Merge branch 'feat/allow_alt_text_search_config' into 'develop'

allow user defined text search config in database

See merge request pleroma/pleroma!3275

2 files changed, 60 insertions, 0 deletions

diff --git a/docs/administration/CLI_tasks/database.md b/docs/administration/CLI_tasks/database.md
index 6dca83167..c53c49921 100644
--- a/docs/administration/CLI_tasks/database.md
+++ b/docs/administration/CLI_tasks/database.md
@@ -141,3 +141,21 @@ but should only be run if necessary. **It is safe to cancel this.**
     ```sh
     mix pleroma.database ensure_expiration
     ```
+
+## Change Text Search Configuration
+
+Change `default_text_search_config` for database and (if necessary) text_search_config used in index, then rebuild index (it may take time). 
+
+=== "OTP"
+
+    ```sh
+    ./bin/pleroma_ctl database set_text_search_config english
+    ```
+
+=== "From Source"
+
+    ```sh
+    mix pleroma.database set_text_search_config english
+    ```
+
+See [PostgreSQL documentation](https://www.postgresql.org/docs/current/textsearch-configuration.html) and `docs/configuration/howto_search_cjk.md` for more detail.
diff --git a/docs/configuration/howto_search_cjk.md b/docs/configuration/howto_search_cjk.md
new file mode 100644
index 000000000..d3ce28077
--- /dev/null
+++ b/docs/configuration/howto_search_cjk.md
@@ -0,0 +1,42 @@
+# How to enable text search for Chinese, Japanese and Korean
+
+Pleroma's full text search feature is powered by PostgreSQL's native [text search](https://www.postgresql.org/docs/current/textsearch.html), it works well out of box for most of languages, but needs extra configurations for some asian languages like Chinese, Japanese and Korean (CJK).
+
+
+## Setup and test the new search config
+
+In most cases, you would need an extension installed to support parsing CJK text. Here are a few extension you may choose from, or you are more than welcome to share additional ones you found working for you with the rest of Pleroma community.
+
+ * [a generic n-gram parser](https://github.com/huangjimmy/pg_cjk_parser) supports Simplifed/Traditional Chinese, Japanese, and Korean
+ * [a Korean parser](https://github.com/i0seph/textsearch_ko) based on mecab
+ * [a Japanese parser](https://www.amris.co.jp/tsja/index.html) based on mecab
+ * [zhparser](https://github.com/amutu/zhparser/) is a PostgreSQL extension base on the Simple Chinese Word Segmentation(SCWS)
+ * [another Chinese parser](https://github.com/jaiminpan/pg_jieba) based on Jieba Chinese Word Segmentation
+ 
+Once you have the new search config , make sure you test it with the `pleroma` user in PostgreSQL (change `YOUR.CONFIG` to your real configuration name)
+```
+SELECT ts_debug('YOUR.CONFIG', '安装和配置Nginx, ElixirとErlangをインストールします');
+```
+Check output of the query, and see if it matches your expectation.
+
+
+## Update text search config and index in database
+
+=== "OTP"
+
+    ```sh
+    ./bin/pleroma_ctl database set_text_search_config YOUR.CONFIG
+    ```
+
+=== "From Source"
+
+    ```sh
+    mix pleroma.database set_text_search_config YOUR.CONFIG
+    ```
+
+Note: index update may take a while.
+
+## Restart database connection
+Since some changes above will only apply with a new database connection, you will have to restart either Pleroma or PostgreSQL process, or use `pg_terminate_backend` SQL command without restarting either. 
+
+Now the search results of statuses should be much more friendly for your language of choice, the results for searching users and tags were not changed, as the default parsing/matching should work for most cases.