Read and Write Apache Parquet

[simonaubertbd]

What's your use case?
Apache Parquet ( https://parquet.apache.org/ ) becomes more and more popular and I think it's like a standard now in the data community, this is no more restricted to Hadoop People. Qlik supports it, Alteryx will support it in the next release, even LibreOffice is working on it, etc, etc.
Why?
-opensource format
-fast

What's your proposed solution?
To have Orange Data Mining support Apache Parquet files for read and write.

Are there any alternative solutions?
To convert parquet files before but seems useless

[markotoplak]

Makes sense indeed. Orange lacks a robust and fast file format.

When I need fast reading, I resort to picked tables, but a robust format like that would be a big improvement.

[simonaubertbd]

Hello @markotoplak do you plan to add it in a future release ? Another point is that it would help the corporate and the research worlds communicate each other.

Best regards,

Simon

[markotoplak]

@simonaubertbd, first one on our list is HDF5 support, then we can also consider Parquet.

But if anyone does Parque we'll gladly merge it.

[simonaubertbd]

@zhuyubei Whoa, pretty impressive, congrats. Is there a pull request for that ? O_o

[zhuyubei]

@zhuyubei Whoa, pretty impressive, congrats. Is there a pull request for that ? O_o

But my code redesigned the whole "Table" class to store data as dataframe instead of numpy. It might not work with the current version.
Let me try to figure out how to implement a parquet reader in current Orange3 version.