("ticking!");
#[cfg(test)]
{
*self.run_loop_iteration.write().unwrap() += 1;
diff --git a/meilisearch/src/main.rs b/meilisearch/src/main.rs
index 050c825a..63f312f6 100644
--- a/meilisearch/src/main.rs
+++ b/meilisearch/src/main.rs
`@@` -25,7 +25,7 `@@` fn setup(opt: &Opt) -> anyhow::Result<()> {
#[actix_web::main]
async fn main() -> anyhow::Result<()> {
- let (opt, config_read_from) = Opt::try_build()?;
+ let (mut opt, config_read_from) = Opt::try_build()?;
setup(&opt)?;
`@@` -56,6 +56,8 `@@` We generated a secure master key for you (you can safely copy this token):
_ => (),
}
+ opt.max_index_size = byte_unit::Byte::from_str("1MB").unwrap();
+
let (index_scheduler, auth_controller) = setup_meilisearch(&opt)?;
#[cfg(all(not(debug_assertions), feature = "analytics"))]
```
</details>
Mainly, these debug changes do the following:
- Set the default index size to 1MiB so that index resizes are initially frequent
- Turn some logs from info to error so that they can be displayed with `--log-level ERROR` (hiding the other infos)
- Add a long sleep between the beginning and the end of the resize so that we can observe the `BeingResized` index status (otherwise it would never come up in my tests)
## Open questions
- Is the growth factor of x2 the correct solution? For a `Vec` in memory it makes sense, but here we're manipulating quantities that are potentially in the order of 500GiBs. For bigger indexes it may make more sense to add at most e.g. 100GiB on each resize operation, avoiding big steps like 500GiB -> 1TiB.
## PR checklist
Please check if your PR fulfills the following requirements:
- [ ] Does this PR fix an existing issue, or have you listed the changes applied in the PR description (and why they are needed)?
- [ ] Have you read the contributing guidelines?
- [ ] Have you made sure that the title is accurate and descriptive of the changes?
Thank you so much for contributing to Meilisearch!
3470: Autobatch addition and deletion r=irevoire a=irevoire
This PR adds the capability to meilisearch to batch document addition and deletion together.
Fix https://github.com/meilisearch/meilisearch/issues/3440
--------------
Things to check before merging;
- [x] What happens if we delete multiple time the same documents -> add a test
- [x] If a documentDeletion gets batched with a documentAddition but the index doesn't exist yet? It should not work
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
Co-authored-by: Tamo <tamo@meilisearch.com>
3490: Fix attributes set candidates r=curquiza a=ManyTheFish
# Pull Request
Fix attributes set candidates for v1.1.0
## details
The attribute criterion was not returning the remaining candidates when its internal algorithm was been exhausted.
We had a loss of candidates by the attribute criterion leading to the bug reported in the issue linked below.
After some investigation, it seems that it was the only criterion that had this behavior.
We are now returning the remaining candidates instead of an empty bitmap.
## Related issue
Fixes#3483
PR on milli for v1.0.1: https://github.com/meilisearch/milli/pull/777
Co-authored-by: ManyTheFish <many@meilisearch.com>
3492: Bump deserr r=Kerollmops a=irevoire
Bump deserr to the latest version;
- We now use the default actix-web extractors that deserr provides (which were copy/pasted from meilisearch)
- We also use the default `JsonError` message provided by deserr instead of defining our own in meilisearch
- Finally, we get the new `did you mean?` error message. Fix#3493
Co-authored-by: Tamo <tamo@meilisearch.com>
3461: Bring v1 changes into main r=curquiza a=Kerollmops
Also bring back changes in milli (the remote repository) into main done during the pre-release
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
Co-authored-by: bors[bot] <26634292+bors[bot]@users.noreply.github.com>
Co-authored-by: curquiza <curquiza@users.noreply.github.com>
Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: Philipp Ahlner <philipp@ahlner.com>
Co-authored-by: Kerollmops <clement@meilisearch.com>
This database can easily contain millions of entries. Thus, iterating
over it can be very expensive.
For regular `documentAdditionOrUpdate` tasks, `del_prefix_fst_words`
will always be empty. Thus, we can save a significant amount of time
by adding this `if !del_prefix_fst_words.is_empty()` condition.
