- DistributionShift in Search object (to be set from model in embed?)
- Fix issue where embedder index wasn't computed at search time
- Accept as default embedder either the "default" one, or the only embedder when there is only one
4223: Update to heed 0.20 r=dureuill a=Kerollmops
This PR brings the v0.20-alpha.9 version of heed into Meilisearch 🎉 The main goal is to test it in a real environment to make the necessary changes if needed. We also want to merge it as soon as possible during the pre-release phase to ensure we catch bugs before the release.
Most of the calls to heed are the same as before, except:
- The `PolyDatabase` has been replaced with a `Database<Unspecified, Unspecified>`. We replaced the `get<T, U>()` by a `remap<T, U>().get()` calls.
- The `Database` `append(...)` method has been replaced with a `put_with_flags(PutFlags::APPEND, ...)`.
- The `RwTxn<'e, 'p>` has been simplified into a `RwTxn<'e>`.
- The `BytesEncode/Decode` traits return a `Result<_, BoxedError>` instead of an `Option<_>`.
- We no longer need to wrap and unwrap the `BEU32` integer when storing/getting them from heed.
### TODO
- [x] Create actual, simple error types instead of using strings in the codecs.
### Follow-up work
- Move the codecs into another member crate (we depend on the uuid one in the meilitool crate).
- Display the internal decoding error in the `SerializationError` internal error variant.
Co-authored-by: Clément Renault <clement@meilisearch.com>
3907: Add telemetry for define field to search on at query time r=dureuill a=ManyTheFish
Add "attributes_to_search_on" telemetry usage counter:
```json
"attributes_to_search_on": {
"total_number_of_use": 12,
},
```
This measures the number of search queries that the user uses `attributesToSearchOn` field.
related to https://github.com/meilisearch/specifications/pull/251
## reviewers:
- `@macraig` for validating the telemetry's name
- `@dureuill` for validating the code
Co-authored-by: ManyTheFish <many@meilisearch.com>
3889: Display the total number of tasks matching a filter/query r=dureuill a=Kerollmops
This PR returns a new field on the `/tasks` routes. The `total` field exposes the total number of tasks that matches the given filter/query. It is useful to display information on a user interface and can help understand when progress is made in processing tasks, i.e., the total number of tasks on `/tasks?statuses=succeeded` will increase over time.
Fixes#3888.
- [ ] Update the specs fo the `/tasks` route.
## How have I implemented it?
I found it much easier to run two times the task filtering system. Once with the original `from` and `limit` parameters and a second time without. The second call will return the total number of tasks that match the query, not only the number of tasks on the current page.
So far, in terms of performance, there doesn't seem to be any issue. I tried different filters with something like 250k tasks. Note that there is a limit of 1M tasks in the queue.
Co-authored-by: Clément Renault <clement@meilisearch.com>
3891: Fix the way we compute the 99th percentile r=dureuill a=Kerollmops
This PR fixes how we compute the 99th percentile by avoiding using float and doing the multiplication and divisions in the correct order avoiding going out of the buffer of timings. You can see the issue on [this rust playground](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021).
When there are a very small number of successful requests, the number is so tiny that the 99th percentile calculus sometimes gives an index out of the buffer. In this example, the `1`/`1.0` represent the number of timings you collected (one). As you can see, the float computation gives us the index `1.0`, with is out of a vector of only one value. This makes the engine generate a `null` value.
```rust
1 * 99 / 100 = 0 // with integers
0.99_f64 * (1.0 - 1.0) + 1.0 = 1.0 // with floats
```
Co-authored-by: Clément Renault <clement@meilisearch.com>
3877: update the total_received properties of multiple events r=dureuill a=dureuill
# Pull Request
## Related issue
Fixes#3814
## What does this PR do?
-fix name of `total_received` for several events
Co-authored-by: Tamo <tamo@meilisearch.com>
3851: Expose lastUpdate and isIndexing in /stats endpoint r=dureuill a=gentcys
# Pull Request
## Related issue
Fixes#3843
## What does this PR do?
- expose lastUpdate in `/stats` endpoint
- expose isIndex in `stats` endpoint
- add a method `is_task_processing` in index-scheduler/src/lib.rs.
## 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: Cong Chen <cong.chen@ocrlabs.com>
Co-authored-by: ManyTheFish <many@meilisearch.com>
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
3867: Add a new link to the cloud pricing page r=curquiza a=Kerollmops
This PR promotes the Cloud by adding a link to the Pricing page to the startup message!
