This seems useful to me as behavior to "opt in" to. We implemented an option to disable the storage of statistics in C++ but having an option to truncate large statistics also seems worthwhile.
On Mon, Oct 28, 2019 at 3:54 PM Xinli shang <[email protected]> wrote: > > Hi all, > > If you know any Parquet application requires the min/max value of > statistics to be real column value, please let me know. Also, If you have > any questions or concerns about this feature, feel free to raise comments > in the design document > <https://docs.google.com/document/d/1Mgb0dDXQJkgjouboDrGa9v06hWGJ0oPiwnmffXShQ_M> > or > in the Jira ticket PARQUET-1685 > <https://issues.apache.org/jira/browse/PARQUET-1685>. > > -- > Xinli Shang
