Lightning: Support Custom File Iteration Algorithm when Constructing MDLoader #38598
Labels
component/lightning
This issue is related to Lightning of TiDB.
type/enhancement
The issue or PR belongs to an enhancement.
Enhancement
The
MDLoader
is the abstraction that stores the result of scanning the data source with file route rules and filters. However, in some situations, we don't need to iterate ALL the data source files to construct a complete MDLoader object. Instead, we just need to pick a few of key files in the data source to construct the information we want in MDLoader, so that the performance for generating anMDLoader
can be boosted.Currently, the file iteration logic cannot be customized, all we have is using the
ExternalStorage
interface, which provides aWalkDir
method to iterate ALL the files in the data source. We want to make the iteration logic customizable.The text was updated successfully, but these errors were encountered: