Sphek (coreseek) solution for table sharding

Regarding sphinx's table sharding process, I have the following questions: 1. For vertical table sharding, SQL _query directly uses join queries. But some people say this is not efficient and the indexing speed is slow? Some people say that it will not affect the efficiency. You can directly use join to solve the problem and want to explain the cause. 2. For horizontal table sharding, I have the following questions about sph1_table sharding:
1. For vertical table shards, SQL _query directly uses join queries. But some people say this is not efficient and the indexing speed is slow? Some people say it will not affect the efficiency. Simply use join to solve the problem and want to explain the cause.
2. For horizontal table shards, I originally imagined to write several more SQL _queries in the data source, but the prompt "warning: key 'SQL _ query' is not multi-value" is displayed when generating the index; it means repeat. I remember that the test was successful, but this time it didn't work. Do you remember me wrong?
If not, consider the following solutions:
① Use union to process SQL _query
② Use views to process SQL _query
③ The source in the Index comes from multiple data sources. Some people on the Internet say it can come from multiple data sources, but no specific configuration method is found.
(Configure a data source for each table shard. duplicate information is inherited)
Which of the following solutions is suitable?
Thank you. Thank you very much.

Reply content:

You have the following questions about how sphtasks Process Table shards:
1. For vertical table shards, SQL _query directly uses join queries. But some people say this is not efficient and the indexing speed is slow? Some people say it will not affect the efficiency. Simply use join to solve the problem and want to explain the cause.
2. For horizontal table shards, I originally imagined to write several more SQL _queries in the data source, but the prompt "warning: key 'SQL _ query' is not multi-value" is displayed when generating the index; it means repeat. I remember that the test was successful, but this time it didn't work. Do you remember me wrong?
If not, consider the following solutions:
① Use union to process SQL _query
② Use views to process SQL _query
③ The source in the Index comes from multiple data sources. Some people on the Internet say it can come from multiple data sources, but no specific configuration method is found.
(Configure a data source for each table shard. duplicate information is inherited)
Which of the following solutions is suitable?
Thank you. Thank you very much.

Sphenders are not mysql, but a search engine.
If the data size is large, it is recommended that the horizontal score be,
1) You can use the built-in distributed index (actually a proxy)
2) send multiple requests and merge the data.

Multiple data sources. When source is configured in indexer, multiple sources are used and separated by commas.