4137: Update version for the next release (v1.4.2) in Cargo.toml r=curquiza a=meili-bot
⚠️ This PR is automatically generated. Check the new version is the expected one and Cargo.lock has been updated before merging.
Co-authored-by: curquiza <curquiza@users.noreply.github.com>
4108: Fix bug where search with distinct attribute and no ranking, returns offset+limit hits r=curquiza a=vivek-26
# Pull Request
## Related issue
Fixes#4078
## What does this PR do?
This PR -
- Fixes bug where search with distinct attribute and no ranking, returns offset+limit hits.
- Adds unit and integration 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: Vivek Kumar <vivek.26@outlook.com>
4089: Use a bufreader and bufwriter everytime there is a grenad<file> r=curquiza a=irevoire
# Pull Request
Wrap all the files we give to a grenad in a `BufReader` or `BufWriter`.
The dump import I tried in the issue went from 2h to 10 minutes on my machine.
I also ran a bunch of benchmarks on my machine, and we're faster by a few seconds everywhere but nothing huge.
-----
The one thing I’m afraid about is if we used to get the inner file in a grenad and then do a read right after without a seek at the beginning of the file or a reopen.
Since we now use a bufreader our read would return the bytes one buffer later and probably completely corrupt what we were supposed to read.
From what I see, it looks like it works, but I may have missed something, I don't know much about this part of the codebase.
This issue should not arise on the bufwriter, though, because if we're not able to write the content of the buffer I ensured that the `into_inner` of the bufwriter should return an internal error.
## Related issue
Fixes#4087
Co-authored-by: Tamo <tamo@meilisearch.com>
4112: Update version for the next release (v1.4.1) in Cargo.toml r=curquiza a=meili-bot
⚠️ This PR is automatically generated. Check the new version is the expected one and Cargo.lock has been updated before merging.
Co-authored-by: curquiza <curquiza@users.noreply.github.com>
4053: Fix the stats of the documents deletion by filter r=Kerollmops a=irevoire
# Pull Request
The issue was that the operation « DocumentDeletionByFilter » was not declared as an index operation. That means the index stats were not reprocessed after the application of the operation.
## Related issue
Fixes#4018
## What does this PR do?
- Move the `DocumentDeletionByFilter` internal operation into the category of the `IndexOperation`. This means that the stats will automatically be re-processed after a batch is processed.
- Update a test to ensure that the stats are valid after each operation
## 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: Tamo <tamo@meilisearch.com>
3997: Refactor empty arrays/objects should return empty instead of null r=Kerollmops a=dogukanakkaya
# Pull Request
## What does this PR do?
At the moment if we select empty objects and array of object properties with dot notations like:
```json
{
"array": [],
"object": {}
}
```
```rs
GetDocumentOptions { fields: Some(vec!["array.name", "object.name"]) }
```
returns null if the array/object has no property yet.
I am not sure if this is expected or it's the correct behaviour but I add my document with a property that is assigned to an empty array/object, later on when I select it, returns null which is kinda weird and unexpected in my opinion.
This PR fixes that issue by returning an empty vector if the array is empty or an empty map if object is empty. This is not added for `permissive-json-pointer/src/lib.rs:224` because `create_array` loops over each item. Selecting a single property that is an object, in an array of objects would result other objects to be empty maps instead of none.
```json
"doggos": [
{
"jean": {
"race": {
"name": "bernese mountain",
}
}
},
{
"marc": {
"age": 4,
"race": {
"name": "golden retriever",
}
}
}
]
```
```rs
GetDocumentOptions { fields: Some(vec!["doggos.jean"]) }
```
Would result in `jean` object and an extra empty object for `marc`.
## 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)?
- [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: dogukanakkaya <doguakkaya27@hotmail.com>
The issue was that the operation « DocumentDeletionByFilter » was not
declared as an index operation. That means the indexes stats were not
reprocessed after the application of the operation.
4028: Fix highlighting bug when searching for a phrase with cropping r=ManyTheFish a=vivek-26
# Pull Request
## Related issue
Fixes#3975
## What does this PR do?
This PR -
- Fixes the bug where searching **only** for a phrase (containing multiple words) along with cropping, highlighted only the first word of the phrase.