The code's behaviour remains completely unchanged.
763: Fixes error message when lat and lng are unparseable r=loiclec a=ahlner
# Pull Request
## Related issue
Fixes partially [#3007](https://github.com/meilisearch/meilisearch/issues/3007)
## What does this PR do?
- Changes function validate_geo_from_json to return a BadLatitudeAndLongitude if lat or lng is a string and not parseable to f64
- implemented some unittests
- Derived PartialEq for GeoError to use assert_eq! in tests
## PR checklist
Please check if your PR fulfills the following requirements:
- [x] Does this PR fix an existing issue, or have you listed the changes applied in the PR description (and why they are needed)?
- [x] Have you read the contributing guidelines?
- [x] Have you made sure that the title is accurate and descriptive of the changes?
Thank you so much for contributing to Meilisearch!
Co-authored-by: Philipp Ahlner <philipp@ahlner.com>
764: Update deserr to latest version r=irevoire a=loiclec
Update deserr to 0.1.5, which changes the `DeserializeFromValue` trait, getting rid of the `default()` method.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
761: Integrate deserr r=irevoire a=loiclec
1. `Setting<T>` now implements `DeserializeFromValue`
2. The settings now store ranking rules as strongly typed `Criterion` instead of `String`, since the validation of the ranking rules will be done on meilisearch's side from now on
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
759: Change primary key inference error messages r=Kerollmops a=dureuill
# Pull Request
## Related issue
Milli part of https://github.com/meilisearch/meilisearch/issues/3301
## What does this PR do?
- Change error message strings
## PR checklist
Please check if your PR fulfills the following requirements:
- [x] Does this PR fix an existing issue, or have you listed the changes applied in the PR description (and why they are needed)?
- [x] Have you read the contributing guidelines?
- [x] Have you made sure that the title is accurate and descriptive of the changes?
Thank you so much for contributing to Meilisearch!
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
733: Avoid a prefix-related worst-case scenario in the proximity criterion r=loiclec a=loiclec
# Pull Request
## Related issue
Somewhat fixes (until merged into meilisearch) https://github.com/meilisearch/meilisearch/issues/3118
## What does this PR do?
When a query ends with a word and a prefix, such as:
```
word pr
```
Then we first determine whether `pre` *could possibly* be in the proximity prefix database before querying it. There are then three possibilities:
1. `pr` is not in any prefix cache because it is not the prefix of many words. We don't query the proximity prefix database. Instead, we list all the word derivations of `pre` through the FST and query the regular proximity databases.
2. `pr` is in the prefix cache but cannot be found in the proximity prefix databases. **In this case, we partially disable the proximity ranking rule for the pair `word pre`.** This is done as follows:
1. Only find the documents where `word` is in proximity to `pre` **exactly** (no derivations)
2. Otherwise, assume that their proximity in all the documents in which they coexist is >= 8
3. `pr` is in the prefix cache and can be found in the proximity prefix databases. In this case we simply query the proximity prefix databases.
Note that if a prefix is longer than 2 bytes, then it cannot be in the proximity prefix databases. Also, proximities larger than 4 are not present in these databases either. Therefore, the impact on relevancy is:
1. For common prefixes of one or two letters: we no longer distinguish between proximities from 4 to 8
2. For common prefixes of more than two letters: we no longer distinguish between any proximities
3. For uncommon prefixes: nothing changes
Regarding (1), it means that these two documents would be considered equally relevant according to the proximity rule for the query `heard pr` (IF `pr` is the prefix of more than 200 words in the dataset):
```json
[
{ "text": "I heard there is a faster proximity criterion" },
{ "text": "I heard there is a faster but less relevant proximity criterion" }
]
```
Regarding (2), it means that two documents would be considered equally relevant according to the proximity rule for the query "faster pro":
```json
[
{ "text": "I heard there is a faster but less relevant proximity criterion" }
{ "text": "I heard there is a faster proximity criterion" },
]
```
But the following document would be considered more relevant than the two documents above:
```json
{ "text": "I heard there is a faster swimmer who is competing in the pro section of the competition " }
```
Note, however, that this change of behaviour only occurs when using the set-based version of the proximity criterion. In cases where there are fewer than 1000 candidate documents when the proximity criterion is called, this PR does not change anything.