<img width="1002" alt="Capture d’écran 2023-06-29 à 17 40 22" src="https://github.com/meilisearch/meilisearch/assets/3610253/b0528c24-fcc2-43ff-a6a1-3ed91716663b">
Co-authored-by: Clément Renault <clement@meilisearch.com>
3866: Update charabia v0.8.0 r=dureuill a=ManyTheFish
# Pull Request
Update Charabia:
- enhance Japanese segmentation
- enhance Latin Tokenization
- words containing `_` are now properly segmented into several words
- brackets `{([])}` are no more considered as context separators so word separated by brackets are now considered near together for the proximity ranking rule
- fixes#3815
- fixes#3778
- fixes [product#151](https://github.com/meilisearch/product/discussions/151)
> Important note: now the float numbers are segmented around the `.` so `3.22` is segmented as [`3`, `.`, `22`] but the middle dot isn't considered as a hard separator, which means that if we search `3.22` we find documents containing `3.22`
Co-authored-by: ManyTheFish <many@meilisearch.com>
3864: Remove `/experimental-features` verbs that weren't in the PRD r=dureuill a=dureuill
Removes:
- POST `/experimental-features`
- DELETE `/experimental-features`
keeping only:
- PATCH `/experimental-features`
- GET `/experimental-features`
The two routes that are described in the PRD.
Following `@guimachiavelli's` [question](https://github.com/meilisearch/documentation/issues/2482#issuecomment-1611845372) about the POST route.
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
3861: Add "meilisearch" prefix to last metrics that were missing it r=Kerollmops a=dureuill
# Pull Request
## Related issue
Related to #3790
## What does this PR do?
- change implementation to follow the spec on metrics name
- regenerate grafana dashboard from the code
## 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!
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
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>
3745: tests: add unit test for `PayloadTooLarge` error r=curquiza a=cymruu
# Pull Request
Add a unit test for the `Payload`, which verifies that a request with a payload that is too large is rejected with the appropriate message.
This was requested in this PR https://github.com/meilisearch/meilisearch/pull/3739
## Related issue
https://github.com/meilisearch/meilisearch/pull/3739
## What does this PR do?
- Adds requested test
## 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!
Co-authored-by: Filip Bachul <filipbachul@gmail.com>
3738: Add analytics on the get documents resource r=dureuill a=irevoire
# Pull Request
## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/3737
Related spec https://github.com/meilisearch/specifications/pull/234
## What does this PR do?
Add the analytics for the following routes:
- `GET` - `/indexes/:uid/documents`
- `GET` - `/indexes/:uid/documents/:doc_id`
- `POST` - `/indexes/:uid/documents/fetch`
These analytics are aggregated between two events:
- `Documents Fetched GET`
- `Documents Fetched POST`
That shares the same payload:
Property name | Description | Example |
|---------------|-------------|---------|
| `requests.total_received` | Total number of request received in this batch | 325 |
| `per_document_id` | `false` | false |
| `per_filter` | `true` if `POST /indexes/:indexUid/documents/fetch` endpoint was used with a filter in this batch, otherwise `false` | false |
| `pagination.max_limit` | Highest value given for the `limit` parameter in this batch | 60 |
| `pagination.max_offset` | Highest value given for the `offset` parameter in this batch | 1000 |
Co-authored-by: Tamo <tamo@meilisearch.com>
3550: Delete documents by filter r=irevoire a=dureuill
# Prototype `prototype-delete-by-filter-0`
Usage:
A new route is available under `POST /indexes/{index_uid}/documents/delete` that allows you to delete your documents by filter.
The expected payload looks like that:
```json
{
"filter": "doggo = bernese",
}
```
It'll then enqueue a task in your task queue that'll delete all the documents matching this filter once it's processed.
Here is an example of the associated details;
```json
"details": {
"deletedDocuments": 53,
"originalFilter": "\"doggo = bernese\""
}
```
----------
# Pull Request
## Related issue
Related to https://github.com/meilisearch/meilisearch/issues/3477
## What does this PR do?
### User standpoint
- Modifies the `/indexes/{:indexUid}/documents/delete-batch` route to accept either the existing array of documents ids, or a JSON object with a `filter` field representing a filter to apply. If that latter variant is used, any document matching the filter will be deleted.
### Implementation standpoint
- (processing time version) Adds a new BatchKind that is not autobatchable and that performs the delete by filter
- Reuse the `documentDeletion` task with a new `originalFilter` detail that replaces the `providedIds` detail.
