Optimize Hive Connector's file path filtering merchanism

Currently Hive Connector hive a merchanism to filter the file path to be scanned:

    private Iterator<HiveFileInfo> getLocatedFileStatusRemoteIterator(Path path, PathFilter pathFilter)
    {
        try (TimeStat.BlockTimer ignored = namenodeStats.getListLocatedStatus().time()) {
            return Iterators.filter(new FileStatusIterator(path, listDirectoryOperation, namenodeStats), input -> pathFilter.accept(input.getPath()));
        }
    }

The filtering is using PathFilter, which only have the path, dropped the information which is available in HiveFileInfo e.g. HiveFileInfo#isDirectory(), I’d suggest we do the filtering using HiveFileInfo:

public interface HiveFileInfoFilter
{
    boolean accept(HiveFileInfo file);
}

If you guys are ok with this, I’d be happy to contribute a patch to optimize this.

[xumingming]

@mbasmanova ok, I will submit a PR recently. After the filtering mechanism is optimized, our private fork of Presto will plugin in a piece of code in BackgroundSplitLoader, which will construct a HiveFileInfoFilter using the following logic:

List<String> patterns = <Get the filename pattern from Hive Metastore>
hiveFileInfoFilter = new FilePatternHiveFileInfoFilter(patterns);

And this pattern will be passed along into HiveFileIterator.

[mbasmanova]

@xumingming Got it. Thanks for explaining. Looking forward to a PR.