---
## Performance
I couldn't use the existing search benchmarks to measure the impact of the PR, but I did some manual tests with the `songs` benchmark dataset.
```
1. 10x 'a':
- 640ms ⟹ 630ms = no significant difference
2. 10x 'b':
- set-based: 4.47s ⟹ 7.42 = bad, ~2x regression
- dynamic: 1s ⟹ 870 ms = no significant difference
3. 'Someone I l':
- set-based: 250ms ⟹ 12 ms = very good, x20 speedup
- dynamic: 21ms ⟹ 11 ms = good, x2 speedup
4. 'billie e':
- set-based: 623ms ⟹ 2ms = very good, x300 speedup
- dynamic: ~4ms ⟹ 4ms = no difference
5. 'billie ei':
- set-based: 57ms ⟹ 20ms = good, ~2x speedup
- dynamic: ~4ms ⟹ ~2ms. = no significant difference
6. 'i am getting o'
- set-based: 300ms ⟹ 60ms = very good, 5x speedup
- dynamic: 30ms ⟹ 6ms = very good, 5x speedup
7. 'prologue 1 a 1:
- set-based: 3.36s ⟹ 120ms = very good, 30x speedup
- dynamic: 200ms ⟹ 30ms = very good, 6x speedup
8. 'prologue 1 a 10':
- set-based: 590ms ⟹ 18ms = very good, 30x speedup
- dynamic: 82ms ⟹ 35ms = good, ~2x speedup
```
Performance is often significantly better, but there is also one regression in the set-based implementation with the query `b b b b b b b b b b`.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
732: Interpret synonyms as phrases r=loiclec a=loiclec
# Pull Request
## Related issue
Fixes (when merged into meilisearch) https://github.com/meilisearch/meilisearch/issues/3125
## What does this PR do?
We now map multi-word synonyms to phrases instead of loose words. Such that the request:
```
btw I am going to nyc soon
```
is interpreted as (when the synonym interpretation is chosen for both `btw` and `nyc`):
```
"by the way" I am going to "New York City" soon
```
instead of:
```
by the way I am going to New York City soon
```
This prevents queries containing multi-word synonyms to exceed to word length limit and degrade the search performance.
In terms of relevancy, there is a debate to have. I personally think this could be considered an improvement, since it would be strange for a user to search for:
```
good DIY project
```
and have a result such as:
```
{
"text": "whether it is a good project to do, you'll have to decide for yourself"
}
```
However, for synonyms such as `NYC -> New York City`, then we will stop matching documents where `New York` is separated from `City`. This is however solvable by adding an additional mapping: `NYC -> New York`.
## Performance
With the old behaviour, some long search requests making heavy uses of synonyms could take minutes to be executed. This is no longer the case, these search requests now take an average amount of time to be resolved.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
736: Update charabia r=curquiza a=ManyTheFish
Update Charabia to the last version.
> We are now Romanizing Chinese characters into Pinyin.
> Note that we keep the accent because they are in fact never typed directly by the end-user, moreover, changing an accent leads to a different Chinese character, and I don't have sufficient knowledge to forecast the impact of removing accents in this context.
Co-authored-by: ManyTheFish <many@meilisearch.com>
709: Optimise the `ExactWords` sub-criterion within `Exactness` r=loiclec a=loiclec
# Pull Request
## Related issue
Fixes (partially) https://github.com/meilisearch/meilisearch/issues/3116
## What does this PR do?
1. Reduces the algorithmic complexity of finding the documents containing N exact words from something that is exponential to something that is polynomial.
