[FLASH-206] Refactor Process About Region Data #36
Conversation
solotzg
commented
Apr 4, 2019
•
solotzg
commented
- move the process about region data, such as write|default|lock cf to a new class RegionData.
- Decode the TiKV key & value at first.
- Optimize and refactor some code.
|
/run-integration-tests |
|
Maybe we should stop making the contributor agreement? It is a private project and those features won't likely be merged back to CH later. |
| auto next(RegionWriteCFDataTrait::Keys * keys = nullptr) | ||
| { | ||
| if (!found) | ||
| throw Exception(String() + "table: " + DB::toString(expected_table_id) + " is not found", ErrorCodes::LOGICAL_ERROR); |
There was a problem hiding this comment.
yes, that's a mistake
| } | ||
|
|
||
| private: | ||
| RegionPtr store; | ||
| std::unique_lock<std::shared_mutex> lock; | ||
|
|
||
| bool found; | ||
| RegionWriteCFData::Data::iterator write_cf_data_it; |
There was a problem hiding this comment.
Maybe storing RegionWriteCFData::Map is better?
There was a problem hiding this comment.
that's also a choice. if so, we need to store the expected table id because removeDataByWriteIt needs.
There was a problem hiding this comment.
It is kind of weird that caching an iterator here, since it won't change.
There was a problem hiding this comment.
Here, I just use this iterator as a pointer to seek TableID and RegionWriteCFData::Map
There was a problem hiding this comment.
I think it is better to cache TableID and RegionWriteCFData::Map & directly here, because an iterator is potentially means changing. And it should run faster since we can skip one pointer reference. But any way, it is your choice.
|
/rebuild |
2 similar comments
|
/rebuild |
|
/rebuild |
|
/run-integration-tests |
|
LGTM |