- Adds unit test case for the above mentioned scenario.
## 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: Vivek Kumar <vivek.26@outlook.com>
4041: Register the swap indexe task in a spawn blocking to be sure to never… r=ManyTheFish a=irevoire
# Pull Request
## Related issue
Fixes https://github.com/meilisearch/meilisearch/issues/4040
## What does this PR do?
- Register the swap indexes task in a spawn blocking task
Co-authored-by: Tamo <tamo@meilisearch.com>
4039: Fix multiple vectors dimensions r=ManyTheFish a=Kerollmops
This PR fixes#4035, making providing multiple vectors in documents possible. This is fixed by extracting the vectors from the non-flattened version of the documents.
Co-authored-by: Kerollmops <clement@meilisearch.com>
4038: Fix filter escaping issues r=ManyTheFish a=Kerollmops
This PR fixes#4034 by always escaping the sequences. Users must always put quotes (simple or double) to escape the filter values.
Co-authored-by: Kerollmops <clement@meilisearch.com>
4037: Update version for the next release (v1.3.3) in Cargo.toml r=curquiza a=meili-bot
⚠️ This PR is automatically generated. Check the new version is the expected one and Cargo.lock has been updated before merging.
Co-authored-by: curquiza <curquiza@users.noreply.github.com>
3994: Fix synonyms with separators r=Kerollmops a=ManyTheFish
# Pull Request
## Related issue
Fixes#3977
## Available prototype
```
$ docker pull getmeili/meilisearch:prototype-fix-synonyms-with-separators-0
```
## What does this PR do?
- add a new test
- filter the empty synonyms after normalization
Co-authored-by: ManyTheFish <many@meilisearch.com>
4016: Define the full Homebrew formula path r=curquiza a=Kerollmops
This PR fixes#4015 by defining the full Homebrew formula path.
Co-authored-by: Clément Renault <clement@meilisearch.com>
4020: Update version for the next release (v1.4.0) in Cargo.toml r=Kerollmops a=meili-bot
⚠️ This PR is automatically generated. Check the new version is the expected one and Cargo.lock has been updated before merging.
Co-authored-by: Kerollmops <Kerollmops@users.noreply.github.com>
Co-authored-by: Clément Renault <clement@meilisearch.com>
4013: Fix the ranking rule by temporarily disabling an assert in the bucket sort algorithm r=Kerollmops a=Kerollmops
This PR temporarily disables an assertion, making the search crash. [I created a tracking issue](https://github.com/meilisearch/meilisearch/issues/4012) to find a better way to fix this.
It no longer reverts a20e4d447c, which seemed to generate unreachable graphs and make the bucket sort ranking algorithm panic because of entering an unreachable state. We discussed that below in the comments.
Temporary fixes#4002, fixes#4006, and fixes#3995.
---
It took me approximately 2 days to find the first bad commit just because I'm bad in `git bisect` x `bash`, i.e. [I misused `%1` with `$!` to kill the most recently backgrounded job](https://unix.stackexchange.com/a/340084/212574)...
<details>
<summary>Here is the script I used to find the invalid commit</summary>
```bash
#!/usr/bin/env bash
set -x
# remove the data
rm -rf data.ms
# build meilisearch
cargo build --release
# ignore this commit if it doesn't compile
if [[ $? != 0 ]]; then
exit 125
fi
# index the dump and start from it
./target/release/meilisearch \
--http-addr 'localhost:7705' \
--import-dump $HOME/Downloads/modified-20230822-083016113.dump &
# wait 10 sec while it indexes the docs
sleep 5
# check if the server crashes on requests
echo '{
"q": "rtx 305",
"attributesToHighlight": [
"*"
],
"highlightPreTag": "<ais-highlight-0000000000>",
"highlightPostTag": "</ais-highlight-0000000000>",
"limit": 21,
"offset": 0
}' | xh 'localhost:7705/indexes/arvutitark_local_orderables/search'
last_exit_code=$?
# Now kill Meilisearch
kill $!
# Clean the potential Cargo.lock
git checkout .
exit $last_exit_code
```
</details>
Co-authored-by: Kerollmops <clement@meilisearch.com>
Co-authored-by: Clément Renault <clement@meilisearch.com>