2. Cache intermediary results between different calls to the `exactness` criterion.
## Performance Results
On the `smol_songs.csv` dataset, a request containing 10 common words now takes about 60ms instead of 5 seconds to execute. For example, this is the case with this (admittedly nonsensical) request: `Rock You Hip Hop Folk World Country Electronic Love The`.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
Displays log message in the form:
```
[2022-12-21T09:19:42Z INFO milli::update::index_documents::enrich] Primary key was not specified in index. Inferred to 'id'
```
742: Add a "Criterion implementation strategy" parameter to Search r=irevoire a=loiclec
Add a parameter to search requests which determines the implementation strategy of the criteria. This can be either `set-based`, `iterative`, or `dynamic` (ie choosing between set-based or iterative at search time). See https://github.com/meilisearch/milli/issues/755 for more context about this change.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
743: Fix finite pagination with placeholder search r=Kerollmops a=ManyTheFish
this bug is reproducible on real datasets and is hard to isolate in a simple test.
related to: https://github.com/meilisearch/meilisearch/issues/3200
poke `@curquiza`
Co-authored-by: ManyTheFish <many@meilisearch.com>
728: Add some integration tests on the sort criterion r=ManyTheFish a=loiclec
This is simply an integration test ensuring that the sort criterion works properly.
However, only one version of the algorithm is tested here (the iterative one). To test the version that uses the facet DB, one has to manually set the `CANDIDATES_THRESHOLD` constant to `0`. I have done that and ensured that the test still succeeds. However, in the future, we will probably want to have an option to force which algorithm is used at runtime, for testing purposes.
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
737: Fix typo initial candidates computation r=Kerollmops a=ManyTheFish
When `Typo` criterion was after a different criterion than `Words` and the previous criterion wasn't returning any candidates at the first iteration of the bucket sort, then the `initial_candidates` were lost.
Now, `Typo`ensure to keep the `initial_candidates` between iterations.
related to https://github.com/meilisearch/meilisearch/issues/3200#issuecomment-1345179578
related to https://github.com/meilisearch/meilisearch/issues/3228
Co-authored-by: ManyTheFish <many@meilisearch.com>
By creating snapshots and updating the format of the existing
snapshots. The next commit will apply the fix, which will show
its effects cleanly on the old and new snapshot tests
723: Fix bug in handling of soft deleted documents when updating settings r=Kerollmops a=loiclec
# Pull Request
## Related issue
Fixes (partially, until merged into meilisearch) https://github.com/meilisearch/meilisearch/issues/3021
## What does this PR do?
This PR fixes the bug where a `missing key in documents database` internal error message could appear when indexing documents.
When updating the settings, before clearing the database and before creating the transform output, we now modify the `ExternalDocumentsIds` structure to get rid of all references to soft deleted document ids in its FSTs.
It used to be that updating the settings would clear the soft-deleted document ids, but keep the original `ExternalDocumentsIds` structure. As a consequence of this, when processing a future document addition, we could wrongly believe that a document was being replaced when, in fact, it was a completely new document. See the tests `bug_3021_first`, `bug_3021_second`, and `bug_3021` for a minimal test case that would have reproduced the issue.
We need to take special care to:
- evaluate how users should update to v0.30.1 (containing this fix): dump? reimporting all documents from scratch?
- understand IF/HOW this bug could have caused duplicate documents to be returned
- and evaluate the correctness of the fix, of course :)
Co-authored-by: Loïc Lecrenier <loic.lecrenier@me.com>
719: Add more members of `filter_parser` to `milli::` & `From<&str>` implementation for `Token` r=Kerollmops a=GregoryConrad
## What does this PR do?
The current `milli::Filter` and `milli::FilterCondition` APIs require working with some members of `filter_parser` directly that `milli::` does *not* re-export to its users (at least when not parsing input using `parse`). Also, using `filter_parser` does not make sense when using milli from an embedded context where there is no query to parse.
Instead of reworking `milli::Filter` and `milli::FilterCondition`, this PR adds two non-breaking changes that ease the use of milli:
- Re-exports more members of the dependent version of `filter_parser` in `milli`
- Implements `From<&str>` for `filter_parser::Token`
- This will also allow some basic tests that need to create a `Token` from a string to avoid some boilerplate.
In conjunction, both of these will allow milli users to easily create a `Token` from a `&str` without needing to add `filter_parser` as an extra dependency.
Note: I wanted to use `FromStr` for the `From` implementation; however, it requires returning a `Result` which is not needed for the conversion. Thus, I just left it as `From<&str>`.
Co-authored-by: Gregory Conrad <gregorysconrad@gmail.com>
