There's a number of JIRA issues, they may not fully capture what
you're after, so we can always open more

* ARROW-3246
* ARROW-3772
* PARQUET-800

On Tue, Jun 25, 2019 at 3:18 PM Keith Chapman <[email protected]> wrote:
>
> Hi Wes,
>
> Thank you for the quick response. In this case I wouldn't mind using the cpp 
> or the java library.
>
> For the parquet-cpp case can you point me to the jira you are referring to.
>
> Regards,
> Keith.
>
> http://keith-chapman.com
>
>
> On Tue, Jun 25, 2019 at 12:38 PM Wes McKinney <[email protected]> wrote:
>>
>> hi Keith -- I assume based on prior discussions you are referring to
>> the C++ library.
>>
>> No -- because the indices are encoded while the dictionary is being
>> built, the dictionary indices would have to be decoded and re-encoded
>> to be able to sort the dictionary. You might consider dictionary
>> encoding with a sorted dictionary using Apache Arrow and then helping
>> with the open JIRA issues about direct dictionary-to-Parquet writes
>> (which would preserve your pre-computed dictionary order).
>>
>> Thanks
>> Wes
>>
>> On Tue, Jun 25, 2019 at 1:51 PM Keith Chapman <[email protected]> 
>> wrote:
>> >
>> > Hi all,
>> >
>> > When writing a parquet file is there a way to force the dictionary to be
>> > written out sorted?
>> >
>> > Regards,
>> > Keith.
>> >
>> > http://keith-chapman.com

Reply via email to