## Example
<details>
<summary>Sample request, response and task result</summary>
Request:
```
curl \
-X POST 'http://localhost:7700/indexes/index-10/documents/delete-batch' \
-H 'Content-Type: application/json' \
--data-binary '{ "filter" : "mass = 600"}'
```
Response:
```
{
"taskUid": 3902,
"indexUid": "index-10",
"status": "enqueued",
"type": "documentDeletion",
"enqueuedAt": "2023-02-28T20:50:31.667502Z"
}
```
Task log:
```json
{
"uid": 3906,
"indexUid": "index-12",
"status": "succeeded",
"type": "documentDeletion",
"canceledBy": null,
"details": {
"deletedDocuments": 3,
"originalFilter": "\"mass = 600\""
},
"error": null,
"duration": "PT0.001819S",
"enqueuedAt": "2023-03-07T08:57:20.11387Z",
"startedAt": "2023-03-07T08:57:20.115895Z",
"finishedAt": "2023-03-07T08:57:20.117714Z"
}
```
</details>
## Draft status
- [ ] Error handling
- [ ] Analytics
- [ ] Do we want to reuse the `delete-batch` route in this way, or create a new route instead?
- [ ] Should the filter be applied at request time or when the deletion task is processed?
- The first commit in this PR applies the filter at request time, meaning that even if a document is modified in a way that no longer matches the filter in a later update, it will be deleted as long as the deletion task is processed after that update.
- The other commits in this PR apply the filter only when the asynchronous deletion task is processed, meaning that documents that match the filter at processing time are deleted even if they didn't match the filter at request time.
- [ ] If keeping the filter at request time, find a more elegant way to recover the user document ids from the internal document ids. The current way implemented in the first commit of this PR involves getting all the documents matching the filter, looking for the value of their primary key, and turning it into a string by copy-pasting routines found in milli...
- [ ] Security consideration, if any
- [ ] Fix the tests (but waiting until product questions are resolved)
- [ ] Add delete by filter specific tests
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
Co-authored-by: Tamo <tamo@meilisearch.com>
3688: Following release v1.1.1: bring back changes into `main` r=curquiza a=curquiza
`@meilisearch/engine-team` ensure the changes we bring to `main` are the ones you want
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
Co-authored-by: bors[bot] <26634292+bors[bot]@users.noreply.github.com>
Co-authored-by: Tamo <tamo@meilisearch.com>
Co-authored-by: dureuill <dureuill@users.noreply.github.com>
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>
3529: Add an analytics on the geo bounding box feature r=ManyTheFish a=irevoire
Fixes#3527
[The specification of the geoBoundingBox](https://github.com/meilisearch/specifications/pull/223) feature has been updated and now introduces a new analytics to follow the usage of the geoBoundingBox feature in the search requests.
Co-authored-by: Tamo <tamo@meilisearch.com>
3524: Update the metrics route r=irevoire a=irevoire
Fixes#3523
Make the metrics available by default without a feature flag.
+ Rename the cli-flag to `experimental-enable-metrics`.
Co-authored-by: Tamo <tamo@meilisearch.com>
3331: Limit the number of concurrently opened indexes r=dureuill a=dureuill
# Pull Request
## Related issue
Relevant to #1841, fixes#3382
## What does this PR do?
### User standpoint
- Limit the number of concurrently opened indexes (currently, the number of indexes that can be concurrently opened is computed at startup)
- When too many an index is opened, the least recently used one is closed and its virtual memory released.
- This allows a user to have an arbitrary number of indexes of an arbitrary size
### Implementation standpoint
- Added a LRU cache map in `index-scheduler::lru`. A more complete implementation (eg with helper functions not used here) is available but would better fit a dedicated crate.
- Use the LRU cache map in the `IndexScheduler`. To simplify the lifecycle of indexes, they are never removed from the cache when they are in the middle of a resize or delete operation. To achieve this, an intermediate `Vec` stores the UUIDs of the indexes that are in the middle of such an operation.
- Upon creating the index scheduler object, compute the total virtual memory that is adressable by using a dichotomic search on the max size of an index. Use this as a base to compute the number of indexes that can be open with 2TiB per index. If the virtual memory address space is lower than 2TiB, then only allow for 1 index of a fraction of that size.
Co-authored-by: Louis Dureuil <louis@meilisearch.com>
