This commit reverts PR #30214, which addressed bug MM-60790 but caused a
performance regression tracked by MM-66782.
This revert has two implications:
1. The performance issue is solved.
2. The original bug is re-introduced.
Re-introducing the original bug seems not to be ideal, but I argue that
the original PR did not actually fix the bug:
- Before that PR, looking for a quoted string would return additional
results: the UX was slightly confusing, because when the user looked
for the word "stateful", the results would contain matches like
"states" (see MM-60790).
- After that PR, looking for a quoted string can timeout, so that the
list of results becomes empty. The UX here may be less confusing,
since the user simply doesn't find what they're looking for, and they
may assume that string is not present in any post, but it's completely
wrong: the result list is empty because the SQL query timed out and
thus the endpoint returned 0 results.
The solution to the original issue should be addressed via
Elasticsearch, which should provide a more refined and precise search
results.
For more information on the investigation on this issue and the
motivation behind the revert, see
https://mattermost.atlassian.net/wiki/x/IYAk_w
Co-authored-by: Mattermost Build <build@mattermost.com>
This PR adds functionality to search by a global search prefix.
This allows Mattermost to be used across multiple data centers
with multiple Elasticsearch instances synchronized using
cross-cluster replication.
While here, we also add tests cases to cover for some missing
search interface methods.
For now, no system console setting is exposed. Because IndexPrefix
is also not exposed. It can be added later if a need arises.
https://mattermost.atlassian.net/browse/MM-63298
```release-note
A new config setting ElasticsearchSettings.GlobalSearchPrefix is added
which can be used to search across multiple indices having a common prefix.
This is useful in a scenario with multiple Elasticsearch instances, where
multiple instances are writing to different indices with different prefixes
using the ElasticsearchSettings.IndexPrefix setting.
```
Co-authored-by: Mattermost Build <build@mattermost.com>
When a search term is a fully quoted string, we want to avoid
stemming if it is supported by the search backend.
For Postgres, it does provide a feature by which if we use the
"simple" search config, then no stemming is performed and an exact
match with the word is done without having to resort
to LIKE queries.
Unfortunately, for ES/OS this is not an option because
the message field is a text field, which means ES/OS will analyze it,
stem it and store it in its root form. Therefore, no exact match
can be possible with ES/OS.
The only solution here is to have yet another keyword field
for message which will store it in its raw form. But this
will effectively double the disk storage for post indices
and not a good design choice.
Ref: https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-term-query.html#avoid-term-query-text-fieldshttps://mattermost.atlassian.net/browse/MM-60790
```release-note
NONE
```
* re-arrange the tests to run only on DB
```release-note
NONE
```
---------
Co-authored-by: Mattermost Build <build@mattermost.com>
* poc - wip
* add search files across teams
* eslint
* fix existing tests
* fix webapp style
* fix test
* add api doc
* change initial state in test
* add tests on API
* add tests on file info layer
* fix file search tags
* add rhs reducer test
* reset team selected when the RHS is suppressed
* change css to reflect UI
* fix style
* fix doc wording
* make getSearchTeam return currentTeamId when value is not set
* await is unnecessary
* revert boolean check and add test
* add comment to getSearchTeam to let dev knows it defaults to currentTeam
* remove redundant team check
* simplfy test
* fix style check
---------
Co-authored-by: Caleb Roseland <caleb@calebroseland.com>
Co-authored-by: Mattermost Build <build@mattermost.com>
* Fixed the issue for DB layer, ES to go
* Handled channel bookmarks
* Handled Bleve
* Lint fix
* Added channel bookmark test
* Skip bleve test
* Used common function
* SKipping ES as indexing logic in unavailable in test
* added rctx to reuired files
* added more file changes after issues on startup
* added context to draft_store
---------
Co-authored-by: Mattermost Build <build@mattermost.com>
* Refactor: Use generic to change func StringInSlice to Contains
* Refactor: Use generic to change func stringNotInSlice to Contains
Make utils.go (dir server/public/utils) and func Contains
* Refactor: Move func Contains from channels/utils to public/utils
Move func Contains from channels/utils to public/utils
Fix import declarations line
* Docs: Add a description of the Contains function
* Test: add TestContains
Add a test code for a Contain function
https://mattermost.atlassian.net/browse/MM-52532
- Replace golint with revive
- Add makezero linter
- Fix all the required linter failures
Some issues in enterprise and public modules
are yet to be fixed. We send this to expediate things.
It was a good decision in hindsight to keep the public module as 0.x
because this would have been a breaking change again.
https://mattermost.atlassian.net/browse/MM-53032
```release-note
Changed the Go module path from github.com/mattermost/mattermost-server/server/v8 to github.com/mattermost/mattermost/server/v8.
For the public facing module, it's path is also changed from github.com/mattermost/mattermost-server/server/public to github.com/mattermost/mattermost/server/public
```
* MM-45956: Optimize FileInfo stats query
We Denormalize Post.ChannelId on FileInfo.ChannelId
```release-note
The file info stats query is now optimized by denormalizing the channelID column into the table itself. This will speed up the query to get the file count for a channel on clicking the RHS.
Migration times:
On a MySQL 8.0.31 DB with
1405 rows in FileInfo and 11M posts, it took around 0.3s
On a Postgres 12.14 DB with
1731 rows in FileInfo and 11M posts, it took around 0.27s
```
https://mattermost.atlassian.net/browse/MM-45956
